Working Note · Method

Two Markets in One: Detecting Latent Price Segments, and the Logic of a Premium Price

Tommaso Marson

Working note · 2026

1The problem with an average

A price decision rests on an estimate of what a market will bear. A single figure carries an implicit claim: the market is coherent enough for one number to represent it. The claim usually goes untested. The figure itself gives no sign of whether it describes one population or averages over several.

An average is faithful to its population only when the population is unimodal, distributed around one centre. Pool two groups with different valuations and the mean falls between them, at a value that may correspond to no one. The number is not imprecise. It can be perfectly stable under re-measurement and still mislead, and its precision hides the problem. This is the pricing case of a general hazard in aggregated data: a relation that holds within every subgroup is distorted or reversed once the subgroups are combined [Blyth, 1972].

2The study

The product was Apple AirPods 4, chosen for a strong public brand and a retail price high enough to make the purchase a genuine decision rather than a reflex. Participants were Italian respondents. Each judged the product at a range of prices in a simple, time-pressured buy / would-not-buy task, so the measure of resistance came from immediate response rather than from deliberated justification. The same participants separately completed a direct, conventional elicitation of the prices they considered cheap, expensive, and unacceptable, a standard stated-price instrument [Van Westendorp, 1976]. The two measures are of different kinds. One reads behaviour under pressure; the other reads a considered statement. That difference makes the second useful as a check on the first.

The study is reported here at the level of its results. The measurement and estimation procedures are LIMEN’s method and are not the subject. The argument depends only on the reported quantities and on the independence of the two instruments.

3A threshold that described no one

Estimated across the whole sample, the derived price threshold (the price at which the population, on balance, tips from acceptance to refusal) fell between the two groups described below. The figure was stable under re-estimation and unremarkable on its face.

It did not survive examination. Between individuals, the variation formed two separated concentrations rather than one spread around that centre: a group whose resistance set in far below that figure, and a group that stayed receptive well above it. The single figure sat in the gap between them. Almost no individual buyer changed their decision there. The single population threshold was withdrawn on that basis. It is not reported here: accurate as an average, false as a description of anyone.

4Two thresholds

Estimated separately, the two groups had two clearly different thresholds. One segment, call it price-sensitive, showed a sharp threshold near €101: acceptance held below it and fell away steeply above. The other, premium-receptive, had a threshold near €149, almost €50 higher, with resistance building gradually across the premium range rather than collapsing at a point. The two segments were of comparable size, roughly half the sample each. The split describes this sample of 52 consumers; it was not set before the study. What had looked like one market with a moderate threshold was two markets with two thresholds and a wide gap between them.

5Testing the segments on an independent measure

Dividing a sample into two groups is not, by itself, a finding. Any dataset can be partitioned into subgroups that differ on the measure used to partition them. That difference is guaranteed by construction. The question that matters is whether the two groups differ on something they were not split by.

They did. The groups were defined solely by purchase decisions under time pressure. Tested on the separate stated-price instrument (a different task, taken deliberately, with no shared mechanism), they diverged clearly in what they said they would pay, at a level far beyond chance (p < 0.001). Directionally, the price-sensitive group’s stated valuations sat near €105 and the premium-receptive group’s near €160. The groups react differently under pressure, and they also say different things about price on an independent test. Two unrelated instruments point to the same boundary. This supports the split; it does not validate it. A test of the split itself, on a larger sample, is still to run [Cronbach & Meehl, 1955; Wedel & Kamakura, 2000].

6The logic of the retail price

Apple’s retail price for the product sits essentially on the premium-receptive segment’s derived threshold: at the top of what that segment will bear, rather than in the comfortable interior of the demand curve. Read against the two-segment structure, the position is coherent rather than coincidental. At the retail price the price-sensitive segment is almost entirely priced out and only a minority of the whole sample would buy. Each of those sales, though, is taken at the edge of the receptive segment’s acceptance, where the margin per unit is greatest.

The price reads as a margin position, not a volume one. With an assumed cost of €60, it gives about 72% of the maximum revenue and about 86% of the maximum profit: a choice for margin and position, not for maximum profit. It anchors at the psychological point where the addressed segment stands on the boundary between yes and no. A single-market view cannot see this; it reads the same price as demand left on the table. The two-market view shows the price doing what a premium strategy is meant to do. More usefully, it also shows the sharp lower threshold that a volume-led strategy would have to respect.

7Scope and limitations

The findings are bounded. The sample came from a single national market, so the results are specific to these respondents and should not be assumed to transfer across markets. The segment proportions are directional at this sample size. The split is descriptive: it was not set before the study, and a test of the split itself is still to run. The independent measure supports it. The lower threshold is a sharp region rather than a finely located point; its position is read from the choice curve, not resolved to the euro. The stated-price instrument was administered after the decision task rather than counterbalanced against it, so the two should not be read as fully insulated from one another. The corroboration is strong on direction and should not be pressed for its decimals. The analysis underlying these results remains under internal review.

8Conclusion

A threshold can be stable and still average over two markets. A number can be precise and describe no one. Here an aggregate figure, precise and reproducible, turned out to be the midpoint of two populations that transact at very different prices. The model revealed the division; an independent instrument supports it. Two measures, one of behaviour and one of stated intent, met at the same boundary. Once the two markets are visible, a premium price that looked like forgone volume becomes a deliberate margin position at the edge of the segment it is meant for. The practical lesson comes before any particular technique: before pricing to a threshold, establish how many thresholds there are.