Method · data as of 2026-10-08
How buyers are ranked
Companies tend to buy what they have bought before. WhoBuys compares a company's website with 4,583 past acquisitions by 159 companies and shows the deals behind every name.
56%
30%
159
The data
Wikipedia, in two ways. First, the per-company acquisition lists, which give one row per deal with a date and what the acquired company did: Google, Microsoft, Apple, Meta, Amazon, Cisco, Oracle, IBM, Adobe, Intel, SAP, eBay, Accenture, Yahoo, Twitter (X), AOL, HP, Dell, Citrix, Gen Digital (Norton), GoDaddy, BlackBerry, Broadcom (CA), Juniper Networks, Nokia, Sony, Rakuten, Electronic Arts, Take-Two Interactive, Red Hat. Second, for buyers without a list (Stripe, Salesforce, PayPal, Uber, Shopify and others), the buyer's own article and the articles of companies "acquired by" it. Each acquired company's own Wikipedia introduction is added, so a precedent describes the business, and links to its source.
The matching
WhoBuys reads the company's homepage and a few linked pages, then compares what it does with every past acquisition by meaning: each description becomes an embedding (OpenAI's text-embedding-3-small), so "product analytics" and "user behaviour tracking" count as close even without a word in common. Buyer and brand names are removed first, so a match is about the business, not who's mentioned. For each buyer, its three closest past deals give a fit score; only similarity above a threshold counts, so hundreds of loosely related deals don't add up, and recent, well-described deals count more. Two adjustments follow: specificity (closer than the buyer's usual match, strength 0.5) and a prior for buyers that acquire often (strength 0.25), both chosen on the backtest among settings where the five busiest buyers fill at most 60% of shortlists. If the embedding service is unavailable, a word-based match (TF-IDF) with its own calibration takes over.
The backtest
Every acquisition since 2010 in the data is a test: we hide it (and any other record of the same company), describe the acquired company in words, rank buyers, and look for the real one.
| Top 1 | Top 3 | Top 5 | |
|---|---|---|---|
| WhoBuys | 30% | 46% | 56% |
| Matching on shared words instead | 29% | 46% | 55% |
| Always the busiest buyers | 6% | 14% | 22% |
| Chance | 1% | 2% | 3% |
The percentages
Each buyer's percentage answers: if one of the 159 buyers in our data acquires this company, how likely is it this one? Scores are turned into percentages with one curve fitted on the backtest, then checked: we grouped every percentage shown in the backtest and counted how often that buyer really was the acquirer. In 4% of past deals, the real buyer had no close precedent at all, so the percentages add up to 96%, not 100%.
| Shown | Average shown | Actually the buyer | Cases |
|---|---|---|---|
| 0–5% | 3% | 3% | 9,545 |
| 5–10% | 7% | 8% | 3,953 |
| 10–20% | 14% | 15% | 1,431 |
| 20–30% | 24% | 33% | 391 |
| 30–50% | 38% | 45% | 213 |
| 50–100% | 69% | 65% | 144 |
Specialists
Big acquirers have bought a bit of everything, so they top many lists. Each company page also shows up to three specialists: buyers with at most 60 deals in our data that once bought a company very close to this one (the single closest past deal decides).
Limits
- Who's in the data decides who can be named. Wikipedia documents large, public acquirers best. Many active buyers are missing or have only a few deals here, and the backtest only covers buyers in the data.
- Fit is not intent. A company that bought something similar may not want another one, and nobody named is known to be interested.
- Matching on words misses context: two companies can describe different businesses with the same words.
- The buyer brief adds context from the company's own site and may name partners or competitors it mentions. Each quote is checked against its source.
- A website that needs JavaScript or blocks automated visits can't be read; paste a description instead.
Not investment advice. WhoBuys is independent and not affiliated with any company named. Acquisition data from Wikipedia under CC BY-SA 4.0: Google, Microsoft, Apple, Meta, Amazon, Cisco, Oracle, IBM, Adobe, Intel, SAP, eBay, Accenture, Yahoo, Twitter (X), AOL, HP, Dell, Citrix, Gen Digital (Norton), GoDaddy, BlackBerry, Broadcom (CA), Juniper Networks, Nokia, Sony, Rakuten, Electronic Arts, Take-Two Interactive, Red Hat and the articles linked from each deal.
Built with open source on GitHub