How we score
A 0-to-100 read on a vendor's innovation posture, computed the same way for every one of the 700+ vendors we track, from information anyone can observe. It is a starting point, not a verdict, and not a promise that a vendor fits your store.
What goes into it
Capability depth
Rewards genuinely advanced capabilities, with rarer ones weighted higher than table stakes. A long list of common features does not inflate the score.
Active development
Recent product updates and real market activity such as launches, funding, and acquisitions. Stagnant vendors score lower.
Integration fit
How many tools it connects to, and how deeply. Two-way integrations count for more.
Ease of evaluation
Published pricing, API documentation, certifications, and complete listings. Transparency scores higher.
Focus
A specialized tool scores higher than a do-everything platform spread thin.
Points add to 100. The same formula runs for every vendor on a regular schedule.
How we read a vendor's AI
- We separate what a vendor's AI claims from what we can independently confirm. A capability we read on a vendor's own site is labeled model-detected. One we can corroborate against other evidence is labeled verified.
- We look for corroboration in artifacts a marketing page cannot easily fake: product documentation, priced AI plans, dated release notes, named AI subprocessors, AI engineering roles, and named customer results. The more of these that line up, the higher the corroboration.
- We cap autonomy at what the evidence supports. When a vendor's materials imply the software acts on its own but we cannot confirm it, we show it as supervised, with a note that vendor materials imply higher and remain unverified.
- We never relabel a vendor as more autonomous, or more AI-driven, than its own evidence shows.
What you cannot do
- You cannot buy a higher score. There is no sponsorship, no featured placement, and no pay to rank.
- Vendors do not score themselves. We use public, observable data, not vendor-submitted marketing.
- No vendor sees who is looking. The score is the same whether you are watching or not.
Where it falls short
- It measures innovation posture, not whether a vendor is the right fit for your store.
- When we can see less about a vendor, the score stays conservative instead of guessing.
- We recalibrate the formula as the catalog grows, and we say so when we do. Last recalibrated June 2026.
See how your stack scores
Connect your AI and check your vendors against the same independent method. No vendor ever knows you looked.
Independent & unbiased — scores can't be bought. Works with Claude & ChatGPT.