What Is a Readout?

Every sprint ends the same way. We write up what happened and we pick the next move.

That write up is the readout. If the hypothesis is the bet, the readout is the accounting. No fluff. No deck of screenshots that says nothing. Just: did the metric we tied to the bet move, what else moved or did not, what we now believe about your buyer, and what we will test next.

If you have been burned by vague agency reports, this page is for you.

What a readout is

A readout is a written judgment at the end of each sprint, tied to the hypothesis from the start of that sprint. It is not a dashboard export. It is not an audit. It is not a strategy doc.

It answers one primary question first. Did the metric we tied to the hypothesis move in the timeframe. That is the headline. Everything else is context for that headline.

Then it shows context. Secondary steps in the funnel. Segment cuts that matter. What we learned about the buyer belief behind the hypothesis.

And it closes with a decision. The next hypothesis, with its own metric and timeframe. Every readout ends with so next we test... in explicit form.

Written, short, and the basis for the next sprint decision. That is the bar. If a readout does not change what we do next, we wrote the wrong readout.

The loop it lives in is always the same: hypothesis → implement → measure → readout. That cadence comes from how sprints work. The shape of the bet comes from what a hypothesis is. The readout is where both get judged.

What you get at the end of a sprint

Four parts. Every time. Same order.

One, the primary metric vs the hypothesis. Moved, did not move, or directional. We name the metric we picked at sprint start, the window, and the outcome. Example shape: Example: primary metric, checkout start rate, illustrative. If we name a real metric, it will be labelled with window and segment. No floating percentages.

Two, what else moved or did not. Secondary steps in the funnel. Funnel shorthand when labelled: Example: add to cart → checkout → purchase. If add to cart went up but checkout did not, that tells us the belief was partly right but the friction moved down the funnel. If nothing moved, that also tells us something. We show both. No cherry picking.

Three, what we learned about the buyer. Did the belief behind the hypothesis validate? Example shape: we believed buyers did not get the offer above the fold. After the change, hesitation there dropped but trust signals still got skipped. That is a learning even if the primary metric was flat. Learnings compound. They are not consolation prizes.

Four, the next hypothesis. Explicit, with its own belief, change, metric, and timeframe. Not we will keep optimizing. A real sentence you can challenge. That is the whole point. The readout does not just report. It recommends the next bet.

We keep it short enough to read in one sitting. If you need the raw numbers, they are attached. But the judgment is in the writing.

Site lift vs Meta spend efficiency (counted separately)

This is the honesty hinge. And it is where most reporting gets fuzzy, so we keep it sharp.

Site lift is more purchases from the same traffic. You send 1000 visits, you got 12 purchases before, you get 16 after. Same traffic, better store.

Meta spend efficiency is lower cost per result from the same site. You spent the same, you got more purchases because targeting, creative, or budget moved. Same store, better acquisition.

They are different levers, different owners, and we count them separately. Always.

Why? Because a Meta change can take credit for a site change if you mix them. Or a site improvement hides behind a bad Meta week. When you blend them, no one knows what to do next. When you separate them, the decision is clear.

So in every readout, site conversion and Meta efficiency are reported apart. We name the handoff point, where Meta hands off to the landing page and PDP. That seam is where most leakage lives, and it is exactly where ownership matters. For the split itself: you own Meta, we own pages, tracking, and readouts. That is the line from how sprints work. We hold it in reporting too.

We do not claim attribution perfection. We do not claim we fix Meta accounts. We do not mix ROAS into site CVR. If a readout needs to show both, it will say Example: Meta efficiency vs site conversion, illustrative split and label it.

One more nuance. We often talk about funnel as a revenue path (acquisition to landing to PDP to checkout). That is helpful shorthand. We use it when labelled Example. But in the readout we will always say which step we mean and which owner it belongs to. No blurred lines.

How Sprint 2 is chosen from Sprint 1 data

Sprint 2 is not preplanned in detail. The backlog is ordered, but the readout reorders it. That is the system.

Three cases, and they are the only three.

Case one, the metric moved and the belief validated. We double down or extend. Could be expand to another template, tighten the message, or roll to the next segment. The hypothesis for Sprint 2 builds directly on the win.

Case two, the metric did not move but we learned why. Maybe the belief was off. Maybe the change was too small. Maybe the friction moved one step down the funnel. We pivot to the next most important hypothesis based on that learning. The readout states the learning plainly, not as an excuse. Then it names the next bet.

Case three, nothing moved and we learned little. That happens. Smaller traffic, noisy window, or a belief that was too far from the real friction. We broaden the lens. Look at the step before or after in the funnel. Check device or traffic source splits. Pick the sharpest new hypothesis we can now form and run it. We do not pretend it was directional when it was not.

In all three cases, the readout does the work of choosing. You do not get a status update, you get a decision: here is what we now believe, so next we test this.

That decision logic is why one at a time matters, which is covered in both what a hypothesis is and how sprints work. One clean bet keeps the choice for Sprint 2 clear.

What directional and up to mean (honesty)

Founders have heard these as weasel words. So we define them. We use them the same way every time.

Directional means the signal points the same way across related steps or segments but is not yet at a threshold we would call conclusive. Example shape: add to cart and checkout start both nudge up, mobile and desktop both nudge up, but neither move alone would clear our bar for proven in this window. It is still useful. It tells us the next hypothesis. We call it directional so you know it is a nudge to choose from, not a win to celebrate.

If you see directional in a readout, it will be labelled with what points which way and why it is not yet conclusive. Smaller traffic stores will see directional more often. We name that plainly. It is not hedging. It is respecting sample size.

Up to means the largest observed move in a specific slice or window, not the average. We use it only when labelled with the slice and window, and we show the average alongside. Example shape, illustrative only: Example: up to X% in Y segment over Z days, average was W%, illustrative. We never write up to without the slice, the window, and the average in the same paragraph. If you see up to without those, call us on it.

Both words exist to keep us honest about uncertainty without being vague. Every readout labels them explicitly so no one mistakes a directional nudge for a proven lift.

And if nothing moved at all, we say so. No softening. The next hypothesis is still there, and the learning still counts.

---

What to do with this page. When you read a Convfetti readout, look for four things: primary metric vs hypothesis, what else moved or did not, what we now believe, and the next hypothesis with its metric. Then check that site and Meta are counted separately and that any directional or up to is labelled with slice, window, and average.

If you have not seen them yet, what a hypothesis is gives you the shape of the bet, and how sprints work shows you the loop these readouts sit in.

See engagement plans on our pricing page for where these readouts live.