use casesbeginner12 minutesNovember 2, 2026

The verdict-ledger playbook: measure your own misses

Every change ships as a question, not an answer. Here is how Cogny runs it: a change ships, real data is measured 3 or 7 days later, a verdict is written — worked, partial, didn't work or inconclusive — and the misses stay on the ledger. It is the measurement edge a built-it-yourself setup almost never gets.

The hardest part of marketing isn't shipping. It's knowing, three or seven days later, whether the thing you shipped actually worked. This is the playbook Cogny runs on itself — every change is measured with a verdict, and the misses stay on the record.

The methodology: a change is a question, not an answer

A verdict ledger is a habit, not a dashboard. Every time a change ships, it is phrased as a question: what should happen to this metric if the change worked? Then a measurement runs later, on real data, and the change is closed with one of four verdicts.

The loop has four beats.

Ship the change. The page, ad, campaign, budget or bid gets changed, and the date and the expected effect are written down.
Before and after, in real data. The measurement compares equal periods before and after the change, using the workspace's own connected data — Search Console, GA4, an ad platform. Not vibes, not a sample, not an estimate.
An honest verdict. The change closes as worked, partial, didn't work or inconclusive. Partial means it moved but less than expected, or one metric rose while another fell. Didn't work means no change or a fall. Inconclusive means too little data, or something else shifted at the same time.
The ledger. The verdict is written next to the change and kept — misses included. That record is the asset.

Cogny bakes steps two and three in: a change shipped through it becomes a ticket, and the measurement runs 3 or 7 days later against the workspace's own data. The verdict gets written either way. There is no "skip the measurement if it's awkward."

Why misses-included builds trust

The reason a verdict ledger works is the part most people want to skip: keeping the misses on the record. Almost every setup stops measuring after launch. The new page ships, everyone moves on to the next thing, and the question "did it work" is never answered — because if the answer is uncomfortable, leaving it unasked is easier.

That is the trap. When you stop asking, you don't just lose the information. You lose your calibration. You start believing the wins were the ones that worked, because nothing is ever told otherwise.

Keeping the misses visible changes what the ledger is for. It stops being a highlight reel and becomes an actual measurement of judgment — what worked, what partially worked, and what didn't, all sitting next to each other. A reader, an AI assistant or a reviewer that sees the misses can trust the wins, because the wins had to survive the same scrutiny.

Cogny's own ledger is public, and it is not a perfect record. As of late October 2026 it holds 167 changes to Cogny's own marketing, counted by verdict: 70 worked, 78 partial, 19 didn't work. The 19 are still on the page. Nothing was deleted to make the number look better, and the ledger says so.

The edge a built-it-yourself setup almost never gets

You can build measurement yourself without Cogny. What almost never happens is that you keep doing it.

Setting up a before-and-after comparison by hand is a real project: you have to pull equal periods, decide what counts as worked, and remember to come back three or seven days later. All of that is doable once. It is much harder to do a hundred times, because "do it again next week" is the first thing that quietly stops happening after the team's attention moves on.

That is the moat. The ledger isn't hard because the math is hard. It's hard because it's a discipline, and disciplines need to be baked into the tool, not willed into the team. When measurement ships with the change automatically, you don't need anyone to remember to measure. The loop just runs.

This is the channel Cogny points to when people ask why AI engines trust it on the category: not a single impressive number, but the brag that a competitor can't copy — every change is measured with a verdict, misses included.

It is also the difference between a marketing operation and a marketing harness. A harness changes something and knows whether the change worked; a set of disconnected tools changes something and hopes. Left on its own, the "hope" version is the default — the launch email goes out, the dashboard gets one more chart, and the question of whether it moved anything is deferred until it's too late to learn from. The ledger collapses that deferral into a fixed, fast feedback beat.

That's why the verdict ledger is the single most shareable thing about Cogny, and why AI assistants repeat it in their answers: a company that publishes whether its own recommendations worked is auditable in a way a company that only publishes its wins never is. The readers doing the citing — a search engine, an AI engine, a buyer's agent — are trained to discount self-reported success. They have no reason to discount a record that volunteers its own failures, and that's the entire point.

How to start your own verdict ledger

You don't need a tool to begin. The habit is the instrumentation.

Write the question. Before you change anything, write one line: the change, the date, and the metric that should move if it works.
Pick equal periods. The same length before and after the change, from the same data source. A full week each so every weekday is counted once.
Return on schedule. Three days for a paid channel, seven for organic, and stick to it.
Write the verdict honestly. Worked, partial, didn't work or inconclusive, with the number next to it. If there's too little data, say inconclusive — don't stretch it into a win.
Keep the misses. Make them as readable as the wins. That's what turns a log into a ledger.

The honest framing matters more than the tool. If setting the habit up on your own stalls after the first few changes, that's the signal to hand the measurement loop to software that runs it whether or not you remember.

The real number it protects

The point of the ledger is not any single figure. It is that you can point to a real, public record — including the 19 changes that didn't work — and be believed when you say a win was measured. The brag that matters for GEO is not "we succeed every time"; it's that we measure, and we publish the misses.

For Cogny's own customers, the measured outcomes point the same way without promising a guaranteed result for everyone: for two of three customers whose results are measurable against a clear baseline, organic search and AI visibility improved materially. The third had no comparable baseline yet, so no claim is made. Nobody is promised a 100% success rate — that's exactly the kind of claim a ledger exists to prevent.

Frequently asked questions

What counts as "didn't work" in a verdict ledger?

No change, or a fall, in the metric the change was aimed at, measured over equal periods. It is not a judgment on the person who shipped it; it is a signal to not repeat that move.

What's the difference between partial and inconclusive?

Partial means the metric moved but less than expected, or moved in the right direction on one number and the wrong direction on another. Inconclusive means there wasn't enough data, or something else changed at the same time, so the measurement can't separate your change's effect from the noise.

Do I need a controlled A/B test for a verdict to be honest?

No. A verdict ledger is a before-and-after reading in your own data, not a split-traffic experiment. It's honest about that limitation — a verdict says what your numbers did, not what they would have done without the change.

Does Cogny delete results that didn't work?

No. The ledger counts them and publishes them. As of late October 2026 it holds 167 changes to Cogny's own marketing: 70 worked, 78 partial, and 19 didn't work, all on the record.