---
name: pricing-research-analysis
description: >
  Analyses and interprets stated-preference pricing research: price sensitivity
  meters, sequential price laddering, monadic price tests, choice-based pricing,
  derived demand curves and survey-based elasticity. States plainly that these
  methods measure stated willingness to pay rather than behaviour, corrects the
  standard over-readings, and reports a defensible price range rather than a
  single point. Use for "what price should we charge", "run the price
  sensitivity analysis", "what is the optimal price point", "build a demand
  curve from the survey", "what is our price elasticity", "will they pay more",
  "we tested three price points", or "the acceptable price range".
category: 06 Specialist and Advanced Analysis
ref: "06.02"
tier: 2
inherits: [K2, K3, K4, K5]
---

# Pricing Research Analysis

## 1. One-line description
Converts stated-preference pricing data into a defensible price range with its assumptions and its distance from actual purchase behaviour made explicit, and refuses the single optimal price point that every one of these methods invites and none of them supports.

## 2. What this skill is used for

**The research problem it solves.** Pricing is the decision research is asked to support most often and is least equipped to answer, because the only reliable evidence about what people will pay is what they paid. Survey pricing measures something adjacent: what people say, in a context with no money at stake, no competitor on the shelf, no salesperson, and no consequence for saying a number that is too low. Everything downstream inherits that limit. Layered on top are method-specific failures that have hardened into industry habit. The intersection points of a price sensitivity meter get quoted as optimal prices, which is not what they are and not what their originator claimed. A sequential price ladder produces a ceiling that reflects the starting point and the respondent's willingness to keep saying yes. A single acceptable price gets extracted from a range and put on a slide with no interval around it. A demand curve gets drawn through four tested points and extrapolated past all of them. This skill supplies the honest framing, the correct reading of each method, and an output that gives a decision-maker a range, a set of assumptions they can challenge, and a clear statement of what would have to be true.

**Where it sits in the research lifecycle.** After fieldwork on a pricing instrument, and before a commercial pricing decision. It runs alongside financial modelling rather than replacing it, and it frequently hands its most important output upward as a warning about what the research cannot establish.

**Typical use cases.**
- Reading a price sensitivity meter correctly, including telling a stakeholder why the intersection is not an optimal price.
- Analysing a sequential price ladder and adjusting for the acquiescence and anchoring it introduces.
- Comparing purchase intent across monadic price cells and turning that into an interpretable curve.
- Deriving a demand curve and stating the assumptions inside it.
- Estimating price elasticity from survey data and reporting the uncertainty honestly.
- Assessing whether the revenue-maximising and volume-maximising prices differ, and by how much.
- Auditing a pricing recommendation before a price change is made.
- Telling a client that the pricing study they ran cannot support the decision they want to make.

**Who uses it.** Research managers and directors running pricing studies; client-side insight and pricing analysts receiving them; commercial and revenue managers who will act on the number; product managers setting launch prices; and consultants who need to know how much weight a supplied pricing deck can carry.

## 3. When to use it

- A pricing instrument has been fielded and the data needs analysing into something a commercial team can use.
- A stakeholder is about to quote an intersection point from a price sensitivity meter as the recommended price.
- Purchase intent has been collected at several price points and needs turning into a demand view.
- A price elasticity figure is wanted, and someone should establish what the survey can actually estimate.
- A price change is being considered and the research evidence behind it needs auditing.
- Revenue and volume implications of a price need separating, because they point different ways.
- A pricing study has come back and the answer looks too good, which usually means the competitive frame was absent.
- Someone is about to present a single price point, and the honest output is a range.

## 4. When NOT to use it

- **Transaction data exists that could answer the question.** Where the organisation has actual sales at multiple price points, promotional history, or a market where prices have varied, that evidence is categorically stronger than any survey and should be analysed first. Stated preference is what you use when revealed preference is unavailable, not an alternative of equal standing. Hand the observational analysis to **05.06 Correlation, Regression and Causal Claim Control** for the confounding structure, and where a price test can actually be run, to **06.04 Experiment and A/B Test Analysis**, which is the only route in this library to a genuine causal price effect.
- **The decision is about a price for an unfamiliar or genuinely novel product.** Respondents cannot state a willingness to pay for something they cannot imagine using, and the numbers they produce will reflect their reference prices for adjacent categories rather than the product. The output will be precise and meaningless. Where the category is new, the honest research answer is qualitative work on value framing and reference points, not a price sensitivity meter.
- **The instrument was a single direct question about what someone would pay.** An open "what would you expect to pay" question with no context produces reference-price recall, not willingness to pay, and it is the single weakest pricing instrument in use. Report it descriptively as a reference price distribution and do not derive an optimal price from it.
- **The competitive frame was absent and the decision is competitive.** A price tested with no competitor present is a price tested in a market that does not exist. Respondents anchor on their own reference prices, and the acceptable range widens substantially. Where the client's actual question is "what should we charge given what they charge", a study without a competitive frame cannot answer it, and saying so is the correct output.
- **The trade-off between price and other attributes is the real question.** If the decision is what to charge for which specification, or how much of one feature buys how much price, that is a trade-off question and belongs to **06.01 Conjoint and MaxDiff Analysis**, which estimates price alongside the attributes it competes with. This skill takes the price output back from there and applies the stated-preference caveats to it, but it does not itself estimate multi-attribute trade-offs.
- **The base at the price point of interest is too small to carry the estimate.** Monadic price testing splits the sample across cells, and a five-cell design on 500 respondents gives 100 per cell before any subgroup. Below the **K4 §7** thresholds, report counts, not curves, and do not fit a function through points whose individual uncertainty is wider than the differences between them.
- **The output will be used as a revenue forecast.** A demand curve from survey data multiplied by a market size produces a revenue number that looks like a forecast and is a chain of assumptions, each of which could be wrong by a factor. Where a revenue projection is genuinely needed, build it explicitly as a model with every assumption named and a sensitivity range around each, and label it a model, not a research finding.
- **The purpose is to justify a price already decided.** Where the price is set and the study is being run to support it, no analysis is honest. Report the full range the evidence supports, including the part that contradicts the decision, per **K4 §4.2**.

## 5. Required inputs

**Required. Without these the skill cannot run. If absent, stop and ask.**
- **The exact question wording, in full, including the product description shown.** Pricing answers are extraordinarily sensitive to how the product was described, what was included, what currency and unit were used, and what context preceded the question. A price for "the subscription" and a price for "the subscription, billed annually, cancel any time, including the premium tier" are different measurements.
- **The instrument type and its full structure.** For a price sensitivity meter, all four questions in their exact wording and order. For a ladder, the starting price, the step size, the direction, and the stopping rule. For monadic testing, the cell allocation method and the price shown in each cell. Without the structure, the data cannot be read.
- **The base for every price point, unweighted**, and the weighting scheme if weighted. Every curve is built on these and a curve without its bases underneath it is not reportable.
- **The sample definition and screening criteria.** Whether respondents were category users, brand users, rejectors, or the general population changes every number and every reading.
- **Whether a competitive frame was shown, and what it was.** Determines whether the results describe a market or a vacuum.

**Optional, and what each one adds.**
- **Actual transaction or sales data at any price.** The single most valuable input, because it allows the stated results to be calibrated against at least one observed point, which is the only external check available.
- **Current price and current volume.** Anchors the demand curve at a known point and turns a relative curve into something with a scale.
- **Unit cost or contribution margin.** Enables the margin view, without which a revenue-optimal price is only half the decision, and usually the less important half.
- **Competitor prices and their sources.** Allows the competitive frame to be assessed even if it was not shown, and allows the results to be read against the market that exists.
- **A previous pricing study with the same instrument.** Establishes how this measure moves for this category, which is the only way to distinguish a real shift from instrument noise.
- **Purchase frequency or volume per buyer.** Converts an intent curve into a volume view rather than a penetration view.
- **Segment or channel variables.** Allows price sensitivity to be described by group, which is usually more actionable than a single market-level answer.

## 6. Questions to ask before starting

1. **What pricing decision is being made, and what are the candidate prices?** Determines whether the study needs to distinguish between two nearby prices, which requires much more precision than establishing a broad range. Default if unanswered: produce a range with its uncertainty and refuse to select within it.
2. **Was a competitive frame shown, and if so what was in it?** Determines whether the output describes a market. Default: state that no competitive frame was present, and that acceptable ranges from context-free pricing questions are systematically wider and higher than what a competitive market produces.
3. **Are respondents category users, and how recently did they buy?** Determines how much the reference price is grounded in behaviour rather than imagination. Default: report the composition and read the results separately for recent buyers, whose stated prices are closer to something real.
4. **What is the current price, and did respondents know it?** Determines whether stated prices are an anchored recall of the current price or an independent judgement. Default: assume anchoring where the product is identifiable, and read the results as movement relative to the current price rather than as absolute levels.
5. **Is the question about penetration, revenue or margin?** These have different optima and the study will be misread if the wrong one is optimised. Default: report all three where cost data permits, and state that they differ.
6. **Will this feed a commercial forecast?** Determines how firmly the boundary needs stating and whether the output should be a model rather than a finding. Default: attach the forecast boundary statement and flag the use as a **K5 §2.5** sign-off point.
7. **How many price points were tested, and how far apart?** Determines the resolution of any curve and where extrapolation would begin. Default: report the curve only across the tested range and mark the tested points on it.

## 7. Step-by-step methodology

**Step 1. State the position that governs everything else, in the output, once, at the front.** Every method in this skill measures stated willingness to pay. Stated willingness to pay differs from purchase behaviour systematically, not randomly, and the direction of the difference is known: respondents overstate willingness to pay for products they like the idea of, understate what they will pay for products where saying a high number feels foolish, and are influenced by the price the question implies. There is no correction factor that fixes this, because the size of the gap depends on the category, the product's familiarity, the respondent's involvement and the question's framing. What can be done is to say so, to prefer relative readings over absolute ones, and to calibrate against any observed price point available. A correct result at this step is a written framing statement that survives into the final deliverable rather than being edited out for confidence.

**Step 2. Reconstruct the instrument and identify what it can and cannot estimate.** Write out the questions in full, in order, with the product description. Then classify: does the instrument measure acceptability (a judgement about a price), intent (a claimed action at a price), or choice (a selection among priced alternatives)? These are three different constructs with three different distances from behaviour, and choice is closest. Record which price points were shown, in what order, and whether the respondent saw more than one. A correct result is a one-page instrument sheet plus a statement of the construct measured.

**Step 3. Clean at the respondent level with pricing-specific checks.** Four checks that generic cleaning misses. *Implausible values:* prices orders of magnitude away from the category, usually a units error (per month versus per year, or a currency confusion). Do not silently drop these; check whether they cluster, because a cluster means the question was ambiguous, which is a finding about the instrument. *Ordering violations in a price sensitivity meter:* where a respondent's "too cheap" price exceeds their "cheap" price, or "expensive" exceeds "too expensive", the respondent did not follow the logic. These cases must be excluded from the meter's cumulative curves because the curves assume ordering, and the exclusion rate is itself a quality indicator: above roughly ten percent suggests the questions were not understood. *Zero and refusal:* distinguish "would pay nothing" from "would not buy" from "declined to answer", which are three different data points frequently coded identically. *Straightlining across a ladder:* consistent yes to every step, or no to every step, which carries no information about the threshold. A correct result is a cleaning log with counts and rates per check, per **K4 §4.4**.

**Step 4. Analyse a price sensitivity meter correctly, and correct the standard over-reading.** The instrument asks four questions: at what price is it so cheap you would doubt the quality, at what price is it a bargain, at what price is it getting expensive, at what price is it too expensive to consider. Build four cumulative curves across the price axis: "too cheap" and "cheap" as descending cumulative, "expensive" and "too expensive" as ascending cumulative. Four intersections result, conventionally named, and here is where the discipline is required. The intersection of "too cheap" and "too expensive" is the point at which equal proportions reject the price for opposite reasons. **It is not an optimal price.** It contains no information about volume, revenue, margin or competition, and calling it optimal is an interpretive leap the instrument does not support. The intersection of "cheap" and "expensive" is the point of equal perceived value framing, and is likewise not an optimum. The two remaining intersections bound a range within which the proportion rejecting on either ground is limited. **The instrument's honest output is a range of acceptable prices and the shape of resistance across that range, nothing more.** Report the four curves themselves, not only their crossing points, because the shape carries most of the information: a steep "too expensive" curve means resistance concentrates in a narrow band, a shallow one means the market is heterogeneous and a single price will be wrong for many people. A correct result is a four-curve chart, a stated range, and an explicit sentence stating that the intersections are not optimal prices.

**Step 5. Analyse a sequential price ladder with its two biases named and, where possible, sized.** A ladder asks about purchase at one price, then moves up or down depending on the answer, until the respondent's threshold is found. It produces a clean-looking demand curve and carries two systematic problems. *Acquiescence:* respondents tend to keep agreeing, particularly to the first price offered and to a series of similar questions, so accepted prices drift upward relative to what a single question would produce. *Anchoring:* the starting price sets a reference, and thresholds found from a high start are systematically higher than thresholds found from a low start for the same population. Where the design randomised the starting price or the direction, this can be quantified directly by comparing the resulting curves, and that comparison should be the first analysis run. Where the design used a single start, the anchoring cannot be measured, and the correct response is to state that the curve's level is conditional on the starting price while its shape is more robust. Also check for a spike at the starting price and at round numbers, both of which indicate the ladder measured the anchor rather than the threshold. A correct result is an acceptance curve with the anchoring assessment attached and the shape emphasised over the level.

**Step 6. Analyse monadic price cells as the between-subjects experiment they are.** In a monadic design each respondent sees one price, so there is no within-respondent anchoring and no acquiescence chain, which makes it the cleanest of the stated-preference instruments and the most expensive, because each price point needs its own sample. Analyse it as a comparison of independent groups: purchase intent (usually top-two-box) per cell, with base sizes, and tests between adjacent cells per **05.02 Statistical Testing**. Two things matter. First, check cell comparability on the profiling variables, because random allocation on a modest sample can leave cells unbalanced on something that matters, and an imbalance on category usage will masquerade as price sensitivity. Second, resist fitting a smooth curve through a small number of noisy points: with four cells of 120, the confidence interval on each point may be wider than the gaps between them, and a fitted curve will imply a precision that does not exist. Plot the points with their intervals first, and fit only if the shape is clear through them. A correct result is a point plot with intervals, a balance check, and adjacent-cell tests.

**Step 7. Derive the demand curve, and write down every assumption inside it.** A demand curve from survey data is built by treating stated intent or acceptance at each price as a proportion who would buy, and plotting it against price. Four assumptions are inside that sentence and each must be stated in the output. *Intent equals behaviour*, which is false in a known direction and by an unknown amount. *The stated proportion is the population proportion*, which requires the sample to represent the buying population, and most pricing samples do not. *Other things hold constant*, including competitor prices, which in reality move in response. *The curve is defined only where it was measured*, so any point between tested prices is interpolation and any point outside is extrapolation with no evidential basis at all. Where an observed price and volume exist, calibrate: shift the curve so it passes through the known point, disclose the size of the shift, and report both the raw and calibrated curves. The size of that shift is itself the most useful number in the study, because it measures the gap between what this sample says and what a market does. A correct result is a curve plotted only across the tested range, with the assumption list beside it and the calibration shift reported if applied.

**Step 8. Compute the revenue and volume views separately, and expect them to disagree.** Volume (or penetration) is maximised at the lowest price tested, almost always, which is why volume optimisation alone is not a pricing answer. Revenue at each price is the price multiplied by the modelled quantity, and its maximum typically sits well above the volume maximum. Where contribution margin is available, compute margin as well, whose optimum sits higher still, because each unit sold at a low price carries the same cost. These three optima are usually different prices, and presenting one without the others is the commonest way a pricing study produces a wrong decision. Also report the flatness: where the revenue curve is flat across a wide band, the choice of price within that band should be made on strategic grounds (positioning, competitive response, channel relationships) rather than on the research, and saying so is more useful than naming a peak that is within noise of its neighbours. A correct result is a three-line chart or table with the optima marked, their differences stated, and the flatness of each curve described.

**Step 9. Estimate elasticity only where the design supports it, and report its uncertainty honestly.** Price elasticity is the percentage change in quantity for a one percent change in price. From survey data it can be computed as an arc elasticity between two tested price points, or from the slope of a fitted demand function. Three constraints. It is a local quantity, valid near the prices where it was estimated and not a constant of the product. Its uncertainty is wide, because it inherits the uncertainty of two proportions and then divides them, so a modest confidence interval on each point becomes a large one on the ratio. And it is elasticity of stated intent, not of sales, which for most categories will be too elastic, because respondents have no switching cost and no habit in a survey. Report the point estimate, its interval, the price range it applies to, and the construct it is elasticity of. Where the interval spans values with different commercial implications (say, elastic and inelastic), the honest output is that the study cannot determine which, not the midpoint. A correct result is an elasticity statement of the form: quantity of stated intent falls approximately X percent for a ten percent price rise between prices A and B, interval Y to Z, on bases n1 and n2.

**Step 10. Assess the competitive frame's effect, present or absent.** Where competitors were shown, read every result as conditional on those competitors at those prices, and say so, because the entire result moves if a competitor repositions. Where they were not shown, state the direction of the bias: context-free pricing questions produce wider acceptable ranges and higher acceptable ceilings than the same product priced against alternatives, because there is nothing to compare against and the respondent's only reference is their own imagination. Where competitor prices are known from desk research, add them to every chart as vertical lines, which is the single cheapest way to make an abstract price curve commercially readable. A correct result is every price chart carrying the competitive reference lines and a statement of what was shown to respondents.

**Step 11. Assemble the range, not the point.** The output of a pricing study is a range with a shape: where resistance begins, where it accelerates, where the revenue view peaks and how flat that peak is, where competitors sit, and how much of the range the evidence can actually distinguish between. Build it by overlaying the outputs from Steps 4 to 9 on a common price axis, and state for each candidate price what the evidence says and what it cannot say. Where a single recommendation is demanded, the correct response is to give the range, name the price within it that the evidence best supports, state the assumptions that would have to hold, and say what would change the answer. That last element matters most: a decision-maker who knows the answer depends on competitor response holding steady can watch for the thing that would invalidate it. A correct result is a price range chart, a candidate-price table, and an explicit statement of what could not be established (**K3 §5.2**).

**Step 12. Bound the commercial extension.** If the output will be multiplied by a market size to produce revenue, build that as a separate, labelled model with each assumption on its own line and a sensitivity range around each: market size, awareness, distribution, conversion from stated intent to purchase, repeat rate. Show the range that results from plausible variation in each. Almost always this range is wide enough to change the decision, which is the point of showing it. A correct result is a model with named assumptions, never a single revenue figure presented as a research finding.

## 8. Analytical framework

    Stated measure → Instrument correction → Curve across tested range
        → Revenue and margin view → Competitive placement → Range with assumptions

The framework's first term is the one that constrains everything after it. Every subsequent operation is arithmetic on a stated measure, and arithmetic does not convert a statement into a behaviour. That is why the framework ends at a range with assumptions rather than at a price: the assumptions are the bridge between what was measured and what will be decided, and making them visible is the analytical contribution.

Read the middle terms as corrections rather than transformations. Instrument correction removes what the ladder's anchor or the meter's ordering logic added. The curve is bounded by what was tested. The revenue view is the first point at which business arithmetic enters, and it is where the volume answer and the revenue answer separate. Competitive placement is what makes the range commercially readable. A pricing output that skips any of these steps will be a number without a context, which is exactly the form in which pricing research does damage.

## 9. Output format

**The framing statement.** Appears once, near the front, in the deliverable itself:

> This study measures stated willingness to pay. Stated willingness to pay differs from purchase behaviour systematically, and no correction factor reliably closes the gap. The results are strongest as relative readings (how resistance changes across the range, how segments differ, where the shape breaks) and weakest as absolute levels. Where an observed price point was available, the calibration is reported.

**The instrument sheet.** Question wording in full, product description shown, price points tested, competitive frame shown or absent, sample definition, base per price point.

**The price sensitivity output.** The four cumulative curves plotted together, the identified range, and a table:

| Curve pair | Price | What this point means | What it does not mean |
|---|---|---|---|

The final column is required. It is where "this is not an optimal price" is written.

**The demand view.**

| Price | Base at this price | Stated intent (top two box) | 95% interval | Indexed volume | Revenue index | Margin index |
|---|---|---|---|---|---|---|

With tested points marked and no line drawn outside the tested range.

**The elasticity statement.** Point estimate, interval, price range of validity, the construct (stated intent, not sales), and the bases.

**The candidate price table.**

| Candidate price | Evidence for | Evidence against | What must hold | Confidence |
|---|---|---|---|---|

**Where the evidence is thin**, use these forms rather than manufacturing a number:
- `[no optimal price identifiable: instrument produces an acceptable range only]`
- `[not estimable: base below 30 at this price point, counts reported instead]`
- `[outside tested range: no evidence at this price]`
- `[anchoring not measurable: single starting price, level conditional on it]`
- `[elasticity interval spans elastic and inelastic; the study cannot distinguish]`
- `[no competitive frame shown: ranges are systematically wider than a competitive context would produce]`

## 10. Quality checks

**K4 §8** runs anyway. These are specific to pricing analysis.

1. Does the framing statement about stated versus revealed preference appear in the deliverable, not only in the working notes?
2. Is the exact product description shown to respondents reproduced in the output?
3. Does every price point carry its base size?
4. Has any price sensitivity intersection been described as optimal, recommended or ideal anywhere?
5. Are the four price sensitivity curves shown, rather than only their crossings?
6. For a ladder, is the starting price disclosed, and is the anchoring effect either measured or stated as unmeasured?
7. Were price sensitivity ordering violations excluded, with the exclusion rate reported?
8. For monadic testing, was cell balance checked on the profiling variables?
9. Does any curve extend beyond the highest or below the lowest tested price?
10. Are volume, revenue and margin optima reported separately, with their differences stated?
11. Is the flatness of the revenue curve described, rather than only its peak?
12. Does the elasticity figure carry its interval, its price range and the statement that it is elasticity of stated intent?
13. Is the competitive frame's presence or absence stated on every price chart?
14. If a revenue projection was produced, is every assumption named with a sensitivity range, and is it labelled a model?
15. Does the output give a range with its uncertainty rather than a single point, and if a point was demanded, does it carry its conditions?

## 11. Common failure modes

| Failure | How to recognise it | How to prevent it |
|---|---|---|
| **Intersection quoted as optimal price** | A single price on a slide, sourced to a price sensitivity meter | The what-it-does-not-mean column is mandatory; the word optimal is banned from meter output |
| **Anchoring read as preference** | A ladder curve with a spike at the starting price | Randomise the start where possible; where not, report the level as conditional and lead with the shape |
| **Acquiescence inflating the ceiling** | Acceptance rates that stay high across implausible prices | Compare against a monadic or single-question benchmark where one exists; state the bias direction where none exists |
| **A curve through noise** | A smooth demand curve fitted through four cells of 100 | Plot points with intervals before fitting; fit only where the shape survives the intervals |
| **Extrapolation past the tested range** | A recommended price nobody was shown | Hard boundary; charts stop where the design stopped |
| **Volume optimum presented as the answer** | A recommendation to price at the bottom of the tested range | Always compute revenue and margin views alongside |
| **The absent competitor** | An acceptable range far above the market, with no competitor anywhere in the study | State the bias direction; add known competitor prices as reference lines |
| **Reference price mistaken for willingness to pay** | An open "what would you expect to pay" question driving a recommendation | Report as reference price distribution, derive nothing from it |
| **Intent converted to volume without a factor** | A market forecast built directly on top-two-box intent | Explicit conversion assumption, sized, sensitivity ranged, labelled a model |
| **AI: producing a precise optimal price** | A single figure with decimal places from a survey instrument | **K3 §4.4** on false precision; the output format is a range |
| **AI: smoothing a demand function through insufficient points** | A fitted equation with an R-squared, on five price cells | Points and intervals first; a fitted function requires justification |
| **AI: dropping the stated-preference framing for confidence** | Clean pricing prose with no mention of the method's limits | Framing statement is a required element of the deliverable |
| **Segment differences reported without bases** | Price sensitivity by segment on cells of 40 | **K4 §7** thresholds enforced on every split |

## 12. AI guardrails

Universal prohibitions are inherited from **K4**. **K4 §3.3** (never overgeneralise beyond the sample) and **K4 §3.5** (never make a recommendation stronger than the evidence supports) govern this skill in full. The following are specific.

1. **Never describe any output of a stated-preference pricing instrument as an optimal price**, including the intersection points of a price sensitivity meter. The correct terms are acceptable range, point of least resistance within the tested range, and revenue-maximising price among the prices tested.
2. **Never present a price recommendation without the stated-preference framing in the same document**, and never move that framing to an appendix.
3. **Never produce a price, curve value or elasticity outside the range of prices actually tested.**
4. **Never convert stated purchase intent into volume, share or revenue without an explicit, named, sized conversion assumption** and a statement that the result is a model rather than a research finding.
5. **Never report an elasticity as a property of the product.** It is local to the prices tested, it is elasticity of a stated measure, and it carries an interval that must appear with it.
6. **Never report a single price where the evidence supports a range**, and where a single price is demanded, never supply it without its conditions and what would change it.
7. **Never analyse price sensitivity curves without excluding ordering violations**, and never omit the violation rate, which is a quality signal about the instrument.
8. **Never compare pricing results across studies with different product descriptions, competitive frames or instruments** as though the difference were a market movement.
9. **Never suppress the flatness of a revenue curve to name a peak.** Where several prices are within noise of each other, say so; a peak reported without its flatness invites a false precision the data cannot carry.
10. **Never treat a pricing figure supplied by a client as validated by this analysis.** Client-supplied prices, volumes and costs are labelled as such per **K2 §6**.

## 13. Best-practice principles

1. **The gap between stated and actual willingness to pay is the subject, not a caveat.** Treating it as a limitation to be noted once and then forgotten is how pricing research produces confident wrong answers. Build the whole output so that relative readings carry the argument and absolute levels carry a warning.
2. **Prefer shape to level, always.** Where resistance accelerates, where the curve is flat, how much the segments differ: these survive the stated-preference problem far better than any absolute price does, because the biases affect all points in roughly the same direction.
3. **One observed price point is worth more than the whole survey.** If the product or a close comparator has ever sold at a known price and volume, calibrate to it, and report the size of the adjustment. That number is the most honest measure of the instrument's bias you will get.
4. **A flat optimum is a finding, and usually a liberating one.** When revenue barely changes across a band, the research has told the business that price within that band is not the lever, which redirects attention to something that is.
5. **Segment-level price sensitivity beats a market-level answer.** Almost every market contains a price-insensitive minority and a highly sensitive majority, and the commercial opportunity is usually in serving them differently rather than in finding the single price that compromises for both.
6. **Ask what happens next, not just what price.** Competitor response, channel reaction and the cost of a price change (including the cost of changing it back) sit outside the study and frequently outweigh the difference between two candidate prices.
7. **Round numbers and thresholds are real.** People respond to price points, not to a continuous variable, and the tested points should reflect where the psychological thresholds are, not be evenly spaced for tidiness. Where a study spaced points evenly across a threshold, the curve will hide the step.
8. **Beware the pricing study run on the wrong sample.** A price tested among existing loyal customers tells you what your least price-sensitive audience says. Whether that generalises to acquisition is the whole question, and it is usually not asked.
9. **The instrument's weaknesses partly cancel across methods.** Where two methods with different biases (a monadic test and a choice-based exercise, say) point to a similar range, confidence rises considerably. Where they disagree, the disagreement is informative and must not be averaged away (**K4 §4.1**).
10. **Never let the pricing number leave without its date.** Price expectations move with inflation, competitor activity and category news faster than most research measures, and a pricing figure quoted eighteen months later is frequently the source of a bad decision.
11. **The most valuable output is often a refusal.** Telling a commercial team that the study cannot distinguish between two candidate prices, and what would, prevents a decision being made on noise, which is worth more than a spurious recommendation.
12. **Write the assumptions where the decision-maker will read them.** An assumption in an appendix is an assumption that will not be challenged, and an unchallenged assumption in a pricing model is where the error will be.

## 14. Worked example

**INPUT**

A fictional not-for-profit continuing education provider, Meridian Learning Trust, is setting the fee for a new professional certificate course. A study of 620 prospective learners (screened as considering professional development in the next twelve months, recruited from an online panel) fielded a price sensitivity meter and a monadic intent test with four cells at 400, 550, 700 and 850. No competitors were shown. The board asks for "the optimal price".

**PROCESS**

*Steps 1 to 3, framing and cleaning.* The framing statement is drafted. The instrument sheet shows the course description was a 60-word paragraph with no mention of accreditation status, which matters because accreditation is the dominant value driver in this category and its absence means the numbers describe a course whose value is undefined. This goes in as a limitation on everything. Cleaning: 51 respondents (8.2%) violated the price sensitivity ordering logic and are excluded from the meter curves, an acceptable rate. Eleven implausible values cluster at approximately one twelfth of the others, consistent with respondents answering per month rather than for the course, so the question's unit wording is flagged as ambiguous rather than the responses simply dropped.

*Step 4, price sensitivity.* Four curves built on n=569. The too-cheap and too-expensive curves cross at 610. The board's request would have this reported as the optimal price. It is not: it is the price at which equal proportions reject on quality-doubt and on cost grounds, and it contains nothing about volume or revenue. The acceptable range bounded by the other two intersections runs 480 to 780. More useful than any crossing point is the shape: the too-expensive curve is shallow between 500 and 700 and steepens sharply above 750, meaning resistance is spread out below 750 and concentrates above it.

*Steps 6 and 7, monadic cells.* Top-two-box intent by cell: 400 gives 44% (n=156), 550 gives 38% (n=154), 700 gives 27% (n=155), 850 gives 14% (n=155). Intervals are approximately plus or minus 8 points on each. Cell balance checked on age, current qualification level and employer funding: the 550 cell has notably more employer-funded respondents (31% against 22 to 24% elsewhere), which would inflate its intent. *The judgement call.* Rather than reweighting the cell, which would introduce its own problems on this base, the analysis reports the imbalance, notes that the 550 point is probably one to two points optimistic, and shows the curve both with and without employer-funded respondents. The shape is unchanged; only the level moves. That is reported.

*Step 8, revenue view.* Indexing volume to the intent proportions, revenue index by price: 400 gives 176, 550 gives 209, 700 gives 189, 850 gives 119. The revenue view peaks at 550 among the prices tested, and 550 and 700 are within the overlap of their intervals, so the study cannot distinguish them on revenue grounds. Margin cannot be computed: no cost data was supplied, and this is stated as a gap rather than assumed.

*Step 10, competitive frame.* Absent from the study. Desk research supplies three comparable certificates at 520, 690 and 900, which are added as reference lines. Against them, the acceptable range from the meter is unremarkable, but the intent drop between 550 and 700 sits exactly where the mid-market competitor sits, which is a plausible and untested explanation worth naming as a hypothesis.

**OUTPUT**

> The evidence supports a fee range of 500 to 700. Within it, revenue among the prices tested is highest at 550 (revenue index 209 against 189 at 700), but the two are within the uncertainty of each other and this study cannot distinguish them on revenue grounds. Stated intent falls from 38% at 550 to 27% at 700 (n=154 and n=155, intervals approximately plus or minus 8 points).

> Price resistance is shallow between 500 and 700 and steepens sharply above 750. The price at which equal proportions reject the course as too cheap and as too expensive is 610. This is a point of balanced resistance. It is not an optimal price and carries no information about volume, revenue or margin.

> **Hypothesis, low confidence.** The intent drop between 550 and 700 coincides with the position of a comparable certificate priced at 690. The study did not show competitors, so this cannot be tested here. A monadic test with the competitive set present would establish it.

> **What could not be established.** Margin-optimal price (no cost data supplied). The effect of accreditation status, which was absent from the description shown and is the dominant value driver in this category; every figure above describes a course whose accreditation is undefined. The effect of the competitive frame, which was not shown and which systematically widens acceptable ranges when absent.

**Researcher sign-off required.** This output will set a fee. The evidence supports a range and cannot select within it on research grounds. The choice between 550 and 700 rests on the Trust's positioning against the accredited competitor at 690 and on cost recovery, neither of which this study measured (**K5 §2.5**).

## 15. Advanced usage

**Combining stated with revealed evidence.** Where any transaction history exists (a pilot, an earlier product, a promotional period), the strongest analysis anchors the survey curve to the observed point and reports the ratio between stated and observed take-up. That ratio is transferable within a category and is worth maintaining across studies, because it turns the next study's stated numbers into calibrated ones.

**Segmented pricing and willingness-to-pay distributions.** Rather than a single curve, model the distribution of individual price thresholds and read the segments off it. This supports tiering, versioning and discount-structure decisions that a single-price analysis cannot, and it usually shows that the market-level optimum serves neither tail well.

**Price and specification jointly.** Where the real question is what to include at what price, hand to **06.01 Conjoint and MaxDiff Analysis** for the trade-off estimation and bring the price output back here for the willingness-to-pay caveats and the revenue view. This combination is stronger than either alone: conjoint handles the competitive set and the trade-offs, this skill handles the translation to a commercial range.

**Testing the price rather than asking about it.** Where the channel allows, a live price experiment is categorically better evidence and belongs to **06.04 Experiment and A/B Test Analysis**. Even a small live test with two prices, correctly randomised, dominates a large survey, and where one is feasible the honest advice is to run it. Where regulatory or fairness constraints prohibit differential pricing, say so, since that constraint is often the real reason survey pricing is being used.

**Reference price and framing effects as their own subject.** Anchoring, decoy effects, bundling and the way a price is expressed (per month against per year, inclusive against exclusive) can move stated willingness to pay by more than the substantive price differences being tested. Where the decision involves how to present a price rather than what to charge, design for that directly rather than reading it off a pricing study built for level.

**Tracking price perception over time.** Where a pricing measure is tracked, the shape and level both drift with inflation and competitor movement, so year-on-year comparison requires the instrument and the product description to be identical and still needs a real-terms adjustment. Hand the trend discipline to **05.05 Trend and Tracker Analysis**.

## 16. Skill chain

**Recommended previous skills**
- **02.01 Survey Questionnaire Design** and **02.07 Scale and Measurement Selection.** Hand over the instrument wording, without which pricing responses cannot be interpreted.
- **04.02 Data Cleaning.** Hands over a cleaned dataset, to which this skill adds pricing-specific checks on ordering violations, units and refusal coding.
- **06.01 Conjoint and MaxDiff Analysis.** Hands over price part-worths and sensitivity curves from a trade-off design, which this skill converts into a commercial range with the stated-preference caveats attached.
- **05.02 Statistical Testing.** Hands over the comparison discipline for monadic cells, including base adequacy and adjacent-cell testing.

**Recommended next skills**
- **08.01 Finding to Insight Development.** Takes the price range and the sensitivity shape and asks what they mean for the proposition, not just the price.
- **08.04 Recommendation Development.** Takes the candidate price table and builds the recommendation with its conditions, which is where the sign-off in **K5 §2.5** belongs.
- **06.04 Experiment and A/B Test Analysis.** Takes over where the organisation can actually test a price, which is stronger evidence than anything this skill can produce.

**Runs well alongside**
- **10.01 Literature Review and Desk Research**, which supplies competitor prices and category reference points that make an abstract price curve commercially readable.
- **05.05 Trend and Tracker Analysis**, where price perception is measured over time.
- **13.03 AI Output Verification**, which audits pricing decks for optimal-price claims, extrapolated curves and unsized intent-to-volume conversions.

---
A Yazi Supplied Skill and resource.