Decide what counts as a conversion before deciding what to bid on it.
For a brand that routes buyers to dealers, specifiers, or a sales team, counting only ecommerce purchases starves the bidding algorithm. It never leaves its learning phase, and the budget buys guesses. The fix is architectural, and it happens before any campaign goes live.
laurenmilligan.pro — MEAS-01Considered purchases do not produce enough conversions to optimize on.
A premium product sold through a mixed channel behaves badly under default measurement. Some customers buy online. Many are routed to a dealer or a rep. Trade and specifier buyers behave differently again, and they are usually the most valuable of the three.
If only completed ecommerce transactions are counted as conversions, the ad platform receives a trickle. Smart Bidding needs volume to learn, and a trickle is not volume. The account underperforms, and the diagnosis is usually written up as a targeting or creative problem when it is neither.
The fix is to treat qualified engagement as the conversion currency. A dealer locator search is a genuine, high-value business outcome, and it happens far more often than a direct online purchase. So does a technical download by a specifier, or a trade programme application. Those are the events worth bidding toward, and they need to be designed, instrumented, and valued deliberately rather than picked up by default.
Primary conversions and secondary signals are not the same thing
Primary conversions are optimization targets. The platform bids toward them. They should be events with real commercial meaning: a dealer contact, a trade inquiry, a quote request, a phone call, a purchase.
Secondary signals are measured and reported but never bid on. Scroll depth, video completion, and newsletter signup are useful for audience building and diagnostics. Optimizing toward them teaches the algorithm to buy cheap, shallow traffic, which is a failure mode that looks like success on a dashboard for about six weeks.
Every event is verified against the live site before it is promised.
Measurement plans are easy to write and easy to write badly. The common failure is a plan full of events that sound right and turn out not to be implementable, discovered in week three of a build when the timeline has no room left.
So every proposed event is checked against the live site first, and marked with its actual status. Confirmed means the underlying page, form, or function exists today and the event is implementable. It does not mean implementation is trivial. Anything that cannot be confirmed is marked unconfirmed and held out of scope pending a conversation, rather than quietly included and quietly dropped.
| Event | What it captures | Status |
|---|---|---|
find_a_dealer_search |
Dealer locator search submitted | Confirmed. Runs as a JavaScript module with filtering, so tracking requires event listeners rather than a page-view trigger. Scoped accordingly. |
trade_inquiry_submit |
Dealer application or trade programme form | Confirmed to exist, not currently measured. Hosted on the marketing automation platform, carrying no Google measurement code. |
technical_download |
Cut sheet, DWG, manual, or catalogue download | Confirmed. Strong specifier and dealer intent signal. |
quote_request |
Quote or consultation request | Not confirmed. No dedicated form found. The general contact form may currently serve this purpose. To confirm at kickoff. |
The finding this method tends to surface
The single most common discovery is that a company's highest-value business-to-business conversion is invisible to the advertising platform. Dealer applications and trade forms are frequently hosted on a marketing automation platform rather than on the storefront, and those templates routinely carry no analytics or tag manager code at all.
The platform records the submission in its own system. As far as Google Analytics and Google Ads are concerned, it never happened. If a company has ever reviewed marketing performance and concluded that it was not producing trade interest, this is a plausible reason why.
The fix has two parts, and the second one is the valuable one. Deploy measurement to the form templates, which makes the submissions visible. Then capture the click identifier on submission and write it onto the lead record, which is the prerequisite for later optimizing toward the dealers you actually sign rather than toward the forms people fill in.
The unglamorous half.
Most established sites have a functional measurement foundation and a layer of accumulated debt sitting on top of it. None of it is unusual and none of it is urgent in isolation. It all becomes material the moment paid traffic is flowing, because each item degrades the quality of the data the bidding algorithm is using.
- Legacy properties still firing. Retired analytics tags left in the container, collecting into a property nobody reads.
- Development artifacts in production. Debug scripts that were meant to come out before launch and did not.
- Consent configuration nobody revisited. Consent Mode defaults set once and never reviewed, which affects both compliance posture and conversion modelling.
- Duplicate and overlapping tags. The same event reported by two mechanisms, inflating counts in a way that looks like growth.
- Coverage that does not match how customers actually convert. The category above.
These are cheap to fix and expensive to leave. They also happen to be the fastest way to establish whether a measurement partner has actually looked at your site or is working from a template.
Consent is named, not assumed
Deploying more tracking onto a site with no consent mechanism is a decision, and it should be made in writing by the people who carry the risk, not absorbed silently into a scope of work.
Where a site has no consent management platform, the options are stated plainly: bring the fix into scope, or defer it in writing with the consent-dependent features held back until it is resolved. Both are defensible. Pretending the question does not exist is not.
No numeric targets before there is data to set them against.
Proposing a target cost per acquisition before measuring anything is a sales tactic, not a strategy. If there is no clean conversion baseline, any number in a proposal is invented, and everyone in the room knows it.
- Days 1–30 Collect a clean baselineEvery confirmed event, running live, measured against the browser-only starting point so the difference the infrastructure makes is known rather than claimed.
- Day 30 Joint target-setting sessionProposed cost per qualified contact, cost per trade inquiry, blended cost per qualified engagement, and return on ad spend where purchase data supports it. You confirm or adjust against business economics only you know.
- Quarterly Targets reviewed in the strategy workshopAgainst actual performance and against what the business now needs, with the written plan updated to match.
The one commitment that can be made before data exists
Within 30 days of the build completing, you will be able to answer "what did our ad spend actually produce" with a specific number.
For most accounts arriving at this point, that is not currently possible, and it is a more useful promise than a cost per acquisition figure guessed in advance.
Find out what is currently invisible.
A preliminary technical review checks what is deployed, what is instrumented, and which of your most valuable conversions the ad platform cannot see. It happens before any recommendation and it is not billed.
Request a technical review