The Millisecond · Issue #1 · 2026-08-12

Four insurance sites, average 97. Here’s the habit.

The habit that separates the four insurance sites in my portfolio from every legacy broker site in Europe. It has nothing to do with the code.

Four insurance sites currently live in my portfolio. Their PSI mobile scores, measured this morning:

morax.gr — 99 (my operator brand, insurance brokerage)
imperial.gr — 96 (client, brokerage)
insurance-greece.com — 98 (client, comparison engine)
insurancedaily.gr — 98 (my operator brand, news publication)

Average: 97.75. Median: 98. Meanwhile, if you audit the top 20 insurance broker websites in Athens on PageSpeed Insights mobile today, you will find zero above 60. Most sit between 24 and 45. I have run this audit twice this year, six months apart, and the distribution barely moved.

The obvious question is what technical choices explain the gap. Static site generators? Edge hosting? Modern image formats? All true — and all downstream of one habit that has almost nothing to do with the code.

The habit: every feature starts at zero

When a typical insurance broker asks their agency for a "quote calculator widget," the answer is yes. Then a live chat, yes. Then a Trustpilot embed, yes. Then a Facebook pixel, yes. Then a heat-map script, yes. Then a "compare providers" plugin, yes. Every yes adds 30–200 KB of JavaScript, one to three third-party network requests, and 40–200ms of main-thread blocking. Nobody says no because nobody is keeping the tab.

On the four sites above, every feature starts at zero. Not "no by default" — zero. The feature must actively earn its place by demonstrating a measurable business outcome that exceeds its performance cost. The chat widget doesn't get added because "chat is good." It gets added if we can show that inbound-to-quote conversion improves by more than the LCP degradation from the widget costs in lost session-continuation rate. If we can't show it, the feature doesn't ship. It doesn't stay in the backlog. It leaves the room.

This is the reverse of the industry default, which is "add first, remove never." The industry default produces the PSI 24 broker sites. Zero-first produces the PSI 97 broker sites. Same code stack would give the same result. It’s not about the code.

Why this habit is hard

It's hard because saying no to a feature feels like leaving revenue on the table. It feels like caution. It feels like the safer decision is to add and see. But every addition compounds. A site launches with 12 features. Two years later it has 34. Each of the 22 additions felt reasonable at the moment. The composite is a website that lost 60% of its speed and 40% of its inbound conversions to features nobody could name if you woke them up at 3 a.m.

The insurance vertical makes this especially punishing because insurance shoppers are already skeptical. They arrive suspicious. A slow-loading site is not a small friction — it is a confirmation of the suspicion. They do not stay to see the quote flow. They leave to the competitor whose site loads.

The tell

Here is the diagnostic you can run in 60 seconds on any broker site, including your own. Open Chrome DevTools (F12). Network tab. Filter to "3rd-party." Reload. Count the requests. If the number is above 15, the site is running features that nobody is auditing anymore. If it's above 25, the site has become a JavaScript delivery vehicle that happens to also show insurance information.

On the four sites above, the third-party count is between 2 and 5. Analytics, one payment integration for the quote flow when relevant, and a font file. That's the whole list. The habit produces the number. Every issue of this newsletter is going to come back to that habit in some form. It is the actual work.

Michael 2026-08-12 · Issue #1 · The Millisecond

Get this weekly

New issue every Friday. One insight, ~500 words, always from real portfolio data. Subscribe →