Pricing psychology: what real buyers did with real prices, and what to change

Part 1Is your price too low? What happened when a company randomized its own prices

Quite possibly, especially if your buyers are businesses or your quality is hard to judge before they buy. When a recruiting service doubled its monthly price for small-business buyers, it lost only about one buyer in four (Dubé & Misra, 2023). In other real markets, the price itself tells buyers what quality to expect (Erdem et al., 2008).

What happened when ZipRecruiter doubled its price?

In 2015, the online recruiting service ZipRecruiter let chance decide which price a new customer saw at its paywall. Dubé and Misra (2023) followed 7,867 prospective buyers: small employers, typically with fewer than 50 staff and 1-3 jobs to fill. Each one saw one of 10 monthly prices, from $19 to $399 (Dubé & Misra, 2023).

Moving from the $99 status quo to $199, a 100% increase, cut conversions by only 25% (Dubé & Misra, 2023). Put plainly, doubling the price lost about one buyer in four, and every buyer who stayed paid twice as much. The researchers' demand model put the profit-maximizing price 230% above the status quo, a chance to raise profits by 55% (Dubé & Misra, 2023). The company had been charging far less than its buyers would bear.

ZipRecruiter acted on it. Shortly after the experiment, it raised its price to $249 for good and kept it there until at least November 2020 (Dubé & Misra, 2023). The higher prices did cost something later, because retention also fell in the pricier cells (Dubé & Misra, 2023). But so few subscribers stayed long at any price that first-month profit outweighed the lost renewals (Dubé & Misra, 2023).

Even with real money and random assignment, the answer was a range, not a number. Among the prices tested, monthly revenue per prospective buyer peaked at $399, but the best price could plausibly sit anywhere from $249 to $399, or above (Dubé & Misra, 2023). That is the honest state of high-ticket pricing evidence. Almost no field study prices a $2,000 to $20,000 program, and a business subscription like this one is the closest evidence there is.

"We want to serve fewer customers and we want to do so at a higher price point."

Jon Preston, founder of Atomic Funnels

The ZipRecruiter data reward exactly that choice, in one market with one product: fewer buyers, a higher price, more profit (Dubé & Misra, 2023). Two conditions come attached. Revenue, not buyer count, has to be the scorecard. And the offer has to live up to the expectations its price sets.

A lower purchase rate can still mean more profit

Gneezy et al. (2012) ran a small, vivid version of this on tour boats. Passengers on 20 cruises per condition could buy a souvenir photo for $5, for $15 or for whatever they chose to pay, and all were told the regular price was $15 (Gneezy et al., 2012). At $5, 64% bought, against 23% at $15 (Gneezy et al., 2012). Under pay what you want, 55% bought (Gneezy et al., 2012). The cheap price pulled in far more buyers.

Yet profit per photo taken barely moved: $3.45 at $15, $3.50 under pay what you want and $3.20 at $5, and the gap between the first two was not statistically significant (Gneezy et al., 2012). The $15 price sold to far fewer people and still earned slightly more per photo than $5. Judge a price test by conversion rate alone and you will crown the cheapest price almost every time.

Does a higher price make buyers expect more?

Price also changes what buyers think they are getting. That is the psychology of pricing at its most basic. From field experiments at a winery, Gneezy et al. (2014) propose that a higher price sets higher expectations, which become the yardstick: when the wine meets them, a higher price goes with higher perceived quality. When a high price came with mediocre wine, visitors rated it worse than the same wine at a low price (Gneezy et al., 2014).

The same signal shows up in years of grocery data. Modeling households' repeated ketchup purchases, Erdem et al. (2008) found price was an important quality signal, and frequent price cuts went with significant damage to brand equity. How often a brand advertised also signaled quality, but less than price did (Erdem et al., 2008).

A meta-analysis (a study that pools the results of many earlier studies) by Rao and Monroe (1989) found a positive, statistically significant link between price and perceived quality for consumer products. The store's name, by contrast, made only a small, nonsignificant difference (Rao & Monroe, 1989). Sellers seem to sense all this.

Stiving (2000) found firms use rounder prices for higher-quality products, more so where buyers can't judge quality before they buy. Coaching, consulting and courses sit squarely in that category. That matters when you pick a price ending.

What to change in your funnel

The practical move is a three-price test on the same offer. Say you sell a $3,000 coaching program: run it at $2,000, $3,000 and $4,500 (illustrative numbers) behind the same VSL and order form. Judge each price by revenue per visitor after refunds. The ZipRecruiter researchers judged the best price by monthly revenue per prospective buyer, not by conversions alone (Dubé & Misra, 2023).

The winery result is the warning label. A higher price raises the bar, so the VSL or webinar pitch has to clear it with proof, specifics and outcomes before the number appears (Gneezy et al., 2014). Track refunds and churn too, since retention fell at higher prices even when profit rose (Dubé & Misra, 2023).

Split traffic evenly across the three prices at the order form and keep everything else identical. Run whole weeks until each arm hits its planned sample.

The price level is only the first reference point buyers use. Next comes whether they see that number in full upfront or meet extra fees at checkout. Later comes how existing customers react when you raise it.

Part 2Should you show the full price upfront or add fees at checkout?

In a large randomized test at StubHub, hiding fees until checkout made more money (Blake et al., 2021). That practice, showing a low price first and adding fees later, is called drip pricing. So the tactic is real. The same test shows its cost: far more buyers walked away at the last step (Blake et al., 2021), and in other studies buyers read a split price as a seller trying to persuade them (Greenleaf et al., 2016).

What happened when StubHub hid its fees?

Picture two people shopping for the same concert tickets on the same day. One sees prices with fees included from the first page. The other sees fees only at checkout. In August 2015, StubHub ran exactly that comparison on several million visitors, split at random, as Blake et al. (2021) describe. Over ten days, the checkout-fee group spent almost 21% more in revenue than the all-in group (Blake et al., 2021).

They also bought more often. The transaction rate rose 14.1% over the experiment, and within a single visit checkout-fee buyers were 12.43% more likely to buy (Blake et al., 2021). For a seller, that means more buyers and more revenue from the same traffic.

The click data show where the money came from. Buyers who saw the lower, fee-free prices were almost 19% more likely to click into a ticket, then almost 45% less likely to finish once the fees appeared (Blake et al., 2021). The low number pulled far more people in, and many bailed when the real total showed up. Enough stayed to win on revenue. Buyers also drifted toward better seats: after the site-wide switch, sellers listed more row-A tickets to meet that demand (Blake et al., 2021).

Were new visitors simply fooled once? It seems not. During the test, checkout-fee buyers spent almost 6% more per purchase, and once everyone saw checkout fees, the two groups spent the same (Blake et al., 2021). What mattered was the price on screen at the moment of choosing, not a lesson the all-in group carried with it. An earlier 2012 test across 99 Major League Soccer games pointed the same way: buyers were 13% less likely to buy tickets to games shown with upfront fees, a borderline result (p = 0.076) (Blake et al., 2021).

Buyers know the fees exist. They just don't add them up.

A supermarket shows why. For three weeks, Chetty et al. (2009) put tax-inclusive price tags on cosmetics, hair accessories and deodorants in one store of a grocery chain, then compared sales with untagged products and two other stores. Posting the all-in price cut quantity sold by 7.6 percent (Chetty et al., 2009). These shoppers were not ignorant. In a survey of 91 customers at that store, the median person correctly said whether 7 out of the 8 products were taxed and named the tax rate within 0.5 percentage points (Chetty et al., 2009).

They knew the tax. They just didn't do the sum in the aisle. With tax left off the tag, only 18% of shoppers named a price within 25 cents of the real total, against 75% when the tag also showed the tax-inclusive price (Greenleaf et al., 2016). That is fewer than one in five shoppers knowing their real price, versus three in four.

Researchers call a base price plus a surcharge partitioned pricing, and experiments reviewed by Greenleaf et al. (2016) show how people handle it. In one set of studies, 23% ignored the surcharge when recalling the total, 54.8% used a rough shortcut, and only 21.9% did the math to within 5% of the true figure (Greenleaf et al., 2016). So about one buyer in five knew what they would actually pay. Across studies, between 12.2% and 35.6% dropped the surcharge entirely (Greenleaf et al., 2016). Accuracy fell from 49% with two price parts to 23.3% with nine (Greenleaf et al., 2016): every extra line item makes the total harder to see.

When does a low base price plus a fee pay off?

eBay sellers found the same lever. Hossain and Morgan (2006) auctioned matched pairs of music CDs and Xbox games to real bidders, changing only how the price was framed. A low opening bid plus a high shipping charge beat a higher opening bid with free shipping in 16 of 20 matched pairs (Hossain & Morgan, 2006). That held only when the minimum total was a small fraction of the retail price. When it was a relatively high fraction, the framing made no systematic difference (Hossain & Morgan, 2006).

Low openings went with earlier first bids, and the auction with the earlier bid earned more in 29 of 38 pairs (Hossain & Morgan, 2006). The researchers suggest bidders keep separate mental accounts for price and shipping, or ignore shipping altogether, though the study did not test that directly (Hossain & Morgan, 2006). A second set of auction experiments by Brown et al. (2010) found that raising shipping charges boosted revenue when the charges were hidden. Sellers with low shipping charges did better disclosing them (Brown et al., 2010). Neither result came from drawing more bidders (Brown et al., 2010).

Taken to the extreme, this turns into a race to the bottom. On one price-comparison site, computer-parts retailers commonly listed a memory module at $1 and revealed a $40 shipping and handling fee at checkout (Ellison & Ellison, 2009).

A split price can backfire on a premium offer

The gains depend on buyers not looking closely, and many do look. In one study in the review, almost half of people believed a split price had more persuasive intent than an all-in price, against 13% who thought the opposite (Greenleaf et al., 2016). Split pricing lowered recalled totals only when the surcharge sat in small font (Greenleaf et al., 2016). That is a poor foundation for an offer sold on trust.

The reverse move has better evidence behind it. In online grocery shopping, free shipping lifted purchases more than an equal discount off the base price (Greenleaf et al., 2016). A named extra given away can do more work than the same money taken off the headline number.

What this means for your order form

No one has tested drip pricing on a coaching program, a course or a software subscription, and there is no field or survey evidence here on business buyers. The closest evidence comes from tickets, groceries and auction items, where a buyer who abandons at checkout costs the seller a few dollars. Your stakes are different. Say you sell a $5,000 program: the buyer who meets a surprise $250 processing fee on the order form is the buyer your webinar spent an hour earning.

So show one all-in price everywhere. The VSL, the pricing page and the order form should name the same number, with any tax or fee stated beside it at the same size. If you want the lift that split framing promises, test a free bonus (a named onboarding session, a done-with-you setup) against an equal dollar discount instead of splitting the price.

The StubHub click data point to the idea that runs through all of pricing psychology. The first number a buyer sees becomes the reference point for every number after it, which helps explain why the fee-free price drew so many more clicks (Blake et al., 2021). Make that first number the true price, presented after the value, not a teaser that gets corrected at checkout.

Part 3Does charm pricing ($997 instead of $1,000) actually work?

Charm pricing, the most famous move in psychological pricing, works with real buyers, but less reliably than its reputation suggests. In a small catalog pilot, prices ending in 9 sold approximately 40% more (Anderson & Simester, 2003). In the largest test, the lift was 7% (Anderson & Simester, 2003). Some tests found no statistically significant difference (Schindler & Kibarian, 1996). For a premium offer, a round price is worth testing against the 9.

What happened when a catalog changed only the last digit?

Picture a regular customer of a women's clothing catalog opening the new issue. Anderson and Simester (2003) mailed different versions of the same catalog to separate random groups of active customers, all of whom had bought in the previous 18 months. In a pilot with four dresses, the $9 ending raised demand by approximately 40% (Anderson & Simester, 2003). A $10 price gap between the two other versions, meanwhile, made effectively no difference in demand (Anderson & Simester, 2003). These shoppers reacted more to the last digit than to ten real dollars.

Bigger tests gave smaller numbers. In a second study, a $9 ending lifted units sold by 22% for new items vs 10% for established ones (Anderson & Simester, 2003). The third and largest test covered 308 items and 90,000 customers per catalog version (Anderson & Simester, 2003). There, the average item sold 16.9 units per version and gained 1.2 units with a $9 ending, a 7% lift (Anderson & Simester, 2003). Trust that figure most: a real but modest bump, not a windfall.

Two details matter if you are pricing an offer. A "Sale" cue, a regular price shown next to a sale price, lifted demand more than the $9 ending did (Anderson & Simester, 2003). And when one catalog manager tried $9.50 endings, buyers did not respond as they did to $9; if anything, demand fell (Anderson & Simester, 2003). A price that merely looks precise is not the same as a 9 ending.

Did other retailers see the same lift?

Schindler and Kibarian (1996) ran a similar test at a national women's clothing retailer, mailing three versions of a clearance catalog to 30,000 existing customers each. The 99-ending version produced 8.0 percent more sales volume than the 00-ending version (Schindler & Kibarian, 1996). But it drew 968 buyers vs 921, a gap of 47 that was not statistically significant (Schindler & Kibarian, 1996). So the gain showed up in how much people bought, not clearly in how many people bought.

Prices ending in 88 did nothing: 923 buyers, two more than the round version (Schindler & Kibarian, 1996). The researchers' explanation is simple. If most buyers skip the last two digits, the price feels 2.1 percent cheaper, and an 8.0 percent sales gain is a normal response to a cut that size (Schindler & Kibarian, 1996).

Older tests were mixed. A south Florida department store chain alternated 98 and 00 endings weekly on 11 products and found no significant sales difference (Schindler & Kibarian, 1996).

Is left-digit bias real outside catalogs?

Left-digit bias means judging a number mostly by its first digit. It shows up clearly at the grocery shelf. Strulov-Shlain (2023) studied 3,500 products sold by 25 US chains. Shoppers treated a one-cent rise from a ninety-nine-cent ending as if it were more than a twenty-cent increase (Strulov-Shlain, 2023). In other words, tipping into a new dollar hurt far more than the penny itself.

Every chain used ninety-nine endings, but priced as if the bias were much smaller than it is (Strulov-Shlain, 2023). That left an estimated 1-4% of potential gross profit unclaimed (Strulov-Shlain, 2023).

An older scanner study of tuna and yogurt buyers found two effects. One was a level effect, where buyers treat the price as lower (Stiving & Winer, 1997). The other was an image effect, where buyers read meaning into the last digits (Stiving & Winer, 1997). Both point the same way: the ending changes how a price feels, not only how it reads.

Buyers read other numbers by their left digit too. At US wholesale used-car auctions, cars showing 79,900 to 79,999 miles sold for approximately $210 more than cars at 80,000 to 80,100 miles (Lacetera et al., 2012). Yet they sold for only $10 less than cars at 79,800 to 79,899 miles (Lacetera et al., 2012). Buyers were paying for the first digit, not the miles. The evidence points to final car buyers, not the professional dealers bidding at auction, as the ones swayed (Lacetera et al., 2012). This is observational data about mileage, not prices, so it shows left-digit inattention in costly purchases, not that 9 endings work at high prices.

Where does charm pricing stop paying?

Not every setting shows the effect. In the supermarket tax test covered under hidden fees, demand did not fall more when tax pushed a price over a whole dollar, though that estimate was imprecise (Chetty et al., 2009).

ZipRecruiter rounded its personalized small-business subscription prices to $9 endings (Dubé & Misra, 2023). The researchers report that this rounding barely changed predicted profit, but the study did not test endings themselves (Dubé & Misra, 2023). So it tells you little either way.

Should a premium offer use a round price?

Round numbers send a message, but which message is unclear. A price-signaling model by Stiving (2000) concluded that endings are not necessarily quality signals in themselves. Yet firms tended to pair round prices with higher quality, more so where buyers can't judge quality before buying (Stiving, 2000). That describes what sellers do. No study here shows buyers reading a round price as a sign of quality.

Where prices get negotiated, round numbers read differently. On an online marketplace, items listed at multiples of $100 drew offers 8-12 percent lower but were 15-25 percent more likely to sell (Backus et al., 2019). Buyers also looked at round listings more often, as if the round number told them something about the seller (Backus et al., 2019). A round price may say "flexible" as easily as "premium," and that matters when the price comes up on a sales call.

No one has tested a just-below price against a round one on a coaching program, course or consulting offer. The closest evidence comes from catalog clothing, groceries, used cars and online marketplaces. No study here tests price endings on business buyers choosing software or services. The nearest clue: professional car dealers seemed less swayed by the left digit than final customers (Lacetera et al., 2012).

What should you change on the order form?

Treat the ending as one more reference point, alongside the anchor and the fees. For a lower-ticket entry offer, a 7 or 9 ending is a cheap test. For a premium program, test round against just-below instead of defaulting to either: say $3,000 vs $2,997. If the price is negotiable on a sales call, expect a round list price to invite lower counteroffers, as the anchoring section explains.

Keep the number identical across the VSL, the webinar slide, the sales page and the order form. If you want a stronger cue than the ending, a credible regular-vs-now comparison did more in the catalog tests, with the cautions covered in the discounts section. Judge any test by revenue per visitor and refunds, not clicks.

Part 4Price anchoring: does the first number buyers see set what they pay?

The first price on the table pulls what buyers offer and pay toward it, in real markets with real money. That makes anchoring one of the most useful ideas in pricing psychology. The pull is weaker and less predictable than the famous lab demonstrations, though, and several of the lab's favorite ways to strengthen it did nothing when people paid for real.

What happens when millions of buyers haggle against a list price?

On eBay, a seller can post a Buy It Now price and invite offers below it, through a format called Best Offer. Backus et al. (2020) analyzed 25,458,516 of these bargaining threads between real buyers and sellers, covering everyday goods with an average list price of $95. That is an enormous record of real price negotiations.

Buyers opened low. The average first offer was $86.60, which was 61% of the list price (Backus et al., 2020). Across all listings, items sold at an average of 83% of list. When bargaining actually happened, the average sale price came down to 73% of list (Backus et al., 2020). In plain terms, a buyer who haggled typically paid a little under three quarters of the asking price.

The most common first offer was exactly half the list price (Backus et al., 2020). Deals closed 45% of the time (Backus et al., 2020), and final prices landed well inside the gap between the seller's list price and the buyer's opening offer (Backus et al., 2020). That is anchoring in the wild. The list price becomes the reference price, the ceiling every negotiation runs down from. These are observational data, so they show the pattern, not what a different list price would have caused.

Where a buyer landed inside that range went with who they were. Patient buyers, the ones who picked the slowest shipping, paid 7.2 percentage points of the reference price less on used goods (Backus et al., 2020). Each 1% more buying experience went with a final price 1.8 percent of the reference price lower (Backus et al., 2020). Experienced sellers got 1.4 to 6.4 percent more, though that edge disappeared once buyer patience was taken into account (Backus et al., 2020). The anchor sets the range. Patience and practice decide the spot.

Does a round asking price invite lower offers?

Round listings drew lower offers and sold more often. In millions of online bargaining interactions studied by Backus et al. (2019), items listed at multiples of $100 went with offers 8-12 percent lower but were 15-25 percent more likely to sell. So a round price meant less money per deal but more deals. Sellers of round-number listings were also more likely to accept a similar offer (Backus et al., 2019), and buyers were more likely to click into those listings (Backus et al., 2019).

Buyers seem to read a round number as a sign the seller will bend, and they adjust their offers to match. These are observational data, so it is not clear a round price alone would buy you a faster yes. Test it. The charm pricing section covers price endings when no one haggles.

Is anchoring as strong with real money as in the lab?

It is real, but smaller and fussier. Jung et al. (2016) ran sixteen field studies in which real customers in commercial settings paid what they wanted (N = 21,997), plus four hypothetical studies (N = 3,174). In the first two studies, anchoring showed up whether or not money was at stake (Jung et al., 2016). Low anchors pulled payments more than high ones (Jung et al., 2016).

What mattered most was where an anchor sat among the payments people actually made, not how far apart the numbers looked on paper (Jung et al., 2016). Moves that boost anchoring in lab studies, like precise numbers or 'most people pay $X' lines, did nothing when people paid real money (Jung et al., 2016). Don't count on them on your order form. Only in the hypothetical studies was anchoring as big and reliable as textbooks promise (Jung et al., 2016). The tour-boat photo test in the price-level section pulled the same lever: every passenger was told the regular price was $15 (Gneezy et al., 2012).

"We want to lean on behavioral psychology and information retention studies to create pages that are statistically proven to elicit consumer behaviors that generate revenue for our business and lead to our desired outcomes."

Jon Preston, founder of Atomic Funnels

The anchoring evidence backs the first half of that and adds a condition to the second. The effect shows up with real buyers (Jung et al., 2016), but a lab-sized result does not carry over to a page automatically (Jung et al., 2016). "Statistically proven" has to mean proven on your own traffic, which is what the testing section covers.

Which number on the page becomes the anchor?

Whichever one you make biggest. When grocery shoppers judged a half-pound product, a prominent $2.00 per pound looked like worse value than $1.00 each (Whitley et al., 2025). For a two-pound product it flipped: $2.00 per pound beat $4.00 each (Whitley et al., 2025). Shoppers anchored on the face value of the most visible number, whatever unit it was in (Whitley et al., 2025).

On a pricing page, the headline figure (total, per month, per session) is your anchor. Choose it on purpose. The payment plans section shows what installment framing does to spending.

Do business buyers anchor too?

Yes, and so do the people selling to them. Researchers looked at transaction data from business markets where salespeople and customers set prices through personal negotiation. Reference prices shaped both how much customers bought and the final price (Bruno et al., 2012). Salespeople carried their own reference prices, and those moved the deal price too (Bruno et al., 2012). The study doesn't say how big the effect was, and no study here measures how a procurement team's counteroffer moves with a vendor's opening number.

What to change in your pitch and on the call

No one has tested anchoring on a webinar pitch or a high-ticket sales call. The closest evidence is the eBay record and the business transaction data, which suggest testing three things. First, name the price before the buyer does, after you have built the value, so your number becomes the reference rather than theirs. Second, on calls, expect counteroffers to start far below list and settle between the two openings (Backus et al., 2020). Decide your floor and your trade, such as a payment plan or smaller scope, before you dial.

Third, treat every number in the pitch as a candidate anchor. Say you sell a $6,000 coaching program. A passing mention of a cheap competitor may pull harder than a big "total value" figure, since low anchors pulled more than high ones in the field (Jung et al., 2016). The other options on your pricing page are anchors too, which is where decoys come in next. The value stack gets its own reckoning in the discounts section.

Part 5Decoy pricing and three-tier offers: does the middle option really win?

Decoys do move real purchases, in stores and online; the pull of the middle option comes mostly from experiments, not sales data. But the field evidence is thinner than pricing psychology blogs suggest, and the effect weakens where most pricing tables live: a cheap plan set against a better, pricier one. So build tiers that differ on features buyers can compare, then test whether a third tier earns its place.

What happened when real shoppers met a decoy?

Picture an option nobody should pick. It is clearly worse than one choice but not obviously worse than another. That is a decoy, and its only job is to make its near neighbor look good.

Doyle et al. (1999) began with buying tasks where people said what they would buy, or what most people would buy, and found the shift there. Then they took it into a store. There, the decoy shifted real in-store purchases toward the option it flattered (Doyle et al., 1999).

Another of their studies turned up something stranger. The decoy still shifted preferences when people could see it was unavailable (Doyle et al., 1999). In other words, a sold-out option can still shape what someone buys. The researchers judged the effect sizeable and of practical significance (Doyle et al., 1999), but their summary reports no effect sizes, so how big it was is unknown.

Two decades later, Wu and Cosguner (2020) looked at an online diamond retailer. Its shoppers weigh stones on several features before spending serious money. The sales data showed those shoppers were still swayed by a decoy (Wu & Cosguner, 2020). These are observational data on expensive goods bought online, which is close to the considered purchase a founder sells. Here too, the summary gives a direction rather than a size.

Does the middle option really win?

The pull toward the middle has its own name: the compromise effect, or extremeness aversion. Neumann et al. (2016) ran a meta-analysis, a study that pools the results of many studies. Across 142 experimental observations, people chose middle options significantly more often than the extremes (Neumann et al., 2016). The summary does not say who took part, and these were experiments, not sales data.

That average hides a wide spread. The pull varied by up to three times its average size, depending on how a study was designed (Neumann et al., 2016). It shrank when options traded price against quality, which is exactly the layout of good-better-best pricing (Neumann et al., 2016). It also shrank when studies compared two-option with three-option sets (Neumann et al., 2016).

Other conditions made it stronger. The pull grew with more features to trade off, non-numeric features, premium top-end extensions and practical products (Neumann et al., 2016).

Even the scorecard changes the answer. The three common ways of measuring the effect can give quite different sizes, and even opposite conclusions about what strengthens it (Neumann et al., 2016). So when someone tells you the middle tier always wins, ask how they counted.

A cheap tier earns its keep by feeding upgrades

A revealing real-market look at tiers comes from Ellison and Ellison (2009). They studied computer-parts retailers selling memory through Pricewatch, a price-comparison site. Its users were hobbyists, gamers, IT support staff and small computer-store owners. Those shoppers sorted by the cheapest low-grade module, so a retailer's rank on that list drove its traffic.

When a site slipped from first to seventh on the cheap-memory list, its sales of medium-quality memory were 65% lower (Ellison & Ellison, 2009). Its sales of high-quality memory were 43% lower (Ellison & Ellison, 2009). The cheap listing, it turns out, was the front door for upgrade sales.

Top rank also changed who walked in. When a site sat first for 256MB memory, 63% of its unit sales were the cheap grade, against 35% on days it sat tenth (Ellison & Ellison, 2009). Bargain hunters flock to the top listing.

Yet enough of them traded up to matter. The average markup (price over cost) on memory was about 12%, against the 2.5% to 4% you would predict from the cheap module alone (Ellison & Ellison, 2009). These are observational data about entry tiers, not decoys. Still, they show what a tier ladder is for: margin on the better options.

Should pricing tiers show features as numbers or images?

How you show features may matter as much as the decoy itself. In a lab study with 240 university recruits in the U.S. and Asia choosing TVs, a decoy lifted the target TV's share from 33% to 57% when picture quality was a number (Frederick et al., 2014). When quality was a photo, the share fell from 53% to 35%, a drop just short of statistical significance (Frederick et al., 2014). The compromise meta-analysis points the other way: non-numeric features strengthened the middle pull (Neumann et al., 2016). So the evidence on numbers versus descriptions is mixed.

The practical upshot: if your tiers differ on things buyers must imagine, show them as numbers (calls per month, hours, seats) where you can. Then test that against a plain-language version.

No field or survey evidence here tests decoys or three-tier menus on business buyers. A procurement team choosing a software plan is untested ground.

What to change on your pricing page and offer stack

Buyers will compare your tiers anyway, so set up that comparison on purpose. Say you sell a $3,000 coaching program (illustrative numbers). Offer a $1,500 self-study tier, the $3,000 core program and a $6,000 tier with one-to-one calls.

List every tier against the same rows, such as calls, response time, templates and community access. That way buyers compare like with like.

A pure price-against-quality menu is where the middle pull weakens. So give the top tier something genuinely different, not just more of the same. The top price also works as a reference point, the same job the first number does in the anchoring section. And it tells buyers what quality to expect, as the price-level section showed.

Then split-test two tiers against three. Judge the winner by revenue per visitor, not by which tier gets picked. Next comes the other comparison sellers lean on: the struck-through total value.

Part 6Do discounts, sale signs and 'total value' stacks raise sales or cheapen the offer?

A believable sale cue can lift sales without touching the price, and a well-timed discount email raises spending far beyond the people who use the code. But sale signs lose their power when overused, and deep discounts can backfire when buyers can't judge quality for themselves. In 70 email experiments, discount offers raised spending 37.2%, and 90% of that gain did not come through redeeming the offer (Sahni et al., 2017).

Same dresses, same prices, more sales: what did the sale sign do?

Picture a household leafing through a women's clothing catalog. Anderson and Simester (2001) worked with a national catalog of this kind and mailed two versions to two samples of 67,000 US households, past and prospective customers alike. In one version, three of eight dresses carried a sale sign. In the other, all eight did. Every price stayed the same. The five dresses that gained a sign sold 57% more (Anderson & Simester, 2001).

That is more than half again as many dresses, sold at an unchanged price. Pulling together catalog and grocery data, the researchers concluded that a sale sign can move demand by more than 50% with no price change (Anderson & Simester, 2001). In other catalog tests, a "Reg $X SALE $Y" format lifted demand more than a $9 price ending did (Anderson & Simester, 2003). This is pricing psychology at its plainest: the format supplies a reference price, the number buyers hold the actual price against.

Signs wear out, though. In grocery scanner data from about 80 stores over 169 weeks, total sales of frozen juice, toothpaste and tuna peaked when about 25% of items carried sale signs, and fell significantly once more than 30% did (Anderson & Simester, 2001). That is store data, not an experiment, so read it as a pattern. When everything is on sale, the sign stops meaning anything.

Discount emails work even when nobody redeems them

An online ticket resale platform ran 70 randomized experiments, emailing discount offers to some customers and not to others. The offers raised average spending by $3.03 during the promotion window, a 37.2% increase (Sahni et al., 2017). The surprise: 90% of that gain did not come through redeeming the offer (Sahni et al., 2017). Nine in ten of the extra dollars arrived some other way.

The effect also outlasted the offer. Spending stayed $1.55 higher in the week after it expired (Sahni et al., 2017). Offers for one ticket genre raised spending in genres the email didn't cover (Sahni et al., 2017). The researchers concluded that emailed offers work partly as advertising: a reason to come back and browse (Sahni et al., 2017).

Who got the email mattered too. Customers who had spent more in the past responded more, and the effect was significantly larger among people who hadn't bought in the previous year (Sahni et al., 2017). So the discount email's real job is to win back attention from people who have drifted. Subsidizing people who would have bought anyway is the expensive side effect.

Do deeper discounts sell more?

In the short run, often yes. About 18,000 mobile-phone subscribers near two rival movie theaters received randomized text-message coupons. At the nearby theater, raising the discount from 20% to 40% off more than doubled the share who bought a ticket, from 2.3% to 5.2% (Dubé et al., 2017). For the rival theater farther away, going from 40% to 60% off lifted purchases from 0.35% to 2.8% (Dubé et al., 2017). The shares are small, but the jumps are large.

Movie tickets are a known product with a known regular price. Nobody wonders whether a cinema seat got worse because the coupon got bigger. That is the setting where deep discounts behave.

A deep discount cheapens the offer when buyers can't judge quality

When buyers can't judge quality, a big discount starts to look like a confession. Across 19,978 Groupon deals for local services such as restaurants, spas and entertainment, an extra 10 percent of discount on top of the average 55.6 percent went with sales dropping 0.63-4.60 percent (Cao et al., 2018). These are deal records, not a randomized test. Still, the pattern is the one you'd expect if discounts signal doubtful quality.

Reassurance softened the drop. It was weaker for deals that had already sold more than 300 vouchers, where buyers could see that others had bought (Cao et al., 2018). It was stronger in cities with higher incomes and more college graduates (Cao et al., 2018). It was also stronger for deals with more Facebook fans and better reviews, where the discount clashed with every other quality cue (Cao et al., 2018).

Take away the uncertainty and the picture changes. For Groupon Goods, physical products with guaranteed returns, the direction flipped: deeper discounts went with more sales (Cao et al., 2018).

In a lab study with students, perceived quality uncertainty climbed sharply only once the discount passed 60% (Cao et al., 2018). The winery visitors in the price-level section showed the same logic: price tells buyers what to expect (Gneezy et al., 2014). A course, a coaching program or a consulting engagement sits squarely on the uncertain side.

What do discounts do to customers over time?

Catalog field experiments varied how deep a current promotion went, then tracked what customers bought afterward. Deeper discounts increased later purchases by first-time customers but reduced later purchases by established customers (Anderson & Simester, 2004). So short-run numbers overstate the gain from discounting to established customers, which pushes prices too low (Anderson & Simester, 2004). They understate it for first-time buyers (Anderson & Simester, 2004).

Promotions also train buyers. In household panel data on nine grocery brands across 29-33 quarters, loyal and nonloyal shoppers alike grew more price sensitive over time (Mela et al., 1997). Price promotions went with a long-run rise in price sensitivity almost four times larger for nonloyal shoppers than for loyal ones (Mela et al., 1997). Promotions train bargain hunters more than loyal buyers. That is one more reason to aim discount emails at lapsed leads rather than at your best customers.

A 'total value' stack is a crossed-out price, so keep it honest

A "total value" line followed by a lower "yours today" price does the same work as "Reg $X SALE $Y": it sets the reference price. A meta-analysis (a study that pools the results of many studies) of advertised reference prices described comparative price advertising as a powerful tool with a strong opportunity for deception that needs careful management (Compeau & Grewal, 1998). Reviews also find buyers react more sharply to a price above their reference point than to one below it (Kalyanaram & Winer, 1995).

No one has tested a value stack on a webinar pitch or a high-ticket sales page. The closest evidence is the catalog sale sign and this reference-price research, and none of it covers business buyers. Treat the stack as a claim your buyer can check, not a number you can inflate.

Say you sell a $3,000 program (illustrative numbers). A discount strategy then comes down to four moves. Build the value stack only from components with a real standalone price. Show one crossed-out price on the offer and order form, not banners on every page, and only a regular price you have actually charged for a meaningful period.

Keep the core price stable rather than permanently "on sale." Save discount emails for lapsed leads, and judge them by total revenue over the following months. Cutting the price that recent buyers paid in full carries its own cost, covered in the price-increases section.

Part 7Do payment plans and subscriptions change what buyers spend?

Payment plans do change what buyers spend, but under narrower conditions than most sellers assume. In real purchase data and experiments, shoppers spent more with buy now pay later (Ashby et al., 2025), and in the experiments, showing the installment price drove the extra spending (Ashby et al., 2025). It did nothing when people paid by card (Ashby et al., 2025). Subscriptions add a second twist. People may choose a plan for the person they hope to become, not the person who actually shows up, as the gym evidence below suggests (Della Vigna & Malmendier, 2006).

Why do gym members pay for visits they never make?

Picture someone signing a gym contract, sure that this time the habit will stick. Della Vigna and Malmendier (2006) studied thousands of people in exactly that spot. The researchers had real contracts and attendance logs for a panel of 7,978 members (Della Vigna & Malmendier, 2006). Most were non-student adults in their early thirties (Della Vigna & Malmendier, 2006).

Members on an unsubsidized monthly flat-rate contract paid over $17 for each visit they actually made (Della Vigna & Malmendier, 2006). Annual members paid over $15 per visit (Della Vigna & Malmendier, 2006). A 10-visit pass sat on the menu the whole time. Yet only 20 percent of monthly members and 24 percent of annual members ended up paying less than $10 per visit (Della Vigna & Malmendier, 2006). For most members, the plan that felt like a deal was the expensive way to use the gym.

The researchers think overconfidence explains the gap. A separate small survey asked 48 gym-goers at a mall how often they expected to go next month. Their forecasts averaged 9.50 visits (Della Vigna & Malmendier, 2006). That was more than twice the 4.17 visits the clubs' members actually averaged (Della Vigna & Malmendier, 2006). The two numbers come from different people, so treat this as suggestive. If the researchers are right, people bought the habit they planned, not the one they had.

Then the membership outlasted the habit. Monthly members let 2.29 full months pass between their last visit and cancelling, paying $185 in that stretch (Della Vigna & Malmendier, 2006). One in five, 20 percent, waited at least 4 months (Della Vigna & Malmendier, 2006). Members who paid more per visit also tended to take a little longer to cancel, a modest link (Della Vigna & Malmendier, 2006). Over a membership, the average monthly member lost $698 compared with buying 10-visit passes (Della Vigna & Malmendier, 2006). That was 47.87 percent of the $1,517 they spent (Della Vigna & Malmendier, 2006). Put plainly, nearly half of what these buyers paid was more than the same visits would have cost on passes.

Two details keep this from being a simple story about foolish buyers. Annual members did attend 9.5 percent more in months 2-4, so some people sorted into plans sensibly (Della Vigna & Malmendier, 2006). Yet monthly members were 18.1 percent more likely to still be members after 14 months (Della Vigna & Malmendier, 2006). They had paid extra for the right to quit any month. They paid for the option to leave, then stayed.

Does buy now pay later make people spend more?

Installment pricing is pricing psychology at its most direct: it changes the number the buyer judges. Ashby et al. (2025) combined real purchase transaction data with a series of experiments. Shoppers spent more with buy now pay later than with other ways of paying (Ashby et al., 2025). In the experiments, the engine was the installment price itself: showing something like 4 payments of $25 drove the extra spending (Ashby et al., 2025). That smaller number made the purchase feel less expensive, and the lighter feeling raised spending (Ashby et al., 2025).

The limits matter as much as the effect. Installment prices did not change spending when people paid by card (Ashby et al., 2025). More installments, or a smaller first payment, pushed spending higher still (Ashby et al., 2025). The published summary reports no effect sizes. So the direction is clear in these studies, while the size is unknown.

"Klarna and Affirm are two very popular companies that have direct integrations with Stripe that allow you to set up payment plans where for three or five payments, you can purchase this."

Jon Preston, founder of Atomic Funnels

The research backs the logic here, with one condition. Klarna and Affirm are buy now pay later providers, the payment mode where the effect appeared. But the lift came from showing the per-payment amount on the page, not merely from offering a plan (Ashby et al., 2025). No field study has tested payment plans on a coaching program, a course or a business purchase. The closest evidence is retail shopping and consumer subscriptions. The per-payment figure becomes a new reference point, much like the first number in a pitch (see the anchoring section).

Which subscription plan do people pick, and what is it worth?

A European internet provider's billing records show the gym pattern in a different market. Its 10,882 residential customers chose between two usage-based plans and a flat rate. On one plan, 48.1% had a flat-rate bias, meaning they bought more than their usage needed (Lambrecht & Skiera, 2006). Only 8.5% had the opposite bias (Lambrecht & Skiera, 2006). More than half of the overbuyers paid at least 100% more than the cheapest plan for their usage would have cost (Lambrecht & Skiera, 2006).

Here is the part sellers notice. The researchers' model put the lifetime value of overbuyers 87% to 135% above customers on their cheapest plan (Lambrecht & Skiera, 2006). Customers who underbought and paid by usage had monthly churn 340%-1040% higher (Lambrecht & Skiera, 2006). This is billing data, so it shows what went with what, not what caused it.

A field experiment on a new mobile phone service points to why. Danaher (2002) offered potential residential customers, most without a cell phone, different mixes of monthly access fees and per-minute charges. The access fee drove whether people quit: churn was much more sensitive to it than to the usage price (Danaher, 2002). The usage price mainly cut how much people used (Danaher, 2002). Both effects were modest for that service (Danaher, 2002). Price elasticities were much smaller than 1, meaning demand moved proportionally less than price (Danaher, 2002).

Prepaying also runs on a clock. In lab experiments on prepaid purchases, Gourville and Soman (1998) found that a past payment's pull on using what was bought fades over time. The researchers labeled this payment depreciation. It may help explain why a prepaid plan starts strong and then drifts.

What should change on the order form and at renewal?

Design the payment schedule on the order form on purpose. Say you sell a $3,000 program (illustrative numbers). Show pay in full at $3,000 next to 3 payments of $1,100, with the per-payment figure as the large number. That figure is what lowered felt expense in the retail studies (Ashby et al., 2025). More installments raised spending there (Ashby et al., 2025), so a 5-payment option is worth testing against 3. Judge it by cash collected and refunds, not sign-ups alone.

For subscriptions and retainers, treat each monthly fee as a moment when buyers decide again. Access fees drove churn in the phone experiment (Danaher, 2002). Remind people before each charge, show them what they used, and make cancelling easy. Revenue from people who stopped using you is fragile. The price increases section shows how perceived unfairness can cost more than the price itself.

Part 8What happens when you raise prices on existing customers?

Existing customers are likely to judge a new price against the one they paid before, not against your costs or your competitors. In a field study of business buyers, a small increase was followed by a revenue drop several times its size. Most of the demand lost to a price premium was a loss the researchers attribute to unfairness, not to the extra dollars. And cutting prices on things customers had just bought at full price backfired too.

What happened when a chemical supplier raised prices on its business customers?

Picture a supplier of adhesives and composites deciding to raise its prices worldwide. That decision produced a rare real-world test with business buyers, from Friess and Kassemeier (2024). The researchers used the company's own transaction records for 966 companies in 44 countries. They compared business customers who got the increase with similar customers who did not, before and after (Friess & Kassemeier, 2024).

A 2.1% average price increase was followed by a 10.6% drop in sales revenue from those customers, averaged across cultures (Friess & Kassemeier, 2024). A second estimate puts the drop at 10.7% (Friess & Kassemeier, 2024). In plain terms, a small change on the invoice went with a revenue loss far larger than the increase itself. Bigger increases went with bigger drops, though that link was only marginally significant until the researchers accounted for culture (Friess & Kassemeier, 2024).

How much the size of the increase mattered varied by country culture (Friess & Kassemeier, 2024). If you sell internationally, roll out increases market by market and watch each one. This was a natural experiment at one company selling a commodity, not a randomized test, so treat it as strong field evidence rather than a law. The lesson for B2B pricing still holds: a small line-item increase is not a small event for the relationship.

Business buyers weigh increases differently from cuts

Transaction data from business markets, where salespeople negotiate price and quantity, point the same way. Business customers react asymmetrically to price increases and price decreases (Bruno et al., 2012). The researchers read this as loss aversion (a loss feels bigger than an equal gain). They found that a longer history with the salesperson might make it stronger, not weaker (Bruno et al., 2012). The study is observational, so these are links, not proven causes.

For a consulting retainer or a software contract, that is a sobering thought. Your longest-standing clients may hold the firmest reference price: the one you set when they signed.

Unfairness can cost more than the price itself

Imagine opening a clothing catalog and finding that your size costs more than everyone else's. A mail-order catalog selling women's plus-size clothing ran a field test of exactly that situation. It mailed four versions of its catalog, each to 31,250 customers, some charging a premium for the largest sizes and some charging every size the same (Anderson & Simester, 2008). The size premium cut gross profit by 6% to 8% compared with the versions without it (Anderson & Simester, 2008). Charging some buyers more left the seller with less.

Then the researchers split the loss in two. For a 10% premium on large sizes, the higher price alone cost an estimated 8% of demand, with an extra 20% loss they attribute to perceived unfairness (Anderson & Simester, 2008). They inferred this from demand patterns rather than measuring how customers felt. Even so, the estimated fairness loss was twice the direct price effect (Anderson & Simester, 2008).

Buyers were not simply trading down to the next size (Anderson & Simester, 2008). And shoppers buying smaller sizes did not seem to care about the premium at all (Anderson & Simester, 2008).

The reaction came from the people paying more, presumably comparing their price with what others paid. That is the core of pricing psychology: buyers read a price against a comparison, and those comparisons are reference points you set. They work the way the anchoring section shows the first number on the table pulling offers toward it.

Cutting prices on recent buyers backfires too

Fairness runs in both directions. In a field experiment by Anderson and Simester (2010), a catalog company mailed customers who had recently paid a high price for an item either shallow or deep discounts on that item and others, 36 items in all. Those who saw the deep discounts placed 14.8% fewer orders over the next 28 months, over $90 in lost revenue per customer (Anderson & Simester, 2010). They averaged 2.78 orders vs 3.27 for the shallow-discount group (Anderson & Simester, 2010).

Many simply stopped buying: 34.0% placed no orders at all, vs 27.1% in the shallow-discount group (Anderson & Simester, 2010). That is roughly one in three customers walking away instead of about one in four. The drop was even stronger for items that were never discounted, so a poor-quality signal cannot fully explain it (Anderson & Simester, 2010).

Timing decided who reacted. The backlash hit customers who had bought within roughly the previous 20 months; customers whose earlier purchase was older were not harmed, and some bought more (Anderson & Simester, 2010). Sales fell 5.22% among customers who had bought within 250 days and rose 7.87% among those whose purchase was more than 750 days old (Anderson & Simester, 2010). This is the long-run cost the discounts section warns about, aimed squarely at the customers who trusted you at full price.

Price quietly drives churn as well as signups

For subscriptions, a higher price can cost you twice. When ZipRecruiter randomized its price for small-business employers, retention fell along with conversion at higher prices, though first-month profit still outweighed the lost renewals (Dubé & Misra, 2023). In a field trial of a new cell phone service, churn was much more sensitive to the monthly access fee than to the per-minute price (Danaher, 2002).

For a membership, a retainer or a software seat, the monthly fee is the access fee. Raise it and watch cancellations for months, not just this week's signups.

How to raise prices without losing the customers you have

Raise prices for new buyers first. Change the pricing page, the VSL and the sales call script. Keep existing clients at their old rate for a set period (grandfathering), or give them clear notice. Say a $3,000 coaching program moves to $3,500 (illustrative numbers): new leads see $3,500, while past clients who renew keep $3,000 for a stated window.

Tell existing clients why. None of these studies tested whether an explanation softens the reaction, but the researchers attribute much of the loss to fairness, so the reason you give is the lever worth testing. Avoid publicly slashing the price recent buyers paid in full; if you run a launch discount, offer them a credit or an extra session instead. Then track revenue per customer for months, because the catalog effect played out over 28 months.

Part 9How big are pricing psychology effects in the real world, and how should you test them?

Pricing effects measured with real money are usually smaller, patchier and more dependent on context than the headline numbers from hypothetical studies. The price that counts is the one your own traffic actually pays, judged by revenue over weeks, with honest room for doubt.

Famous pricing psychology effects shrink when the money is real

Picture a buyer who can pay whatever they like. As the anchoring section showed, Jung et al. (2016) ran sixteen field studies with real pay-what-you-want buyers. The textbook-sized anchoring effect appeared only in the paper's separate hypothetical studies (Jung et al., 2016).

That gap matters if you price from a survey. A "what would you pay?" question behaves like the hypothetical studies, not like your checkout. A pooled look at 53 willingness-to-pay studies, mostly with students, found that newer studies report noticeably smaller anchoring effects (Li et al., 2021). Use surveys to find the words buyers care about. Use real price tests to find the number.

Even clean field tests disagree with each other

Price endings show how messy real data gets. A national women's clothing catalog mailed 30,000 customers prices ending in 99 and another 30,000 prices ending in 00. The 99 version drew 47 more buyers, a gap too small to call real (Schindler & Kibarian, 1996). Yet it still sold more volume, as the charm pricing section showed. Counting buyers alone would have missed that.

Older tests split the same way. A department-store test across 11 products found no statistically significant difference between 98 and 00 endings (Schindler & Kibarian, 1996). Supermarket scanner data showed 13 of 21 brands selling more with 9 endings. That was not an experiment, though, and Schindler and Kibarian (1996) judged it inconclusive.

Price level is no tidier. ZipRecruiter randomized its monthly price across 10 cells for 7,867 prospective small-business customers. Revenue per prospect peaked at $399 among the prices tested (Dubé & Misra, 2023). Once the researchers accounted for statistical uncertainty, the best price could have been anywhere from $249 to $399, or above $399 (Dubé & Misra, 2023). That was a large, careful test, and it still produced a range, not a single number.

"By testing three prices out of the gate, you're gathering insight from the market about what the price of your product should be."

Jon Preston, founder of Atomic Funnels

The research backs the idea. Randomizing prices on live buyers is how ZipRecruiter learned it could charge well above where it started (Dubé & Misra, 2023). One condition: with three cells and modest traffic, expect to learn which region of prices works, not the exact best price.

Short tests can point you the wrong way

Most price tests stop at the first purchase. That hides the costs that arrive later. Take a trial of a new cell phone service with mostly first-time residential customers. Looking only at usage, and ignoring the customers who quit, underestimated price sensitivity by about 45% (Danaher, 2002). In plain terms, buyers looked far less bothered by the price than they really were.

A direct-mail catalog's experiments found a related trap. Short-run numbers overstated how much established customers responded to discounts, so prices set from them would be too low (Anderson & Simester, 2004).

Timing changes the picture too. During StubHub's fee experiment, buyers shown fees only at checkout spent almost 6% more per purchase than buyers shown all-in prices (Blake et al., 2021). After the whole site switched to checkout fees, the two groups spent the same (Blake et al., 2021). The gap belonged to the format buyers saw while the test ran. So recheck a winner after it becomes the default.

Rivals react as well. In a field experiment with 18,000 mobile subscribers near two competing movie theaters, targeted prices looked like a 1.18% and a 3.82% revenue gain over a single price (Dubé et al., 2017). That assumed each theater's rival would stand still. Once the rival could respond, the gains fell to 0.14% and 2.44% (Dubé et al., 2017). The researchers warn that tests holding a competitor fixed can mislead if a permanent change would provoke a response (Dubé et al., 2017). A price test is a snapshot of one market at one moment.

Price sensitivity is not a fixed number

Loyal grocery shoppers were half as price-sensitive as brand switchers (Mela et al., 1997). On a computer-parts price-comparison site, demand for one retailer's cheapest memory modules was extremely price-sensitive (Ellison & Ellison, 2009). Those sensitivities also drift, so last year's figure may not hold this year (Mela et al., 1997). Your warm list and your cold traffic will not react to a price the same way. Test them separately.

Sellers can misjudge their buyers, too. Across 25 US grocery chains, retailers used prices ending in nines but priced as if left-digit bias were much smaller than it was (Strulov-Shlain, 2023). How you score a test can change the winner too (Neumann et al., 2016). Decide the metric (revenue per visitor net of refunds) before launch.

How to run a price test on your own funnel

Treat every published effect size, meaning how big a difference was, as a hypothesis for your traffic, not a forecast. A few rules follow from the evidence above.

• Change one thing at a time. Price, ending, payment plan and fee display all move the buyer's reference point. Test them together and you can't tell which one worked.
• Score by revenue per visitor, not conversion rate. A higher price can sell to fewer people and still earn more, as the tour-boat photos in the first section showed.
• Count refunds, cancellations and renewals. Ignoring the people who quit is how the phone trial misjudged price sensitivity.
• Run full weeks and fix your sample size before you start. Buyers can behave differently on different days. Stopping the moment a variant looks ahead produces false winners. The test boxes in each section give starting points.
• On low-traffic, high-ticket funnels, test big differences. Say a $3,000 program makes 40 sales a month (illustrative numbers): you can tell $2,000 from $4,000, not $2,970 from $3,000.

Testing prices this way is slow, but it beats borrowing someone else's elasticity. Atomic Funnels offers a pricing split test with 3 or 5 price points as an optional upgrade to its builds, and the traffic picks the winner.

Part 10How to apply pricing psychology to your funnel

Applied to a funnel, pricing psychology comes down to a chain of reference points that your buyer meets in order. First the price level, then how it is displayed and how it ends. Then the order of numbers in the pitch, the tiers, the discounts, the payment schedule and the next price change. Set each one on purpose, then let your own traffic pick the winner, judged by revenue per visitor over months.

None of these moves has been field-tested on a $2,000-$20,000 program, and a few (when to reveal the price, decoys with business buyers, grandfathering) have no direct test at all, so each one is a bridge from the closest evidence.

Start with the price level. That is where the biggest money hid in the ZipRecruiter test. When ZipRecruiter randomized its monthly price for small-business employers, doubling the price cut conversions by only a quarter (Dubé & Misra, 2023). The data pointed to a best price well above what it had been charging (Dubé & Misra, 2023).

There was a catch: retention also fell at the higher prices (Dubé & Misra, 2023). So your scorecard has to run past the first sale.

Move 1: Test three price points before you polish anything else

Say you sell a $3,000 program. Send comparable traffic to $2,000, $3,000 and $4,500 versions (illustrative numbers), with everything else held fixed. Judge the result by revenue per visitor net of refunds, not by conversion rate. In the tour-boat photo study from the first section, very different purchase rates produced similar profit per photo (Gneezy et al., 2012).

Atomic Funnels builds can include a pricing split test with 3 or 5 price points as an optional upgrade, with traffic deciding the winner (see split-test architecture).

Whatever price wins, the VSL has to earn it. At a winery, a high price paired with a disappointing product was judged worse than the same product at a low price (Gneezy et al., 2014). A premium number raises the bar your delivery must clear.

Moves 2 and 3: Show one all-in price, and pick the ending that fits your position

On the order form, the first total a buyer sees should be the total they pay. Shrouded fees raised StubHub's revenue, but they also drove far more people out at the final checkout step (Blake et al., 2021). A high-ticket sale runs on trust. If you want a sweetener, test a free bonus against an equal discount instead of adding a surprise line at checkout, as the hidden-fees section explains.

Charm endings and decoy tiers, the tactics you see everywhere online, are refinements. The price level is the decision. In the catalog tests, $9 endings sometimes sold more (Anderson & Simester, 2003). In a separate catalog, 99-cent endings lifted sales volume but did not bring in significantly more buyers (Schindler & Kibarian, 1996).

Firms also tend to use round prices for higher-quality products (Stiving, 2000). For a premium program, a round price may say more than one ending in 7 or 9. Settle the level first, then test the ending.

Move 4: In the pitch, build value first, then say the number yourself

The first number on the table shapes everything after it. On eBay's Best Offer listings, buyers opened well below the list price, and the deal usually landed between the two opening numbers (Backus et al., 2020). In field studies with real pay-what-you-want buyers, anchors moved payments (Jung et al., 2016). The big, reliable effect, though, showed up only in hypothetical tasks (Jung et al., 2016).

Applied to a pitch, that points to a sequence. In the VSL and the webinar pitch, build the value case first, then state your price before the buyer invents one. On a sales call, say the list price plainly and early. Expect counteroffers to start from it, and decide your walk-away number before you dial.

Move 5: Build tiers buyers can actually compare

A decoy option shifted real in-store purchases (Doyle et al., 1999). But a meta-analysis (a study that pools many experiments) of 142 experimental observations found that the pull toward the middle option weakens when options trade price against quality (Neumann et al., 2016). That trade-off is what a pricing page does.

So make your tiers differ on concrete, countable features in the offer stack: number of coaching calls, months of access, done-for-you templates. Test two tiers against three by revenue per visitor.

Move 6: Keep the value stack believable and discount rarely

A "total value" line in your offer stack is a comparison price, and comparisons work only while buyers believe them. A sale sign alone lifted a clothing catalog's dress sales with no price change (Anderson & Simester, 2001). Yet grocery categories sold less once sale signs covered too many items (Anderson & Simester, 2001).

Deep cuts carry their own cost. Deeper Groupon discounts went with lower sales (Cao et al., 2018). And deep discounts on items customers had recently bought at full price led to fewer orders over the following months (Anderson & Simester, 2010).

On the order form, tie any offer to a dated reason (for example, 'launch pricing ends Friday at midnight, when the cohort closes'). Test a deadline email against no deadline and score revenue over 30 days. Save emailed discounts for leads who have gone quiet: in one ticket platform's field test, people who had not bought in the past year responded more to emailed offers (Sahni et al., 2017). A permanent markdown on a premium program tells buyers what it is really worth.

Move 7: Offer installments where checkout can show them

Show the per-payment amount beside the pay-in-full total. Installment display lifted spending only with buy-now-pay-later, not with cards (Ashby et al., 2025). On the order form, that might read "3 payments of $1,100" beside "$3,000 today" (illustrative). The payment-plans section covers the evidence and its limits.

Be honest about subscriptions and renewals. In a small survey, gym-goers forecast more than twice as many visits as members actually made (Della Vigna & Malmendier, 2006), so buyers may pick plans for the person they hope to become. Clear renewal dates and easy cancellation are the fair response to that gap.

Move 8: Raise prices on new buyers first

Existing customers are likely to judge a new price against the one they paid. After a European chemical supplier raised prices for business customers in 44 countries, the revenue that followed fell by far more than the increase itself (Friess & Kassemeier, 2024). In a catalog test, the demand loss the researchers attributed to perceived unfairness was larger than the direct effect of the higher price (Anderson & Simester, 2008).

The cautious reading follows from that. Change the price for new buyers first. Give existing clients notice or keep their rate, explain the reason, and track revenue for months.

What should you test first?

The table follows the buyer's path. Start at the top: price level and display carry the most money, and the finer moves only pay off once those are right.