Sizing a market from a sample
Sizing a market means taking what was measured and scaling it to what was not. Every step in that scaling is an assumption, and the final figure inherits the uncertainty of all of them.
A market size quoted without a range is a point estimate from a chain of assumptions presented as a fact.
What does the chain look like?
A measured sample, a universe count, a weighting assumption, and frequently an assumption about how the unmeasured part differs from the measured part. Each link has its own uncertainty and the uncertainties compound.
Writing the chain out is the most useful thing to do with a market size. It shows where the answer is fragile.
Where does the universe count come from?
Commercial store databases, government counts, trade estimates or the provider's own fieldwork. In fragmented channels these disagree, sometimes substantially, because defining what counts as a store in scope is itself a judgment.
Two market sizes built on different universe counts will differ by that ratio before any behavioural assumption is made.
What is the most fragile assumption?
Usually that unmeasured stores behave like measured ones. Panels recruit stores that are willing and reachable, and willingness correlates with characteristics that may also correlate with sales.
Where that assumption can be tested — against an independent benchmark, against shipment data, against a fieldwork sample — it should be. Where it cannot, it should be stated.
How should a market size be presented?
As a range with the main assumptions named, and with a sensitivity showing how the figure moves if each assumption changes. That takes one table and it prevents the figure being quoted as a precise fact.
The sensitivity table is also the most persuasive part in a professional audience, because it shows the work rather than the conclusion.
What about triangulation?
Building the estimate two independent ways — bottom-up from stores and top-down from a category total — and comparing is the standard check. Agreement increases confidence; disagreement locates the weak assumption.
It is more work and it is the difference between a number you can defend and one you hope nobody probes.
What should a buyer of a market size ask?
How the universe was counted, how the sample maps to it, what the main assumptions are, and what the range is. Four questions, and a provider who answers them has produced an estimate rather than a headline.
How stale can a market size be?
Universes change as stores open and close, so an estimate more than a couple of years old should be treated as indicative. Re-basing periodically is part of maintaining the figure rather than an optional refresh.
Frequently asked questions
Why do published market sizes differ so much?
Because they define the market differently, count the universe differently, or scale differently. Comparing two market sizes without comparing their definitions compares arithmetic rather than markets.
Is a bottom-up or top-down estimate better?
Neither in general, and having both is considerably better than having one. Where they disagree, the reconciliation usually reveals the assumption that was carrying the most weight.
How wide should the range be?
Wide enough to reflect the assumption uncertainty honestly, which is usually wider than commercial audiences expect. A narrow range on a fragmented channel estimate is a claim about knowledge that is hard to support.
Can shipment data validate a market size?
Partly, for your own brand, which then extrapolates to the category if your share is known — and share is usually derived from the same estimate. The circularity is worth noticing rather than relying on.
What about a market that is growing?
Sizing and growth are separate estimates and growth is usually the steadier of the two, because sampling errors partly cancel in a ratio. A confident growth figure alongside an uncertain level is a normal and honest combination.
Should the universe include very small outlets?
That is a definitional choice that materially affects the answer, so it should be stated. The inclusion threshold is one of the main reasons two estimates differ.
How do I use a market size responsibly?
As an order of magnitude for planning rather than as a precise denominator. Using a fragile estimate as the denominator for a share target creates a target that moves when the estimate is re-based.
Does the independent channel make this harder?
Yes, because the universe is large, fragmented and not centrally registered. The uncertainty is genuinely larger and should be reported as such rather than smoothed over.
What is the most honest presentation?
The chain of assumptions, the range, and the sensitivity. Public statistical practice, such as the Census Bureau's published methodology, is a useful model for how carefully this is usually documented.
How should a market size be used in a target?
As a range, with the target expressed in absolute terms rather than as a share of an uncertain denominator. A share target against a re-baseable estimate moves when the estimate is revised, through nobody's doing.
### What should a brand actually do with this?
Build the estimate twice, from the top down and from the bottom up, before quoting it anywhere. The reconciliation between the two is where the fragile assumption reveals itself, and it takes less time than defending a single figure later.
Then publish the range internally alongside the point. A figure that circulates without its range will be treated as precise within a quarter, regardless of how it was originally presented.
Working with retail data? See NRS Insights data solutions or contact NRS Insights.