Price Anchoring: Does Showing a Higher Price First Work?
A price anchoring test measures whether a higher reference price shifts revenue per visitor, not just which tier gets clicked. Full guide.
📚 This article is part of the guide How to A/B Test Pricing: The Complete Guide.
A price anchoring test is one of the few pricing experiments where you change nothing about the actual price and still move revenue. The complete guide to A/B testing pricing covers why pricing tests behave differently from a button or headline test; this article goes one layer deeper into a single, specific lever inside that guide: showing a higher reference price first, whether that is a premium plan at the top of a tier table, a crossed-out “was” price on a product page, or a decoy tier that almost nobody buys. This piece covers what the bias actually is, how to design the A/B test around it, which metric tells you whether it worked, a worked example with real significance numbers, and where the tactic turns from smart framing into something that erodes trust.
What Price Anchoring Actually Is
Anchoring is a well-documented cognitive bias: the first number a person sees becomes a reference point, and every judgment that follows leans on it, even when that number has nothing to do with the decision at hand. The classic demonstration comes from Amos Tversky and Daniel Kahneman, published in Science in 1974: participants spun a rigged wheel that only landed on 10 or 65, then estimated what percentage of United Nations member countries were African. People who saw 10 guessed a median of 25%; people who saw 65 guessed a median of 45%. The wheel had no logical connection to the question, and it still moved the answer by twenty points.
On a pricing page, the mechanism is the same. The first price a visitor sees becomes the ruler they use to judge every price that follows. A $199/month plan looks expensive if it is the first number on the page. The same $199/month plan looks like an afterthought if, a second earlier, the visitor already saw a $499/month plan. The middle plan’s price never changed. The perception of it did.
A closely related, equally well-established effect reinforces this: the decoy effect, described by Joel Huber, John Payne, and Christopher Puto in a 1982 paper in the Journal of Consumer Research. They showed that adding a third option, one that is worse than one of the two main choices but not worse than the other, can increase how often people pick the option that dominates the decoy, even though nothing changed about the two original options. Dan Ariely popularized a market version of this in Predictably Irrational: in a study of Economist magazine subscriptions, adding a print-only option priced the same as a print-plus-web bundle (a choice nobody rationally prefers over the bundle) pulled a majority of respondents toward the bundle, even though the print-only option itself was barely chosen at all.
Translated to a SaaS tier table, the same principle plays out: an expensive Enterprise-style plan at the top doesn’t need to convert on its own to be useful. It can exist purely to make the plan underneath it look reasonable by comparison. The same logic drives “compare at” pricing in ecommerce: a crossed-out reference price next to the current one, or a pricier item shown before the budget option in a product grid.
How to Design the A/B Test
The design itself is simple. What takes discipline is isolating the variable and choosing the right metric:
- Control (A): the pricing page as it stands today, in whatever order it currently uses (usually ascending: cheapest plan first).
- Variation (B): the exact same plans, same prices, same features, with only the order changed: the anchor tier appears first, followed by the rest.
- Isolate one variable. Change the order only. If you also change visual weight (a “most popular” badge), the price itself, or the feature list in the same test, you will not know which change caused the effect.
- Stable randomization. The same visitor should see the same order for the entire length of their decision, including return visits before they convert, which matters most in B2B where the decision rarely happens on the first visit.
- Run length. Cover at least one full decision cycle for your product, not a few convenient days. See the B2B row in the context table below.
The same lever shows up in two other common forms worth testing on their own: a “compare at” or “was” price (a crossed-out higher price next to the current one, common in ecommerce and some SaaS annual-plan pages), and a decoy tier (an option deliberately positioned to be dominated by one of the other two, the Economist-subscription pattern above). Both are anchoring tests in disguise, and both should be measured with the same metric discipline as reordering a tier table.
The Metric That Actually Decides, and Why the Obvious One Misleads
Here is the most common mistake, and the reason so many price anchoring tests you will find online read as inconclusive even when the underlying psychological effect is real. Measuring only “how many people chose the anchored tier” or “how many people chose the middle tier” is not the right primary metric. Anchoring can redistribute which tier people choose without changing how many people convert overall, and it can even lower overall conversion (some visitors see the table and bounce because it now looks pricier at first glance) while still raising average revenue because of a richer plan mix.
The correct primary metric is revenue per visitor (ARPU) across the entire funnel: sum the revenue from every plan sold on both sides of the test and divide by the total visitors who saw the pricing page. Track these as supporting metrics, not as the decision itself:
- Overall conversion (guardrail): it should not drop meaningfully. If it does, the anchor is pushing people away, not just repositioning their choice.
- Plan mix: what share of conversions each tier represents, useful for understanding the mechanism, not for deciding the test alone.
- Average order value / initial ticket: a quick signal that the mix shifted, but not the same thing as ARPU (a lower overall conversion with a higher ticket may or may not net out ahead).
A Worked Example, With the Calculator’s Real Numbers
Take a SaaS pricing page with three tiers, Starter at $29, Growth at $79, and Scale at $199, and 8,000 visitors per variation.
Guardrail check: overall conversion does not move. On both sides, 320 visitors out of 8,000 converted to some plan, a 4.0% rate on each side:
Two-sided two-proportion z-test. "Not significant" almost always means not enough sample, not that the versions are equal.
Paste 8,000 visitors and 320 conversions on both sides of the calculator above and the result is p = 1.0, no difference at all. That is expected, and it is exactly the trap: read only the overall conversion rate, and this test looks like it did nothing.
Now look at what happened inside those same 320 conversions:
| Plan | A · no anchor | B · anchor (Scale shown first) |
|---|---|---|
| Starter ($29) | 200 (62.5%) | 130 (40.6%) |
| Growth ($79) | 88 (27.5%) | 150 (46.9%) |
| Scale ($199) | 32 (10.0%) | 40 (12.5%) |
| Group revenue | $19,120 | $23,580 |
| ARPU (revenue over 8,000 visitors) | $2.39 | $2.95 |
Revenue per visitor rose from $2.39 to $2.95, a relative gain of roughly +23.3%, without changing overall conversion or a single sticker price. And if, instead of ARPU, you had measured only “who upgraded to Growth or Scale” (120 of 8,000 = 1.5% in A versus 190 of 8,000 = 2.375% in B), the test also comes back clearly significant: paste those numbers into the calculator above (visitors 8,000/8,000, conversions 120/190) and the result is p is approximately 0.00006, a z-score around 4.0, and a 95% confidence interval on the absolute lift of roughly +0.45 to +1.30 percentage points, with B winning. The two readings tell the same story from different angles: the guardrail (overall conversion) holds steady, and both the upgrade-mix metric and aggregate ARPU move by a statistically solid margin.
Anchoring Across Contexts: Ecommerce, SaaS B2C, and SaaS B2B
The psychological mechanism is identical everywhere, but the test design should adapt to the decision cycle and purchase type:
| Context | Where the anchor sits | Main thing to watch |
|---|---|---|
| Ecommerce (physical product) | A “premium” item at the top of a category, or a crossed-out list price next to the current one | Short decision cycle, the test can run in days, but watch the overall-conversion guardrail closely because visitors are often comparing you against a competitor open in another tab |
| SaaS B2C, self-serve | An annual or “Pro” plan pinned at the top of the pricing table | Also a short cycle, but watch churn: if the anchor pushes people into a plan above what they actually use, first-cycle cancellation rises and masks the short-term ARPU gain |
| SaaS B2B, consultative sale | An Enterprise plan at the top of the pricing page, ahead of the sales demo | The decision rarely happens on the first visit; the test needs weeks to cover the real cycle, and the ARPU metric should track closed contracts, not clicks on “request a demo” |
The thread through all three: an anchor should never be judged by the immediate click, only by the revenue that actually lands, measured over whatever window matches that sales cycle.
The Ethics Line: When Anchoring Turns Into a Deceptive Practice
Honest anchoring uses a real plan or price, one that some genuine share of customers actually choose and that delivers the value it promises. The problem starts the moment the anchor is fabricated purely to look expensive: an Enterprise tier padded with features nobody ever buys, or a “list price” that was never actually charged to anyone, existing only to make everything else look cheap by comparison. That is no longer a test of value perception, it is manipulation with nothing behind it, and it tends to backfire in two specific ways. Customers who spot the trick (through market comparison, a public review, or simply asking a sales rep a direct question) start associating the brand with dishonesty, and it drives up perceived unfairness and refund requests once someone feels the anchor was never a genuine option. Enterprise sales teams also end up fielding leads generated by a plan the company never truly intended to sell at that price, which breaks trust before the first real conversation even starts.
A separate trap deserves its own mention: testing an anchor change on visitors who remember the old price. Returning customers, especially long-tenured SaaS accounts checking the pricing page before a renewal, are not a blank slate the way a first-time visitor is. If they recall what they used to see and now find a much higher anchor at the top, the reaction is not “this makes the middle plan look reasonable,” it is “did the price change on me?” A price anchoring test should generally be scoped to new-visitor traffic unless you deliberately want to measure the reaction of an informed returning audience, which is a different, riskier experiment with its own guardrails (support tickets, churn, and complaint volume, not just ARPU).
A simple check settles most gray areas: if you removed the anchor tomorrow, would it still make sense as a real, sellable product at the price shown, for some genuine segment of customer? If yes, it is honest anchoring. If no, it is an empty decoy.
Common Mistakes in Price Anchoring Tests
- An anchor that is too obvious. An absurd price jump ($29 next to $2,990) signals manipulation instead of establishing a reference point, and attentive visitors notice the tactic, which can produce the opposite reaction of what was intended. Anchors work best when the jump is large but plausible within the category.
- Measuring only the anchored tier or the middle tier. As covered above, this ignores the effect on aggregate revenue and can call “no effect” on a test that actually moved meaningful revenue, or the reverse.
- Testing without enough time for a B2B decision cycle. B2B pricing rarely converts on the first visit. If the test runs for only a few days, you are measuring the fast deciders only, a biased slice of the audience, and the required sample size at your real traffic volume may not have been reached yet. Calculate the sample from your actual conversion rate with the statistical significance calculator before declaring any winner, and run at least one full decision cycle of your customer base (ask sales for that number if you don’t know it off the top of your head).
- Changing the price at the same time as the order. If you reprice tiers in the same test that reorders the table, you lose the ability to isolate the anchoring effect on its own.
- Ignoring returning-customer exposure. As covered above, a returning visitor who remembers the old price reads a new anchor very differently from a first-time visitor. Scope the test to new traffic unless the returning-customer reaction is exactly what you set out to measure.
Make This Automatic on Donnu
The mistake that invalidates the most price anchoring tests is not statistical, it is a metric mistake: reading only the conversion rate on the anchored tier and ignoring what happened to revenue across the whole funnel. Donnu A/B reads the test by the metric that actually decides whether the anchor worked, aggregate revenue per visitor, using the same significance engine behind every other experiment you run, so you do not have to build this calculation by hand every time you test the order of your pricing tiers.
Start a 14-day free trial and run your next pricing experiment measuring what actually moves the result. See also the complete guide to A/B testing pricing and multi-armed bandits for pricing tests, a useful next step once you have a few anchoring wins and want to route more traffic to the winner automatically.
Read also: The complete guide to A/B testing pricing · Multi-armed bandits for pricing tests · A/B test statistical significance calculator
Leia em português: Ancoragem de preço em teste A/B
References
- Tversky, A. & Kahneman, D. Judgment under Uncertainty: Heuristics and Biases. Science, 185(4157), 1124-1131, 1974. cs.tufts.edu/comp/150AIH/pdf/TverskyKa74.pdf.
- Huber, J., Payne, J.W. & Puto, C. Adding Asymmetrically Dominated Alternatives: Violations of Regularity and the Similarity Hypothesis. Journal of Consumer Research, 9(1), 90-98, 1982. academic.oup.com/jcr/article/9/1/90/1839380.
- Ariely, D. Predictably Irrational: The Hidden Forces That Shape Our Decisions. HarperCollins, 2008. danariely.com/books/predictably-irrational.
Frequently asked questions
- What is a price anchoring test?
- It is an A/B test that measures whether showing a higher reference price first (a pricier plan at the top of a tier table, a crossed-out "was" price, a premium item shown before cheaper ones) changes how visitors judge the value of everything they see afterward, and whether that shift moves revenue per visitor, not just clicks on one tier.
- Is price anchoring manipulative or unethical?
- It depends on whether the anchor is real. Anchoring with a genuine plan or price that some real customers actually choose and that delivers the value promised is honest framing. Inventing a plan or "was" price nobody was ever meant to buy, purely to make the rest look cheap, is deceptive, and it is the version that backfires once customers notice.
- What metric should decide a price anchoring test?
- Never the conversion rate to the anchored tier alone, and never the click rate on the middle tier alone. The primary metric is revenue per visitor (ARPU) across the whole pricing page, because anchoring can redistribute which tier people pick without changing, or even while lowering, overall conversion, and a test that only reads one slice of the funnel can call a false win or a false loss.
- How much traffic do I need to test price anchoring on a SaaS pricing page?
- More than it looks like on paper. Pricing pages tend to have low volume and a long decision cycle, sometimes days or weeks between a first visit and a signed contract, so the test needs to run long enough to capture that full cycle, not just same-day clicks. Compute the required sample from your actual conversion rate with the calculator embedded in this article before calling a winner.
- Does price anchoring work the same way in ecommerce as in SaaS?
- The underlying psychology is identical, but the test design changes with the decision cycle. Ecommerce anchors (a premium item shown first, a crossed-out list price) usually resolve in days because purchase cycles are short; SaaS anchors, especially in B2B where sales are consultative, need weeks to cover the real time between a first pricing-page visit and a signed deal.