The STP Framework 2026: Targeting Sets the Cost Floor

Three STP framework steps, one of which carries a number

The sequence is associated with Philip Kotler's marketing texts, where it is set out as a chain from segmentation through targeting to positioning. We did not read a copy hosted on a primary domain, so the attribution stops at the author and the idea.

Segmentation describes. It cuts a population into groups on an axis, and the output is a set of buckets with sizes attached.

Targeting decides. It picks which buckets you will serve, and that decision produces a single number: how many buyers you are building the business for.

Positioning communicates. It sets what you claim relative to the alternatives, and in a pricing case it usually resolves to a ratio against a named rival's price.

Only the middle step commits capital. That is why it dominates, and a published pricing case shows by how much.

The mechanism is simple enough to state before the numbers arrive. A cost based price is fixed cost divided by expected users, plus a markup. Segmentation does not appear in that formula. Positioning does not either. The user count does, in the denominator, where a factor of two becomes a factor of two in the price.

That also explains a common failure in the room. A candidate who spends forty seconds naming five demographic segments and then picks all of them has left the denominator undefined, so every price that follows is unsupported. Picking one segment and naming its size is the shorter and better answer.

The published pricing case, filter by filter

Oliver Wyman publishes a case in which an Indian startup has to price a digital therapy product for autistic children before launch, checked 30 September 2026 at oliverwyman.com. The published data is unusually complete.

The firm's own chain runs in three approaches. On value, the product replaces a third of a 300,000 rupee doctor bill, so 300,000 divided by three is 100,000 rupees, and the page calls that the ceiling: this "suggests a potential price of INR 100,000 per year. However, this should be seen as the upper limit".

On volume, the page writes that "a realistic goal is to capture 25% of the calculated user base (25% of 4.2 million, which is roughly 1million)". By hand, 25 percent of 4,200,000 is 1,050,000, and the page announces its own rounding with the word roughly.

On cost, the fixed cost is divided by that user count. Ten billion rupees over one million users is 10,000 rupees a user a year, and the page applies a markup: "Using the cost-based approach and assuming a 50% markup, the product can be priced at INR 15,000 (INR 10,000 * 1.5)." By hand, 10,000 times 1.5 is 15,000 rupees.

All four of those operations check out. The published band of 15,000 to 100,000 rupees follows from them.

The 42 percent the case never shows

One link in that chain is missing, and it is the link that carries the whole cost based price.

The page names the filters: "The addressable market can be calculated by applying filters such as income levels, internet/tablet penetration, and English-speaking population to the total market size." It gives none of their values. The 4,200,000 figure appears without a derivation.

The ratio is still readable. 4,200,000 over 10,000,000 is 42 percent, so the filter chain implies a combined pass rate of 42 percent. Apply the one filter the case does publish, removing the 50 percent low income group, and 10,000,000 becomes 5,000,000. The remaining two filters then have to pass 0.42 over 0.50 jointly, which is 84 percent.

Eighty four percent of middle and high income Indian households with an autistic child having both a tablet and English language use is a strong assumption, and it may be the intended one. The point is that it is an assumption, it is doing most of the work, and a candidate who states it out loud has done something the published answer does not. That inference is ours, from the firm's own two numbers, and our research file labels it as such.

Oliver Wyman's own preparation page lists using numbers as one of its six method steps, and asking the right questions as one of its six evaluation criteria, checked 30 September 2026 at oliverwyman.com. Asking what the penetration filter is worth is the right question here, and it is worth asking before the price is computed rather than after.

How the STP framework target count moves the price floor

Hold the fixed cost and the markup, and vary only the targeting decision.

Each row by hand. Ten billion over 500,000 is 20,000 rupees, times 1.5 is 30,000. Over 1,050,000 it is 9,523.81, which rounds to 9,524, and times 1.5 is 14,285.7, rounding to 14,286. Over 2,000,000 it is 5,000, times 1.5 is 7,500. Over 4,200,000 it is 2,380.95, rounding to 2,381, and times 1.5 is 3,571.43, rounding to 3,571.

The price floor runs from 3,571 to 30,000 rupees across that range, which is a factor of 8.4. Now compare the other assumption in the same formula. Doubling the markup from 50 percent to 100 percent takes the price at one million users from 15,000 to 20,000 rupees, which is a factor of 1.33.

So the targeting step moves the cost based price 6.3 times further than the markup does, since 8.4 over 1.33 is 6.3. Segmentation produced the buckets and positioning will set the final number, and the decision that moved the floor by 8.4 times was the choice of how many users to build for.

A positioning example as a ratio to the nearest rival

The third step resolves into a comparison, and the case publishes enough to make it arithmetic rather than adjectival.

Three rivals are priced at 50,000, 60,000 and 75,000 rupees, with published feature differences: the cheapest has basic modules without remote support, the middle one adds gamification without progress tracking, and the dearest bundles consultations and a device. The client's product is digital only with gamification and progress tracking.

The published recommendation is a subscription "priced between INR 15,000 and INR 30,000 annually". Read as ratios, that is 30 percent to 60 percent of the cheapest rival's price, since 15,000 over 50,000 is 30 percent and 30,000 over 50,000 is 60 percent.

That ratio is the positioning statement, and it is falsifiable. A product at 30 percent of the cheapest competitor either is or is not credible to a middle income buyer, and a pilot answers it. Four steps turn STP into something a client can test.

1. Segment on an axis that changes behavior or cost. Here income changes ability to pay, which changes the reachable count.
2. State the target as a single number, and say what it implies. One million users, implying a 42 percent filter pass rate.
3. Divide fixed cost by that number before you talk about markup. The denominator is the decision; the markup is a preference.
4. Express positioning as a ratio against one named rival. 30 to 60 percent of the cheapest incumbent, and say what would disprove it.

Simon-Kucher publishes the position that framework knowledge helps "especially in our core focus areas such as pricing, marketing, digitalization, and sales", checked 30 September 2026 at simon-kucher.com. This is that territory, and the arithmetic above is what the knowledge is for.

For a contrast in which the published segmentation does not move the answer at all, see customer segmentation. The pricing family is in pricing strategy cases, and the mechanics of cutting a population are in segmentation analysis. Bain publishes a related promotion case costed against ten million customers at 100 dollars of average annual spend, checked 30 September 2026 at bain.com, and its calculation tables are images we did not read, so we reconstruct nothing from it.

What a three step recital costs

The weights here are ours, read out of src/lib/grading. Firms publish criteria and method steps; a weighted score is not something any of them puts in public.