How this benchmark was made

This is a self-assessment

Keel’s team wrote the criteria, chose the weights, and scored Keel. No independent analyst, auditor or customer was involved. Nobody outside this project has validated a single score. When a vendor scores itself first in its own study, the correct response is scepticism, and we would rather hand you the reasons than have you find them.

We wrote the criteria against our own feature list

83 of the 92 criteria are traceable to a specific Keel feature ID in our FEATURES.md. Of the nine that are not, four are certifications we lack and five are pricing properties we have. Every criterion in the study is therefore something we built, something we are missing, or something we price well.

A study built from our roadmap will be won by our roadmap. The raw scores put Keel more than a point clear of the field. That gap measures the rubric, not the market.

Competitor scores are floors, not measurements

Competitors were scored from public product pages, analyst summaries, review sites and published case studies. There was no hands-on testing of any competitor product. Commercial vendors routinely ship more than their documentation reveals, particularly in enterprise tiers.

Our methodology specified an “Unknown” marker for capabilities we could neither confirm nor rule out, to be excluded from the average rather than scored as zero. We did not apply it — not once, across 2,576 competitor cells. Every documentation gap was recorded as a hard zero or one.

Assume every competitor here scores higher in reality than it does on this page.

How we scored ourselves, and where we got it wrong

Our rubric defines a 3 as “best-in-class or uniquely strong — generally reserved for the top one or two competitors on any given criterion.”

We gave ourselves a 3 on 87 of 92 criteria. We never once gave ourselves a 2. That is not a plausible distribution, and it means we did not apply our own boundary to our own product.

Re-graded against the running code rather than the feature list, Keel scores 2.39, not 2.89. The largest corrections are in Human-in-the-Loop, Core AI and Integrations, and they are driven by capabilities that are built but not yet wired into the request path.

There is a second study, and it puts us second

A separate internal workbook scored a broader ten-vendor panel that weights market proof alongside capability. On that panel Keel places second at 7.95, behind Datadog at 8.01. We are publishing both.

What we would tell you to do instead

Run a two-to-three week proof of concept against the top two alternatives in whichever pillar matters most to you. No vendor’s spreadsheet — including this one — is a substitute for that.

Check our working

Every score, every criterion, both workbooks — including the one that puts us second.

Download the raw scoring (JSON)