Size a market with the denominator, not a top-50 list
A search returns the companies that matched. Market sizing needs the ones that did not, too — that is what a denominator is. Canonical delivers the full company population into your warehouse so segment counts are arithmetic instead of extrapolation.
01 — The problem
Where this breaks down today
Every TAM model starts with the same question: how many companies actually exist that look like this? Answer it with a search tool and you get a ranked list capped at some result limit, which tells you those companies exist but nothing about how many more do. So the number gets extrapolated from a sample, the sample's bias goes unexamined, and the model carries a confidence it has not earned.
Bulk delivery makes segment sizing a GROUP BY. With the full company universe in your own warehouse, you count the population directly, slice it by geography, headcount band, industry, and founding cohort, and re-cut the segmentation as many times as the analysis needs without a new data pull for each cut.
Sizing from a query-based tool means every segment count is a floor, not a count. Result caps hide the tail, so the segments most likely to be underestimated are exactly the small and non-US ones a growth case depends on. Comparing two segments is worse than estimating one: if both are capped, the ratio between them is meaningless. And because each new cut of the data is another set of queries, analysts stop exploring alternative segmentations long before they have found the one that actually explains the market.
02 — Why bulk delivery
Why an API per record is the wrong shape
Sizing is a population question, and a ranked result set is a sample with an unknown selection rule. You need the rows that did not match in order to compute a share, a penetration rate, or a coverage gap against your own customers. Cross-tabulation multiplies the problem: twelve geographies by six headcount bands by nine industries is 648 cells, which is 648 queries against an API and one GROUP BY against a local table. And a defensible model has to be reproducible — an auditor needs to re-run last quarter's number against the same data, which a live endpoint cannot give you.
03 — How it works
How Canonical solves it
Canonical delivers the complete company dataset to Snowflake, BigQuery, or your S3 bucket, with the firmographic and location attributes market models are built on: industry, business model, headquarters, founding year, and headcount, plus funding history where you need capital formation as a proxy for market maturity. Count segments directly, cross-tab them freely, and compare them honestly, because every cell comes from the same population. Coverage extends into the long tail, so the small-company end of a market — usually the part a TAM case rests on — is measured rather than guessed. Pin an analysis to a specific delivery and the number is reproducible months later.
-
Define the segmentation
Bring the dimensions your model cuts on — geography, industry, headcount band, founding cohort. We scope the delivery to cover them and send a sample so you can sanity-check segment counts you already have a view on.
-
Count in your own warehouse
Deliveries land in Snowflake, BigQuery, or S3. Segment sizing becomes ordinary SQL, so re-cutting the segmentation costs an analyst a query rather than a new data pull.
-
Refresh and track the shift
Take weekly, monthly, or quarterly snapshots and keep the ones your models cite. Comparing deliveries over time turns a static TAM into a view of how the market is actually forming.
04 — What's delivered
The attributes this use case relies on
Attribute groups, not an exact schema — the delivery schema is finalized with you during onboarding. See everything a delivery contains.
-
companyCanonical name, domain, founding year, and description.
-
headquartersCity, region, and country of the primary HQ.
-
firmographicsIndustry, business model, and offering attributes.
-
employee_countLatest estimated headcount.
-
fundingTotal raised, round count, and the latest round's series, date, amount, and lead investor.
05 — FAQ
Frequently asked questions
Why can't I size a market with a company search tool?
Because a search returns matches up to a result cap, which gives you a floor rather than a count, and no visibility into what the cap hid. Sizing needs the whole population so you can compute shares and compare segments on the same basis. Search is the right tool for finding companies, not for counting them.
How is this different from your funding market intelligence page?
Different job. Funding market intelligence is about round activity — who raised, at what stage, backed by whom — explored interactively through search or the MCP server. This is the full firmographic population delivered to your warehouse for counting and modeling. Teams often want both: one to watch the market move, the other to size it.
Can I compute our market share and penetration?
Yes, and that is usually the first thing teams run. Join your customer list to the delivered population and penetration by segment falls out directly, including the segments where you have no customers at all — which is the interesting half of the answer.
Is the long tail actually covered, or just the well-known companies?
Long-tail coverage is Canonical's reason for existing, and it matters most in exactly this use case, because small and non-US companies are where capped searches undercount worst. Verify it the honest way: run the sample against a segment you already know well and check the count against your own knowledge.
Can I reproduce a number from last quarter?
Yes, if you keep the delivery. Each snapshot is a complete, self-consistent copy of the universe at that point, so retaining the ones your models cite means an analysis can be re-run against the same data later.
Start with a sample dataset.
Tell us what you're building and we'll scope the delivery around it, then send a sample so you can test the join before committing to anything.