The number moved. Was it you?

You shipped the new onboarding on Tuesday. By Friday sign-ups are up eleven percent week-over-week, and the team is pleased, and someone drops a green chart in the channel. It feels like the loop closed — you had a hypothesis, you shipped it, the number went up. But sit with the chart for a second and the good feeling gets shaky. Up since Tuesday. Up compared to a quiet week. Up while marketing was also running a push and a competitor was also having an outage. The number is higher. That is genuinely all you know.

Here’s the uncomfortable part: higher is a level, and a level is almost useless for telling whether you did anything. A level is where the line happens to be sitting when you look at it, and a line sits where it sits for a hundred reasons — the season, the day of the week, a link that went around, a price a rival changed. Reading a level and crediting your deploy is like walking into a room that’s warm and taking credit for the sun. The warmth is real. Your involvement is not established.

Higher is a level. You need a break.

What you actually want isn’t a higher number. It’s a break — a place where the line changes shape, and does it right where your change landed. A metric that was drifting along and then kicks up the day you shipped is telling you something a metric that’s merely elevated never can. The level answers “where is it now”; the break answers “did something happen here.” Only the second one has your name anywhere near it.

This is why the honest version of the question is so much narrower than the celebration makes it sound. Not are sign-ups up — they’re always up or down against something. The question is: did the series do something different at the exact moment my change reached production, that it wasn’t already doing? That’s not a marketing question or a vibes question. It’s a question about the shape of a line at a specific timestamp, and it has a real answer.

A rising number is an invitation to check, not a medal to accept. The tide lifts everyone in the harbor; the wake is the only part you made.

The deploy is a dated event. The metric is a noisy line.

Once you frame it that way, the engineering problem comes into focus, and it’s smaller and more stubborn than “analytics.” You have two objects. A deploy is a precise, dated event — it happened at 3:11pm on the fourteenth, no ambiguity, already logged by the machine that shipped it. A business metric is a noisy, continuous line — sign-ups a day, activation, revenue, jittering around from a thousand causes, never still. The whole game is welding the first onto the second: drawing the dated event as a vertical line straight through the noisy one, on a single shared time axis, so a human can look at where they cross and read whether the shape broke there.

That sounds trivial. It is exactly the thing almost no tooling does. Your deploy tracker knows when you shipped and nothing about what happened after. Your analytics dashboard knows the metric in exquisite detail and nothing about your deploys. They live in different tabs owned by different teams, and the one act that would make attribution readable — putting the event and the outcome on the same picture — is the one act neither tool is built to perform. So the question never even gets asked, because asking it means pulling two timestamps into alignment by hand, and nobody does that on a Friday.

Alignment is necessary, not sufficient — and that’s the discipline

code-results does that welding, and the temptation with a tool like this is to oversell what the welding buys you. Lining a deploy up against a metric that moved is still a correlation. Ship into a holiday and revenue climbs for reasons that were coming regardless; ship a genuinely great change the week a rival undercuts you and it reads as flat. Timing does not manufacture cause. Anyone who tells you their tool draws a deploy on a chart and therefore proves the deploy worked is selling you a story with good typography.

What alignment actually buys is that the question becomes askable at all, and it kills the two ways teams routinely fool themselves. It stops you crediting your deploy for a rising tide — the line was already climbing before you shipped, and now you can see that. And it stops the crueler failure, the real win that drowns: a change that genuinely worked but landed in a loud week, so nobody ever saw the small clean step it put in the line. Put the event on the noisy line and both of those become visible in a glance. You still have to go investigate the break — pull the segment, check the confounds, rule out the campaign. But now there’s a specific, dated break to investigate, instead of a warm room and a shrug.

Why a studio this small cares

We didn’t arrive at this from theory. We build and run our own products — that’s the one rule the studio runs on — and operating a thing is where you learn, painfully, that the number being up is not the same as your work being the reason. We’d ship something we believed in, watch a metric wander, and have no honest way to say whether the wander had anything to do with us. So we built the smallest thing that made the question answerable: the deploy, drawn on the line.

It won’t hand you causation. Nothing honest will. But it will stop you accepting medals for the tide, and stop you missing the wins you actually made — and between those two, most of the self-deception in shipping quietly lives. The number moved. That was never the interesting part. Whether you moved it is the only thing worth knowing, and you can’t know it until the deploy and the line are drawn on the same page.