Measuring what clients actually care about
Lighthouse scores do not appear in board reports. Mapping technical metrics to commercial ones is part of the job.
A perfect Lighthouse score has never once persuaded a finance director of anything. “Checkout completion rose four points after the largest contentful paint dropped below one second” has, repeatedly.
Pair every technical metric with a commercial one#
Before a performance project starts, we agree which business number is expected to move and how it will be measured. Page load pairs with conversion or bounce. Accessibility conformance pairs with tender eligibility. Build time pairs with release frequency, which pairs with time to ship a fix.
If no commercial number can plausibly be attached, that is worth knowing early. Some technical work is maintenance and should be argued for as maintenance rather than dressed up as growth.
Measure before, honestly#
Take the baseline before touching anything, from field data rather than a lab run, segmented by device class. Desktop-on-fibre numbers hide the problem: on a recent project the median mobile load time was 4.2 seconds while the lab score was a comfortable 94.
Report the ones that did not move#
On the Halden project, organic traffic rose 84% and enquiries tripled. Time on page fell, and we reported that too, because a catalogue where engineers find the tolerance faster should show exactly that.
- 3.1s → 0.7s Product page load
- +84% Organic sessions
- 3.1x Qualified enquiries
Reporting only the favourable metrics trains clients to distrust the favourable metrics.