Blog / Methodology
Methodology
How Fit My Deck's investor-thesis dataset is built: sources, parsing, sample sizes, date ranges, suppression rules, and our corrections policy.
The corpus
Every statistic in our dataset articles is computed from Fit My Deck's investor catalog: a corpus of roughly 1,800 investors across the US, EU and UK. For each investor we collect their publicly published materials — investment thesis pages, fund websites, and public deal announcements — and parse them into a structured profile: stated stage appetite, ticket-size ranges, sector interests, geography, and recent co-investments.
Parsing is performed by large language models against each investor's own published text, then normalized into a fixed taxonomy (six stage labels, five ticket-size buckets, thirteen sector groups). Profiles carry the date they were last enriched; every dataset article states the as-of date of the snapshot it was computed from.
Sample sizes and suppression
Aggregate statistics are only published when the underlying cell contains at least 15 investors (n ≥ 15). Cuts below that threshold are suppressed rather than reported. Each chart and each headline statistic carries its own sample size.
Stated preferences are not observed behavior: when an investor lists multiple stages or ticket ranges, they count toward each. Shares can therefore sum to more than 100%.
Named investors
Claims about a named investor come only from that investor's own published thesis or public announcements, and always link the source. Outliers and mismatches (for example, stated stage vs stated ticket size) are reported as market observations with n-of-many context, never as assessments of an individual firm.
Versioning and corrections
Dataset articles are versioned. When the underlying corpus is re-analyzed (quarterly), articles are updated in place with a new version number, a new as-of date, and an updateddateModifiedtimestamp. Previous revisions are retained internally.
If you believe a statistic is wrong, or that your firm's profile misrepresents your published thesis, email hello@fmd.vc. Verified corrections are applied to the corpus and affected articles are recomputed and republished with a correction note.
Citing our data
You are welcome to cite and reproduce our statistics and charts with attribution to Fit My Deck and a link to the source article. Every dataset article exposes machine-readableDatasetstructured data including version and date modified.