Reproducing any row in the index by hand

The formula

The composite index is the sum of each axis subscore multiplied by its weight, divided by 100. The weights sum to exactly 100, so the index is bounded at 0 and 100 and is reported to two decimal places.

There is no curve, no normalisation against a moving cohort mean, and no editorial adjustment. If a product improves on an axis and nothing else changes, its index goes up by exactly the subscore change multiplied by the weight.

Worked example

Take a product scoring 80 on agent performance, 90 on reliability, 75 on scalability, 70 on SEO and GEO, 85 on API and MCP, 70 on integrations, 80 on design, 75 on speed, 70 on value and 60 on code ownership.

The contributions are 14.40, 16.20, 9.75, 8.40, 7.65, 5.60, 5.60, 5.25, 3.50 and 1.80. The sum is 78.15, which is the number that would appear in the index column.

Where each subscore comes from

Two of the ten are computed directly from measurement rather than assigned. Reliability is the strict pass rate over the 45 recorded executions. Speed is the fastest median in the cohort divided by the product median, expressed as a percentage. The other eight are scored against published criteria and are open to challenge with evidence.

Challenging a number

Send the evidence to the lab address on the contact page. If a measurement is wrong we correct it, note the correction on the product page and record it in the change log. We have done this twice.