Verdicts: Scale, Hold, or Cut, with the numbers that made them

By the HonestTag team · Published July 12, 2026

Dashboards are good at producing numbers and bad at producing decisions. You can stare at a wall of metrics and still not know what to do with Tuesday's budget. Verdicts close that gap with a fixed, published rule applied to your store's own numbers, and they refuse to speak when the data is too thin to judge.

What a verdict is

A verdict is a weekly judgment per channel, and per campaign once campaign matching is on. It is issued each Monday and then held immutable for the week. That cadence is deliberate: it prevents whipsaw, and it enforces one change per conversion-lag window, because ad results arrive on a delay and reacting faster than the data means reacting to noise.

Every verdict shows its work. It prints the exact numbers that produced it, names which rule in the ladder fired, and gives a re-check date. History is stored, so the track record is auditable later: "we showed Cut on this date, at these numbers." A judgment you cannot audit afterward is just a vibe with a timestamp.

Verdicts are part of the judgment layer, which starts on the Truth tier; see pricing. Correctness is never paywalled; breadth and judgment layers are.

First, the evidence floors

Before any rule runs, the data has to clear a set of floors, each published in numbers. If any one of them fails, the state is No verdict, and the verdict names the failed floor and what unlocks it:

• Fewer than 30 new customers store-wide in the trailing 30 days
• Fewer than 20 first-click new customers for the channel in 30 days (15 for a campaign)
• Under $500 of channel spend in the trailing 7 days ($300 for a campaign). The spend floors are counted in your store's currency, 500 and 300 of it (500.00 EUR on a euro store), and every amount a verdict prints is in your store's currency
• A campaign younger than 14 days
• Unresolved mixed currencies or stale spend data
• No NCAC ceiling resolvable

"No verdict" never means you did something wrong. The floor did not clear, that is all. It usually just means a channel is small or young, and the honest answer at that size is "not enough evidence yet."

The ceiling: your economics, not ours

Scale and Cut are comparisons against a ceiling: the most a new customer can cost you and still be worth acquiring. If you set an explicit NCAC target, that is the ceiling. If you have not, HonestTag derives one from your margins: AOV x gross margin percent x an LTV multiple. Either way, the source of the ceiling is cited inside every verdict, so you always know what you were judged against.

If no ceiling is resolvable at all, the state is No verdict. A verdict without a merchant-owned threshold would be the tool inventing your economics, and we will not do that.

The rule ladder

Five rules, checked in order, first match wins. The order itself is published:

1. Change in flight. You noted a change inside the lag window, so the verdict is Hold and the judgment waits until the window closes.
2. Cut. First-Click NCAC ran more than 25% over your ceiling two weeks running, AND NMER x new-customer count shrank.
3. Scale. NCAC came in at least 20% under your ceiling, AND NMER x new-customer count grew or held.
4. Signals conflict. The verdict is Hold, and both conflicting numbers are named.
5. Everything flat. Hold.

Channel revenue in these verdicts is first-click new-customer revenue, gross of refunds. Refunds carry no per-channel split, so netting them here would be invented precision; store-level MER and NMER stay net.

No scoring, no weighting, no blending, no model learning your preferences. Every verdict is a deterministic, published rule applied to your data, the same rule for every store, every week.

One consequence worth spelling out: NMER decaying while you scale is expected, not alarming. If spend rose, NMER fell, but NMER x new-customer count grew, the verdict is still Scale, and it says the decay happened, in the open, rather than hiding it.

The re-check date

Every verdict ends with a date to look again. That date is not a guess. Once a channel has 20 or more measured orders, the re-check interval is that channel's own median click-to-order lag plus 3 days, kept between 7 and 14. Before that volume exists, a 7-day default applies. The check-in date is one the store's own conversion behavior justifies, not an asserted default that happens to fit a weekly email.

A read, not advice

A verdict is a read of your own data: a fixed, published rule applied to your store's numbers and your targets. It is not advice. The tool cannot see your cash position, your inventory, or your contracts. If the verdict says Cut and you know a container of stock lands Thursday, you know something the rule does not. The decision is yours; the verdict just makes sure the numbers under it are honest and named.

The refusal is the feature

Most attribution dashboards print a prescriptive recommendation on any data that exists: ROAS to two decimal places on $30 of spend, an arrow telling you to raise budget off a sample of four orders. In July 2026, we reviewed 80 tracking and attribution apps in the Shopify App Store; zero published how their numbers are computed, let alone when their numbers should not be trusted.

A judgment engine that knows when not to speak is what makes it trustworthy when it does speak. Every floor, band, and rule above is published in numbers on the in-app methodology page, and the mechanics are on the product page. If a verdict ever surprises you, you can trace it to the exact rule and the exact numbers, and disagree with it on the merits.

Frequently asked questions