Steganox Engineering · 30 April 2026
The first two seconds
A visitor arriving from a search result has no investment in your page. If nothing meaningful has rendered by the time their patience runs out, they leave, and they do not report it, so the loss never appears in any dashboard you look at. You see the traffic that stayed.
This is why performance work often produces conversion improvements that look implausible. You are not persuading more of the same people. You are being seen by people who previously never got far enough to be persuaded.
Where the weight usually is
In our audits the culprits repeat. Uncompressed hero imagery, often several megabytes for something displayed at a fraction of its native size. Third-party tags: analytics, chat widgets, heat maps, ad pixels. Each individually reasonable, collectively a second and a half of blocked main thread. Fonts that block text from rendering until they arrive. And increasingly, a heavy interactive or 3D element loaded on the critical path because it looks impressive in a design review.
That last one is worth dwelling on, because the fix is rarely to remove it. It is to load it after the page is usable. The visual impact is identical; the difference is that the headline, the navigation and the call to action have already rendered while the decorative layer is still downloading.
Budgets, not one-off fixes
An optimisation pass is worth doing and will regress within two quarters unless something prevents it. We set explicit budgets: a page weight ceiling, a limit on blocking scripts, a target for interaction latency. Then we enforce them in continuous integration, so the pull request that would breach one fails before it merges.
It changes the conversation. Instead of an annual argument about why the site got slow again, adding a new tag becomes an explicit trade against something already there.