Method
How we count, including the parts that make us look worse.
Every attribution tool makes choices about what counts as evidence. Most do not publish them, which is what lets two tools report different revenue for the same week and both sound certain. These are ours. You should read them before you trust a number we show you, and you should ask anyone selling you an alternative for the same page.
The chain
A number is only as good as the weakest link behind it.
Signet counts along one chain, and every number it reports names the link it rests on. Where a link is missing, the number is missing too.
- ClickA person clicks an ad. The ad platform tells us which ad, and that is the last thing in this chain the platform can confirm on its own.
- CaptureYour own site records the arrival, on your domain, with identity you collect. This is the first link that belongs to you rather than to a platform.
- BookThe visitor books a call. The booking carries the captured identity forward, which is what lets a call be tied back to an ad rather than guessed at.
- ReadThe call is recorded and read against a rubric you wrote. Publishing the rubric mints a version, and the call records which version read it.
- BindA payment arrives and is matched to that sale. Only now is there closed revenue, and only now will Signet credit an ad with it.
The chain is why your closed revenue is the basis for every verdict Signet reaches. A click is not evidence that anything worked. A closed deal is.
Whose number is it
We never add our numbers to the platform's and show you the total.
Impressions and clicks come from the ad platform, which is the only party that knows them. Outcomes come from your own site, which is the only place they happen. These are different kinds of fact with different reliability, so they are stored as separate fields carrying their own class and there is no field that sums across them. A blended figure is not discouraged here. It is impossible to express.
- Signet-measuredOutcomes Signet measured on your own site, tied to the exact ad each person clicked. For example, a booking counted here means Signet saw that person arrive from this ad and then book — Meta is not the source of this number.
- Platform-reportedNumbers Meta reports about its own delivery — impressions, clicks, and spend. For example, only Meta knows how many times it showed an ad, so these stay clearly labeled as Meta's and are never mixed into what Signet measured itself.
- Pre-tagging historyClicks from before Signet managed your ad URLs cannot say which placement or device they came from. For example, every placement and device number today is Meta's own report, labeled as history until Signet's URL tagging is switched on.
These are the product's own definitions, imported into this page rather than restated, so the method described here cannot drift from the method used.
Revenue
Money we cannot tie to an ad is three different situations, not one problem.
A single unattributed total is the most misleading number in this category. It reads as a matching failure, and usually most of it is revenue no ad could ever be credited for. Signet splits it, and the three buckets add up to your ledger total exactly.
The split is what tells you whether to fix your tracking or leave it alone. Pre-tracking revenue is not a problem to solve. Organic revenue that stays unbound for months is.
- Ad-attributedMoney you were paid that we matched to a specific ad-driven sale, so its value flows into ROAS. For example, a $500 payment linked to the booked call an ad won becomes the revenue credited to that ad.
- Pre-trackingMoney from customers whose relationship began before Signet started recording your sales, so no ad-driven sale exists to credit. For example, a subscriber who signed up last year still pays every month, but that revenue predates measurement and is not from your ads.
- OrganicMoney from a customer we started measuring but have not tied to an ad-driven sale yet. For example, someone who signed up during the measured period without a recorded booking sits here until that sale is found.
Ad-attributed plus pre-tracking plus organic equals the ledger total. The total never implies a failure it does not describe.
The floor
Below 30 observations you get the shortfall, not the rate.
One sample floor applies everywhere: verdicts, creative rankings, KPI cards, table cells. No surface is allowed its own threshold, because a product with per-screen floors is a product where confidence depends on which page you opened.
An ad with no closes from two leads is not a failing ad. It is an unproven one, and those are different claims. Signet says the second and shows you what it would take to say more: 3 of 30 leads needed.
- Needs more dataNot enough evidence yet
- MaturingStill closing, not final yet
- SolidEnough evidence to trust this
Insufficient states carry a dashed outline as well as a color, so the state survives a grayscale render or a color-blind reader.
The vocabulary
A verdict can say six things. Two of them are admissions.
The list is closed. Signet cannot invent a seventh reading to describe an awkward case, so an ad that has not earned a conclusion gets Needs more data or Maturing rather than a softened version of a real verdict.
Every verdict carries the evidence it rests on in the same view, never a footnote and never a hover. Split is the one people ask about: it means your first-touch and last-non-direct readings disagree, and rather than pick the flattering one, Signet tells you they disagree.
- Scale-readyReady to scale
- HoldPerforming acceptably
- SplitYour two models disagree
- Kill-candidateLosing money
- Needs more dataNot enough evidence yet
- MaturingStill maturing
Rendered by the product's own verdict components, from the product's own closed list of codes. A code that leaves the product leaves this page with it.
Evidence in the open
Every rate shows the sample it came from, at the same size as the rate.
A percentage with its denominator hidden is a rhetorical device rather than a measurement. Signet prints the metric, the count behind it, the unit, and the window, beside the number rather than underneath it in a lighter grey.
This is also the fastest way to audit us. If the evidence line looks too thin to support the verdict beside it, believe the evidence line.
CAC $212 · 48 closed · 90d
CAC $260 · 19 closed · 90d
CAC · 3 closed · 90d
What we cannot count
Four things Signet will not tell you, and will not pretend to.
- Which single headline closed the deal, under dynamic creative.The platform mixes your assets and reports at ad grain. The single-element answer is not in the data anyone receives, so a tool that ranks headlines there is ranking a model of your ads rather than your ads.
- Which placement or device a historical click came from.Clicks from before Signet managed your ad URLs carry no placement dimension. Those numbers stay the platform's, labeled pre-tagging history, until managed tagging is switched on.
- That an ad caused a result, from observation alone.Comparing what already ran is observational. Ads are confounded with audience, budget and timing, so Signet says one performed better than another, never that one caused the difference. A causal claim needs a test the product designed.
- Revenue from a sale nobody recorded.Signet counts payments it can tie to a sale in your system. A deal closed off the record is invisible to it, and the honest consequence is that your bound revenue is a floor rather than a ceiling.
None of these are gaps waiting on a roadmap. Three are properties of what the platforms return, and the fourth is a property of your own records. A vendor who quotes you a number for any of them has invented it.
The trade
You will see fewer numbers here than a competing tool will show you.
That is the whole trade, stated plainly so nobody discovers it in week three. Every figure Signet withholds is one another tool would have printed with a confident font and no denominator. The ones it does print, you can act on, and the state marked solid means what it says: enough evidence to trust this.
Read the method before you trust the number.
Every module is part of one platform. You get all of it, and the tiers differ by scale rather than by which capabilities are switched on.