AI Pricing Research: How to Test Product Prices & Bundles
Learn how to compare prices, bundles and promotions with buyer reactions, costs and returns. Includes worked calculations, pilot choices and a study brief.
Pricing research compares responses to prices with the costs and returns of selling the offer. It helps you choose prices, bundles or promotions to test in market. This guide shows how to define the offer, compare buyer reactions and check whether higher interest is worth a lower margin. Pricing is one part of a market research plan: test the offer and alternatives before connecting stated choices with the economics.
Every pricing decision has two parts: how the complete offer changes the buyer’s choice and what each sale contributes after the relevant costs. Keep both visible for a first price, a price change or a promotion.
Book a demo to compare the complete priced offers your buyers would face. See our comparisons with human research and check simulated choices alongside the margin assumptions and paid market test.
Pricing research at a glance
Study choice | What the decision needs |
|---|---|
Purpose | Compare feasible prices for a complete offer and choose candidates for a real-market pilot. |
Buyer evidence | Separate stated interest or modeled qualification from observed purchases. |
Commercial inputs | Keep net receipts, variable costs, substitution and contribution assumptions visible. |
Method | Choose price perceptions, a price ladder, independent price cells or conjoint for the unresolved question. |
Next evidence | Use a suitable market pilot to observe paid trial, repeat buying and full-window contribution. |
How do you define the offer for pricing research?
Write down the exact pack size, inclusions, buying occasion and channel. A supermarket six-pack differs from a home-delivered subscription. Keep the product description the same when you want to understand price alone.
Example: a simulated foldable-phone price comparison
In the published foldable-phone study, lowering price from $1,999 to $1,599 increased simulated qualification in both matched groups. The next question was whether that wider audience was worth the lost margin. A pricing study should connect those two parts of the choice.
The table shows the qualified share at each tested price in the published simulated study.
Cohort and assigned profiles | Qualified share, $1,999 / $1,599 |
|---|---|
Core: 2,103 | 5.52% / 11.27% |
Expansion: 2,897 | 6.18% / 12.50% |
The same profiles evaluated matched 256GB menus at both prices: core n=2,103 and expansion n=2,897. Qualification means meeting the study’s simulated funding and purchase-decision requirements. These simulated results have no supplied intervals and come from different groups than the main-offer comparisons.
Read our published comparison.
Example: pricing a grocery six-pack
This hypothetical example follows sparkling-tea brand Lumen. It needs to know whether a cheaper six-pack would attract enough extra buyers to justify a lower margin. Use the study to choose feasible prices for a market test, keeping buyer responses apart from cost and contribution assumptions.
Use the brief below to keep Lumen’s target buyers, price options and market test together:
Brief field | Example to adapt |
|---|---|
Decision | Select two prices for a real-market pilot |
Offer | Six 250 ml cans of Lumen sparkling tea |
Audience | Adults who bought ready-to-drink tea within the past month |
Context | One-off purchase through a grocery retailer |
Price options | $12, $15 and $18, tax treatment stated consistently |
Main measure | Stated purchase likelihood at the displayed price |
Diagnostics | Value for money, expected alternative, reason for rejection |
Next evidence | Retailer economics and a controlled market pilot |
Include current buyers and plausible switchers, then compare them separately. One average can hide a price that suits frequent buyers but loses the new customers you hoped to reach.
The retailer also needs to know where Lumen’s buyers would come from. SKIM’s Unilever case describes virtual shelves for Unox and Andrélon, with brands, prices and pack sizes in a retail setting and market, shopper and financial inputs. Ask whether Lumen’s lower price wins competitors’ buyers, adds purchases or shifts them from a pack you already sell. Compare that substitution with net receipts and contribution before recommending a range.
Compare stated interest with the price index
Compare three independent groups, each with 100 responses to the same Lumen product at one price. Positive interest combines “definitely” and “probably would buy” answers:
Price | Interest and price × interest index, USD |
|---|---|
$12 | Stated interest: 60 of 100 (60%) Price × interest: $7.20 |
$15 | Stated interest: 50 of 100 (50%) Price × interest: $7.50 |
$18 | Stated interest: 35 of 100 (35%) Price × interest: $6.30 |
Price multiplied by interest gives a revenue index per person who qualifies for the study, under a simplified stated-interest assumption. It is not revenue per visitor or a forecast. Here, $15 has the highest index while $12 has more stated interest. Which matters more depends on the business objective.
With an assumed unit cost of $7, the contribution indexes for the same comparison are $3.00, $4.00 and $3.85: (price − cost) × stated-interest share. The contribution comparison narrows the apparent gap between $15 and $18. Retail fees, promotions and repeat purchase would change it again.
Small groups leave the price ranking uncertain
If these counts came from two independent random human samples, the 60% versus 50% interest gap would have an approximate 95% interval of −3.7 to +23.7 percentage points. At the stated $7 cost, the $15-minus-$18 contribution-index gap is $0.15. Its approximate interval would be −$1.14 to +$1.44 under the same binomial assumptions. The small advantage cannot establish the population ranking.
These calculations use invented counts and do not adjust for comparing several prices at once. Treat $15 and the other prices as pilot candidates. Check costs, risk and uncertainty before committing. Generating simulated answers does not make human-sampling intervals valid.
View an example study to check the audience, offer and findings before planning your own.
Check our simulated comparison at two prices and our pricing comparisons with published human studies. Use the first to judge the analysis returned and the second to check whether the tested audience and task resemble yours.
Price sensitivity: what changes when the price changes?
Price sensitivity concerns how buyers respond to a price or price change. Price elasticity compares a percentage change in quantity sold with a percentage change in price, as Circana’s definition explains. Lumen compares stated interest under displayed offers. It does not estimate elasticity from actual sales. Keep that difference clear when choosing a formula or reading a chart.
Choose the method for the question still open
Match the question to what prices people see. Four methods make those choices clear:
Uncertainty | Method and interpretation |
|---|---|
Where is a plausible price range? | Van Westendorp’s Price Sensitivity Meter asks for prices that seem too cheap, a bargain, expensive but still worth considering, and too expensive. It describes acceptable-price perceptions, rather than actual demand. |
How does stated buying change across a price ladder? | Gabor-Granger asks about buying at specified prices, then changes the price for the same respondent. The basic design does not include competitor prices. |
Which complete price offer should progress? | Monadic testing shows each independent group the same offer at one price. Each group avoids earlier test-price exposure; comparable audiences and adequate evidence in each cell still matter. |
Should features, packs or prices change together? | Choice-based conjoint presents alternative configurations with different attributes and prices. It estimates tradeoffs within the choices shown, so the alternatives and features included shape the question. |
See Qualtrics’ pricing-method guide for the question formats.
The Lumen example uses independent groups: each evaluates one complete offer at one price. If the retailer needs to understand switching between brands or packs, include realistic alternatives.
For subscription packages, usage limits and overages, see our SaaS pricing research guide.
Check answer quality and whether earlier prices affect later answers. Someone who sees three prices may use the first as a reference. Independent groups need enough answers for every audience-price comparison. For a Sapien study, agree what people see, buyer groups, response definitions and assumptions in the proposal. Keep them in the final report.
Use our pricing research methods guide to compare the questions each approach can answer before choosing the design for your offer.
Define what buyers can choose instead: their usual product, a competing bundle, waiting or buying nothing. Put those options in the brief. A price that wins against weak alternatives may fare differently against what people already buy.
If you are selecting a provider for a simulated choice study, compare Sapien and Subconscious AI on how the study is designed, operated and validated.
For simulated research, check the product and pricing validation comparisons. Check which population, question and measure were compared with the human reference. Agreement on a stated response does not show actual revenue or repeat buying; ask what evidence supports your planned decision.
The same result can serve different business objectives
Use the Lumen example to look beyond the largest column. At the assumed $7 cost, $15 matches the $12 contribution index when its stated-interest share is 3/8 = 37.5%. It matches the $18 index at 3.85/8 = 48.125%. Its 50% in this example therefore clears the latter comparison by only 1.875 percentage points. These are break-even calculations for the indexes, not estimates of real demand.
Business constraint | Interpretation and next evidence |
|---|---|
Acquire more trial buyers | Read the result: $12 has higher stated interest, with lower contribution index Evidence to obtain next: Actual trial cost and repeat behavior |
Fund each unit sold | Read the result: Inspect margin after retailer deductions, rather than gross price Evidence to obtain next: Net receipts and variable costs |
Protect contribution per eligible buyer | Read the result: $15 and $18 are close under the stated assumption Evidence to obtain next: Sensitivity to response uncertainty and cost |
Maintain a premium position | Read the result: A higher price may change what people compare the offer with Evidence to obtain next: Alternatives and credibility at that price |
Build a sheet with net receipts, unit cost and interest share as separate inputs. Change one assumption at a time before combining pessimistic cases. If retailer fees reduce net receipts, recalculate the index. Do not subtract them from a forecast the study never made.
Read contribution beside stated interest
Price | Interest and contribution index, USD |
|---|---|
$12 | Stated interest: 60% Contribution index: $3.00 |
$15 | Stated interest: 50% Contribution index: $4.00 |
$18 | Stated interest: 35% Contribution index: $3.85 |
Index calculation: (price − $7 pack cost) × stated-interest share, per person who qualifies for the study. Revenue indexes are $7.20, $7.50 and $6.30. These are not observed revenue or profit.
Use a pilot to test buying and contribution
Choose prices the business can offer, then study which buyers they attract.
Take $12 and $15 into a feasible pilot for Lumen. The first tests broader trial; the second tests higher contribution. Keep $18 as a sensitivity case until there is a clear business reason for its narrower appeal. Record net receipts, paid first purchases and repeat buying. Base the final choice on the returns from customers actually acquired.
If you are choosing who should deliver that work, compare our pricing research consulting routes by research, business interpretation and implementation responsibilities.
Example: a temporary price cut
A temporary discount needs its own comparison after setting a regular price. A two-week price cut may sell more packs but reduce contribution. Its lower margin also applies to people who would have paid the usual price.
For a separate hypothetical promotion exercise, assume $15 is Lumen’s regular price. The retailer receives $15 per pack before sales tax, or $12 during the promotion. Variable cost is $7 per pack either way. There is no supplier funding or other change in fees, costs or product mix. The campaign adds $60. These are planning assumptions, not results from the earlier interest groups.
Without the promotion, suppose the retailer would sell 100 packs in the two-week window. Contribution after the specified variable cost would be 100 × ($15 − $7) = $800. Selling 160 packs at the promotional price would leave 160 × ($12 − $7) − $60 = $740. A 60% increase in packs would still leave $60 less contribution under those assumptions.
The promotional price needs 172 packs to match the $800 comparison: ($800 + $60) ÷ $5. That is 72 more packs than the assumed no-promotion baseline. It is a break-even requirement, not a forecast that the discount will generate them.
Early buying can erase the campaign’s gain
Reaching the campaign-period threshold is not enough if the extra sales replace later full-price purchases. Continue the example at 172 promotional packs. Suppose the next period would normally sell another 100 packs at $15, but sells only 80 because 20 purchases happened early.
Across both periods, the no-promotion contribution would be 200 × $8 = $1,600. The promotional path would leave 172 × $5 − $60 + 80 × $8 = $1,440. The specified contribution is $160 lower, even though the campaign met its short-window threshold. The calculation shows the cost of bringing those 20 purchases forward. It does not show that Lumen’s customers would actually do this.
Van Heerde, Leeflang and Wittink's promotion research distinguishes cross-brand effects, purchases borrowed from other periods and category expansion. This matters because a sales increase can contain several mechanisms. Do not use its reported scanner-data shares as default assumptions for a new promotion.
Decide whose result matters. Lumen may gain from a competitor while the retailer shifts sales between products it already stocks. A bigger Lumen pack might replace another Lumen pack. Check that substitution before calling a sales increase additional growth for the business.
Check the offer before measuring buying behavior
First check the complete promotional offer. Is the price for one pack or conditional on buying several? Can buyers identify the quantity, saving, campaign dates and any eligibility rule? Show realistic alternatives, including the usual product, waiting and buying nothing. Explore whether the offer changes the intended choice or mainly rewards a purchase already planned.
Sapien can compare simulated reactions to those prices, bundles and terms across buyer groups. The responses can reveal an unclear condition or an objection worth testing. They cannot tell you how many actual purchases will be extra, replace another purchase or happen earlier.
For the commercial pilot, define a no-promotion comparison and a measurement period long enough to cover the normal buying cycle. NIST's experimental-design guidance explains random assignment to experimental units; choose a practical unit for the retailer and allow for differences between stores, periods or customers in the design.
Keep actual net receipts, units, variable costs, campaign costs and refunds alongside availability and exposure. Check the period after the offer ends, with the relevant repeat-purchase and substitution evidence. Stockouts or a simultaneous display change can alter the comparison and should be part of the record.
Move forward with terms buyers understand and a required sales increase the business considers plausible. Use the pilot to check whether contribution holds across the full buying period. Campaign-period sales alone answer only part of that question.
Connect buyer responses with the costs in your brief
Start with the commercial choice: a first price, a price change, a promotion or a bundle. Sapien's managed studies use simulated audiences built from data to compare those scenarios across buyer groups. The useful output combines simulated responses, explanations of switching or rejection and a recommendation with its assumptions.
Agree the comparison before the study starts. Bring the information you have and use the table to find what still needs agreement:
Scope field | What to agree |
|---|---|
Decision and objective | The price or offer choice, who will make it, and whether trial, contribution, switching or another outcome matters most. |
Complete offer and context | Inclusions, quantity, purchase occasion, channel and consistent tax treatment. |
Feasible alternatives | Candidate prices or bundles, the usual or competing option, and whether buyers can wait or choose nothing. |
Audience and comparison | Buyer eligibility, geography, relevant groups, price exposure and the comparisons the report must support. |
Questions and interpretation | Response definitions, diagnostic follow-ups, eligible bases, source status and the limits of the measures. |
Commercial constraints | Net receipts, variable costs, capacity or other conditions that could change the recommendation. |
Deliverables and next evidence | The comparison, explanation, recommendation and sensitivity view; remaining evidence, responsibilities and timing. |
The report should let you check the recommendation against the offer, audience, question and answer count. Put simulated results beside the business assumptions. For each price or bundle, ask what changes the choice, which groups respond differently and which evidence would answer the questions still open.
You can start before settling every candidate price. Record unknown costs and limits, then refine sensitive business inputs while planning the study. Agree which gaps must be filled before a recommendation can be made.
A Sapien study can compare Lumen’s complete six-pack offers, show differences between buyer groups and explore switching or rejection. Use those findings to select offers for the market test and the cost assumptions the pilot must check.
