Deploy impact · production
Five findings from the first full week after the V2 pixel rolled out to production. Three of them explain a nearly five-fold jump in tracked traffic. The remaining two are the reason we trust it.
Pixel visits per day · Jul 15 → Jul 28
The tallest bars on the left are the busiest days of the previous week — the peak the old pixel was already catching. After the deploy, ordinary weekdays look like this. The change is not subtle; it is the moment measurement started matching behavior.
If only the race condition were fixed, both numbers would have grown in lockstep. They didn’t. The gap between household growth and visit growth is real users’ in-session navigation — every click through the app that used to leave no trace at all.
That gap is where the SPA refire lives, and it’s the entire reason the top-line number is 4.8× and not 1.3×.
Growth by dimension
3.7×ratio gapThe extra fires are not landing on the same handful of URLs. They’re spreading across a much broader universe of URLs, which is what session navigation looks like from the server’s side — new routes appearing, not old routes hit twice.
URLs vs visits
1:1growth parityBefore V2, an average household appeared with roughly two distinct URLs across the entire weekday. After V2, the same household is generating pixels for almost eight distinct URLs per day. This is arguably the most important number on the page: it says the previous data was seeing the surface — the first landing page, maybe one more — but almost nothing of the in-session journey that came after.
URLs per household
3.7×session depthThis is the dedup validator. If V2 were over-firing on the same URL — e.g. pushState plus replaceState without dedup, or the 50ms debounce failing to coalesce — this ratio would spike past 3. It didn’t. The location.href !== lastUrl guard is doing its job.
A half-percent shift across a week of weekday traffic is not a signal — it’s indistinguishable from noise. Every extra visit V2 captured landed on a distinct URL: the hallmark of session navigation, not repeat fires on the same page.
Where we’d see over-firing
+0.01×essentially flatHousehold resolution didn’t degrade under the higher volume. The mapping from IP to household is holding steady; the extra fires aren’t confusing the identity layer.
Identity layer stability
+0.00×unchanged