Team first, technology as the ticket: what buyers actually pay for in very-early AI analytics deals
Asked (summary):
What makes a small AI analytics startup most attractive for a very early acquisition — its team, its customers, its data model, its distribution, or its technology — and what evidence shows that?
The evidence base is 18 official acquisition announcements in AI, data and analytics from June 2024 to September 2026 by Databricks, Salesforce, IBM, Microsoft, Snowflake and Qlik, read against the FTC's study of 819 unreported Big Tech transactions (2010–2019) and a workforce study of roughly 4,000 U.S. startup acquisitions. The case set is purposive — chosen to illuminate, not to estimate a population — and announcements are buyer-authored, therefore promotional.
How the five factors rank, and what each one is worth
Bars count how many of the 18 official announcements explicitly cite each factor. Rank reflects the strategic reading: team and technology are what the buyer pays for; data models and customers de-risk the deal; distribution usually flows from buyer to target, not the reverse. Hover a bar to see the actual cases behind it.
Primary acquisition valueProof / de-riskingUsually supplied by the buyer
Findings the chart cannot say alone
In the FTC's 819 unreported Big Tech transactions, 76.7% imposed non-competes and 79.1% used deferred or contingent pay for founders and key employees — contractual proof that keeping people was central to the deal's value. ftc.gov
Team retention is the fragile part: 33% of acquired workers left within a year versus 12% of comparable ordinary hires, across about 350,000 workers in roughly 4,000 U.S. startup acquisitions. lsandc.com
Tabular shows team and technology moving as one asset: Databricks bought the original Apache Iceberg creators together with their stewardship of the format's community. databricks.com
Distribution runs the other way at this stage — Salesforce promised to bring Waii "to millions of users," and Databricks will put Row Zero in front of all its customers on every major cloud. salesforce.com
Evidence matrix: 18 official announcements against the five factors
A dot means the announcement explicitly cites that factor; hover or tap a dot for the wording. Dot colour follows the factor's role: black for team, blue for technology, gold for data model and customers (proof), grey for distribution. Zoomin and Apromore are the counterexamples — already-installed traction made customers and ecosystem fit the headline, which is why customers rise in weight after the very earliest stage.
Target
Buyer
Announced
Team
Technology
Data model
Customers
Distribution
Source
The two quantitative anchors
FTC study of 819 unreported transactions, 2010–2019
Alphabet, Amazon, Apple, Facebook and Microsoft reported 819 non-HSR transactions, among them 101 explicit hiring events against 91 patent acquisitions. Of operating-company deals above $1m, 65% fell between $1m and $25m, and between 39.3% and 47.9% of targets were under five years old. Non-competes bound founders and key employees in 76.7% of deals; more than 79% used deferred or contingent compensation — the structure of the contracts says the people were the asset being secured.
J. Daniel Kim's research on roughly 4,000 U.S. high-tech startup acquisitions (1990–2011, about 350,000 workers, Census matched data) finds 33% of acquired workers gone within a year versus 12% of comparable regular hires, and acquired employees 15% more likely to leave over three years, driven by organisational mismatch. If the team is the asset, it is also the asset most at risk of walking out.
Secondary-source summary of the study · lsandc.com
Founder checklist: the most acquirable profile
A small, exceptional team with scarce domain and engineering judgment that cannot be cleanly separated from the code
A narrow capability already working on real enterprise data — a working wedge, not a demo
Proprietary semantic, evaluation, governance, compiler or migration know-how rather than a generic model wrapper
Clean IP ownership and enterprise-grade security posture
Two or three referenceable design partners proving accuracy, workflow fit and willingness to pay
An obvious integration path into a named buyer product — the buyer supplies the distribution
Method: 18 official AI/data/analytics acquisition announcements (Jun 2024–Sep 2026) from Databricks, Salesforce, IBM, Microsoft, Snowflake and Qlik, coded for explicit mention of team, technology, data model, customer and distribution evidence; plus the official FTC study PDF of 819 non-HSR transactions (2010–2019) and one secondary summary of J. Daniel Kim's startup-acquisition workforce study. Purposive case set including some mature comparators (Informatica, Hakkoda); counts describe a pattern in buyer-authored announcements, not a population estimate or causal proof. Duplicate secondary retellings of the FTC study were cut; evidence excerpts in the matrix are trimmed to fit.