8 Scale and Population

How many customers your commitments require you to find

Once costs are understood as commitments, a constraint appears that demand curves cannot supply on their own.

A demand curve tells you how the people you studied respond to price. It says nothing about how many such people exist, how many you could ever reach, or how many would have to buy before your commitments are covered. While you had made no commitments, none of that mattered. The moment you have fixed costs, it decides everything.

This chapter is about the arithmetic those commitments impose. Not market sizing as a marketing exercise, but scale as a condition: given what you have agreed to pay whether or not anyone buys, how many customers must you actually find?

Market Size Is Not Demand

Entrepreneurs talk about market size as though it were part of demand. It is a different quantity answering a different question, and confusing the two hides the one that matters.

Demand describes how customers respond to price: whether quantity is sensitive or stubborn, whether cutting the price brings people in. Scale asks how many people could show up at all. A venture can face intense demand from a small population, or listless demand from an enormous one, and the shape of the curve tells you nothing about which situation you are in.

Population — the number of people who could plausibly have the problem. A ceiling on demand rather than a forecast of it.

Treat population as a constraint rather than an estimate. It is a ceiling on how much demand could ever materialize, a limit on how much contribution your fixed costs can draw from, and a bound on how wrong an optimistic projection is allowed to be. Population does not tell you what will happen. It tells you what cannot.

This is also why the chapter sits here rather than earlier. With no commitments, population is close to irrelevant; you can serve whoever turns up. With commitments, it becomes decisive, because the commitments have to be covered by however many people exist to cover them.

It is worth saying plainly that this is not market sizing in the usual sense. We are not estimating a share of an imagined market or forecasting growth, which is the move the chapter on the familiar tools took apart. We start from demand you learned, costs you chose, and commitments you are considering, and then ask whether a large enough reachable population exists to support them. Scale here is a test rather than an aspiration.

Existing Is Not Being Reachable

A population can exist and still be useless to you.

Existence asks who could plausibly have the problem. Access asks which of those people you can find, talk to, earn trust from, and serve, repeatedly, at a cost you can bear. The second question is answered by structure rather than effort: whether channels exist, whether gatekeepers stand in the way, what regulation applies, how much switching costs bind people to what they already use, whether the social risk of trying something new is high.

A population that exists but cannot be reached is economically equivalent to one that does not exist.

Entrepreneurs tend to treat access as a marketing problem to be solved later with more spend. Some access barriers do yield to time and money. Others are structural and will not, and the difference has to be assessed before commitments are justified rather than after.

Your early demand experiments do not settle this. They were run with people who were easy to find, willing to talk, and already aware of the problem, which is exactly right for learning demand and no evidence at all about reach. Recall that in executing the experiment the response rate itself was the first read on what acquisition costs; this is the same finding arriving with consequences attached.

So access shrinks the population you are allowed to use. Not the number who might benefit, but the number who are reachable, engageable, and realistically serviceable. That is not pessimism. It is the only version of the number that can carry a commitment.

Sample Is Not Population

You are now holding two numbers that have never met. A demand curve fitted to the people who answered your instrument, and a reachable population you have just argued down from everyone who might have the problem. The curve describes a few dozen people. The commitment is owed by all of them.

Putting the two on the same plane takes one multiplication. Count everyone who passed your screen and gave a usable answer, call that $\mathsf{n}$, and take the reachable population as $\mathsf{N}$. The ratio between them is the multiplier:

\[ \mathsf{k} = \frac{\mathsf{N}}{\mathsf{n}} \]

Multiply the quantity at every price by $\mathsf{k}$ and the curve you fitted for a sample becomes a curve for a market. Nothing else moves. Prices stay where they were and the shape is untouched, because the claim you are making is exactly that another $\mathsf{k}$ groups of people, otherwise like the ones who answered, would respond the way those people did. Apply it to the fitted curve rather than to the raw answers; the fit is where you already made your assumptions, and rescaling first only buries them.

Hand your AI or the app the fitted curve, the count $\mathsf{n}$, and the population $\mathsf{N}$, and the rescaled curve comes back. Ask for $\mathsf{k}$ itself by name, because it is the one number in the operation you can judge on sight. What had to be settled before any of it runs is $\mathsf{N}$, and the reductions in the previous section are that judgment. No arithmetic downstream repairs a wrong one.

Two things go wrong here, both quietly.

Who belongs in $\mathsf{n}$. Everyone who passed the screen, including the ones who told you they would buy nothing. A zero is a finding about the population rather than a missing row. Drop the non-buyers and $\mathsf{n}$ falls while $\mathsf{N}$ holds still, so $\mathsf{k}$ rises and every quantity downstream rises with it. Suppose forty-six people passed your screen and eleven of them would take none of it even at a price of zero. Rescale on the thirty-five who would buy and you have lifted the whole curve by about a third, and nothing in the output will look wrong.

What $\mathsf{k}$ assumes. That the sample behaves like the population. It is the strongest assumption in the entire method and the arithmetic never tests it. Getting from one person to a market put the hazard plainly: a sample skewed toward enthusiasts gives you a curve too high and too flat, which reads as raising the price is cheap. Multiplying by $\mathsf{k}$ multiplies that distortion along with everything else. If you recruited through your own network, this is the place to say so rather than the last page.

So look at $\mathsf{k}$ before you look at anything it produced. A curve built on forty people being multiplied by three thousand is a defensible thing to do and an uncomfortable thing to watch, and the discomfort is the correct response to it.

Adoption Is Not Penetration

One more distinction, and it is the one that trips up otherwise careful people.

Adoption is a customer-level decision: whether this person, facing this price, says yes. It is governed by demand, and your demand curve already contains it. Penetration is an aggregate outcome: what fraction of a population ends up adopting. It is the cumulative result of awareness, access, timing, trust, substitutes, and eventually competition, and none of that appears in a demand curve.

Fixed costs do not care who adopts. They care how many do.

Which is why early evidence cannot give you a penetration rate, however encouraging it looks. Thirty percent of the people we interviewed said they would buy is a statement about thirty percent of the people you could find, who were more motivated, better informed, more tolerant of friction, and more forgiving than the population behind them. That does not make the evidence useless. It makes it local.

Scaling a local result up to a population is the commonest way this goes wrong, and it is worth seeing on one picture.

The axis below is quantity, not price. That is deliberate: cost is a function of how much you make, so plotted this way your cost structure is a single straight line whose intercept is the commitment $\mathsf{f}$ and whose slope is the cost per unit $\mathsf{c}$, exactly as the previous chapter wrote it. That line never moves in what follows. The only thing the slider changes is how many customers exist to travel along it.

viewof kScale = Inputs.range([1, 12], {value: 1, step: 0.5,
  label: "Population, as a multiple of your sample"})

Ffix = 1000000
cVar = 200
cost = q => Ffix + cVar * q

// Sample demand Q = 2500 - 2P, so inverse demand P = (2500k - Q) / 2k
// and revenue R(Q) = Q * P(Q). Scaling multiplies the whole curve by k.
rev = q => q * (2500 * kScale - q) / (2 * kScale)
qCap = 2500 * kScale

// break-even where R(Q) = C(Q):  Q^2 - 2100kQ + 2000000k = 0
disc = 2100 * kScale * (2100 * kScale) - 4 * 2000000 * kScale
be = disc > 0
  ? [(2100 * kScale - Math.sqrt(disc)) / 2, (2100 * kScale + Math.sqrt(disc)) / 2]
  : null

qs = d3.range(0, qCap + 1, Math.max(5, qCap / 400))

Plot.plot({
  width: 700, height: 430, marginLeft: 86, marginBottom: 46, marginRight: 92,
  style: {background: "transparent", fontSize: "13px", color: "#6B7280"},
  x: {domain: [0, 31000], label: "Quantity \u2192", tickFormat: d3.format(","),
      labelAnchor: "right"},
  y: {domain: [0, 10000000], label: "\u2191 Dollars",
      tickFormat: d => `$${d3.format(".2s")(d).replace("G","B")}`,
      grid: true, labelAnchor: "top"},
  marks: [
    be ? Plot.rect([{x1: be[0], x2: Math.min(be[1], qCap)}],
          {x1: "x1", x2: "x2", y1: 0, y2: 10000000, fill: "#2F855A", fillOpacity: 0.08}) : null,
    be ? Plot.text([{x: (be[0] + Math.min(be[1], qCap)) / 2}],
          {x: "x", y: 9400000, text: () => "revenue covers cost here",
           fill: "#2F855A", fontSize: 11, textAnchor: "middle"}) : null,

    // the cost structure: ONE line, unchanged by anything the slider does
    Plot.line(d3.range(0, 31001, 250).map(q => ({q, y: cost(q)})),
              {x: "q", y: "y", stroke: "#C2410C", strokeWidth: 2.5}),
    Plot.text([{q: 30500, y: cost(30500)}], {x: "q", y: "y", dx: 6, dy: -6,
              text: () => "total cost", fill: "#C2410C", textAnchor: "start", fontSize: 12}),
    Plot.dot([{q: 0, y: Ffix}], {x: "q", y: "y", r: 4, fill: "#C2410C"}),
    Plot.text([{q: 0, y: Ffix}], {x: "q", y: "y", dx: 6, dy: -12,
              text: () => "f", fill: "#C2410C", fontSize: 12, textAnchor: "start"}),

    Plot.line(qs.map(q => ({q, y: rev(q)})),
              {x: "q", y: "y", stroke: "#2563EB", strokeWidth: 2.5}),
    Plot.text([{q: qCap * 0.5, y: rev(qCap * 0.5)}], {x: "q", y: "y", dy: -14,
              text: () => "revenue", fill: "#2563EB", fontSize: 12, textAnchor: "middle"}),
    Plot.ruleY([0], {stroke: "#E1E6E3"})
  ]
})

html`<p style="margin-top:.4rem;color:#4A5751">
At ${kScale}\u00d7 your sample, the best revenue available is
<strong>$${d3.format(",.0f")(rev(1250 * kScale))}</strong> against a cost of
<strong>$${d3.format(",.0f")(cost(1250 * kScale))}</strong> at that quantity.
${be
 ? `Revenue covers cost between <strong>${d3.format(",.0f")(be[0])}</strong> and
    <strong>${d3.format(",.0f")(Math.min(be[1], qCap))}</strong> units.`
 : `<span style="color:#A33A2A">The revenue curve never reaches the cost line. No price
    works, because there are not enough customers to reach one.</span>`}
</p>`

Start at one and the revenue curve is a small hump near the origin while the cost line climbs steadily away from it. Nothing is wrong with the cost structure and nothing is wrong with the demand; the sample simply does not contain enough people to pay for a commitment sized for a population. Push the slider up and the hump grows until it finally rises through the cost line, and the two crossings are the quantities between which this venture works at all.

Somewhere around one and a half to two times the sample, the curve first touches the line. Below that there is no price and no effort that helps. That threshold is not a forecast about the market. It is a fact about the commitment, and it was available before anything was spent.

The Penetration Your Commitments Require

Here is the part that turns all of this into a number.

Penetration usually gets talked about aspirationally. If we get just five percent of the market. That framing quietly reverses the logic, because penetration is not something you aim at. It is something your cost structure demands, and you can calculate it before you have any idea whether it is achievable.

Every sale contributes what is left after the cost of making it, which is the price minus the variable cost. That amount is the sale’s contribution, and the word is literal: it is what the sale contributes toward the fixed costs you have already committed. Divide your fixed commitment by that contribution and you have the number of units required before anything is covered:

\[ \mathsf{q_{required}} = \frac{\mathsf{f}}{\mathsf{p - c}} \]

Contribution — what one sale leaves after its own variable cost, $\mathsf{p - c}$, and therefore what it contributes toward fixed cost. If you have met contribution margin in accounting, this is the per-unit form of it.

That is a count of sales, and the population you can reach is a count of people, so one conversion stands between them: how many units a buyer takes in the period. Where the decision is yes or no, that number is one and the two counts coincide. Where buyers take several — the case the how-many elicitation was built for — divide the required sales by units per buyer first, or you will report a penetration several times worse than the one you face.

Then divide by the population you can actually reach, and you have the share of them who must buy.

Work it once, in the simple case. Commit $50,000. Charge $40 against a variable cost of $25, so each sale contributes $15. You need 3,334 sales, and if each buyer takes one, 3,334 buyers. If the population you can genuinely reach is 20,000 people, you need one in six of them, which is an extraordinary number for a new offering nobody has heard of. If it is 200,000, you need one in sixty, which is merely ambitious.

Same commitment, same price, same cost. The only thing that changed was who you can reach, and it moved the requirement by a factor of ten.

viewof fCost = Inputs.range([5000, 250000], {value: 50000, step: 5000, label: "Commitment"})
viewof pPrice = Inputs.range([10, 120], {value: 40, step: 1, label: "Price"})
viewof cUnit = Inputs.range([1, 100], {value: 25, step: 1, label: "Cost per unit"})
viewof uPer = Inputs.range([1, 12], {value: 1, step: 0.5, label: "Units one buyer takes"})
viewof popn = Inputs.range([2000, 400000], {value: 20000, step: 2000, label: "Reachable population"})

contrib = pPrice - cUnit
qReq = contrib > 0 ? fCost / contrib : Infinity
buyersReq = qReq / uPer          // sales are not people unless uPer is 1
penReq = buyersReq / popn
fmt = d3.format(",.0f")
pct = d3.format(".2%")

html`<div style="border-left:3px solid #146B45;padding:.6rem 0 .6rem .9rem;margin:.6rem 0">
${contrib <= 0
 ? `<p style="margin:0;color:#A33A2A"><strong>No price above cost.</strong> Every sale loses
    money, so no quantity and no population can cover the commitment.</p>`
 : `<p style="margin:0 0 .35rem">Each sale contributes <strong>$${fmt(contrib)}</strong>.
    Covering the commitment takes <strong>${fmt(qReq)}</strong> sales${
      uPer === 1 ? `` : `, which is ${fmt(buyersReq)} buyers at ${uPer} units each`}.</p>
    <p style="margin:0">Out of ${fmt(popn)} reachable people, that is
    <strong style="font-size:1.15em">${pct(penReq)}</strong> of them —
    about <strong>one in ${fmt(1/penReq)}</strong>.
    ${penReq > 1
      ? ` <span style="color:#A33A2A">More people than you can reach. This cannot work at
          this price with this commitment.</span>`
      : penReq > 0.20
        ? ` <span style="color:#A33A2A">A fifth of everyone reachable is a very hard
            number for something new.</span>`
        : penReq > 0.05
          ? ` <span style="color:#8A6A12">Demanding, and worth being honest about.</span>`
          : ` <span style="color:#1B6E45">Modest relative to the population.</span>`}</p>`}
</div>`

Move the sliders and notice which one has the most leverage. Cutting the commitment lowers the requirement proportionally. Raising the price does too, and faster than most people expect, because it widens contribution rather than merely adding revenue. Enlarging the reachable population lowers the share without touching the count, which is why the honest version of that number is the one filtered by access rather than the one from the industry report.

Notice also what the calculation is not. It is not a forecast. Nobody is claiming this penetration will occur. It answers a conditional: if this venture is to work, how much demand must ultimately materialize? That question comes before forecasting, before competition, and before any conversation about market share.

Which reframes the question you should be asking. Not is this a big market, which invites the answer you want, but is it plausible that this many reachable people will adopt, at a price that clears cost, in time to cover what I have already committed to? You can hold that against what you know about reach and trust and how new the idea is, and get an answer you would be willing to defend.

Scale, then, is not a strategy. It is not chosen because you want to grow. It is inherited from the commitments you make, and scaling never rescues a structure that does not work; it only amplifies whatever is already there.

Before you accept a population curve

Every profit number in the rest of this book is computed on the rescaled curve rather than the one you fitted. Four things before it goes forward.

Say $\mathsf{k}$ out loud. Multiplying by twelve and multiplying by twelve thousand are different acts, and only one of them is defensible from forty interviews.
Say who is in $\mathsf{n}$, including how many of them would buy nothing at any price. If you cannot say how many zeros you had, you do not know what you divided by.
Say the reductions that produced $\mathsf{N}$, in order, and name the ones where you took the low end of a range.
Say the required penetration as a fraction, not a percentage. One in six and 16.7% are the same number, and only one of them is hard to say without flinching.

If you cannot do all four, what you have is a larger sample rather than a market.

Putting It to Work

Ask yourself — how many, out of how many?

Take the commitment you are actually contemplating and run the arithmetic once, by hand. Deliberately by hand: this is the one calculation in the book worth refusing to delegate, because it is a single division and the whole of its value is in the moment you see the number. Divide it by the contribution one sale makes, and you have the number of sales you need. If a buyer takes more than one, divide again by units per buyer to turn sales into people.

Now the harder half. Write down the population you can genuinely reach: not everyone with the problem, but the ones you could find, contact, earn trust from, and serve again next month at a cost you can carry. If your honest answer is a range, use the low end.

Divide the buyers you need by the population you can reach, and say the result out loud as a fraction. One in six. One in forty. One in three hundred. Then ask somebody who knows the market whether a new and unknown offering takes that share of the people it can reach.

If the number embarrasses you, you have learned something before spending anything, which is the cheapest way this lesson is ever available.

The move: Calculate the penetration your commitments require before you decide whether it is achievable. The requirement is arithmetic; only its plausibility is a judgment.

You now know how customers respond to price, what serving them costs, what you would be committing to, and how many people would have to buy for that commitment to make sense. What remains is to put those together and see what comes out, which is the question the whole book has been circling.