In lead generation, the form is the beginning, not the end. Among 100 form submissions there is spam, browsers, students, competitors, and a handful of customers. Google sees them all as identical and Smart Bidding chases them all equally. The answer is not to filter them by hand: it is to score every lead and send the score back to Google as a value, so it learns to bring in the good ones. This lesson is the method; the Suite's Lead Rating is its implementation.
Why the form is not the conversion
A real anonymised case (module 8 of the Basic level): an account with a CPA of €39 "meeting" a €40 target. Once the leads were scored: 75% were browsers or spam; the CPA per useful lead was €160. Google had spent months optimising towards the easiest form. Once it was taught the difference (a high value for the good ones, a low one for the bad), in two months useful leads rose by a third on the same spend.
The three layers of the score
A score from 0 to 100 per lead, adding up three layers that become known at different moments:
A · Behaviour on the site (known instantly)
What they did before submitting the form: pages viewed, time, sections (pricing, case studies, contact), device, time of day, whether they came back, how fast they filled the form in (a form submitted in 4 seconds is a bot), whether the fields are coherent (a valid phone number, a company email versus a disposable one). This is the layer that catches spam and browsers without waiting for anybody.
B · Positive actions (minutes or hours)
They confirmed the email, answered the call, downloaded the quote, replied on WhatsApp, booked an appointment. Each action raises the score.
C · Source campaign and channel (historical)
Leads from certain campaigns, keywords and areas close more often. A ranking of campaigns by historical close rate (from the CRM) weights the score: a lead from the campaign that closes at 30% starts higher than one from the campaign that closes at 5%.
(A fourth layer, D, is reserved for business signals: sector, company size, product of interest.)
Recalibrating with the CRM
The score is a prediction; the CRM holds the truth: which leads became customers. Every month, comparing score against outcome recalibrates the weights: if leads who "viewed pricing" close twice as often, that signal weighs more; if "downloaded a PDF" predicts nothing, it weighs less. This is supervised lead scoring: it learns from the business, not from hunches.
And the definitive signal: the sales rep's manual rating (bad / good / sale) per lead, feeding the model as ground truth.
Sending the score back to Google
Three offline conversions per lead (lesson 1), uploaded with the GCLID:
| Conversion | When | Value |
|---|---|---|
| Raw lead | The same day | The score (or an expected value derived from it) |
| Qualified lead | When the CRM confirms it | The average value of a qualified lead |
| Sale | On closing | The real amount |
Smart Bidding therefore receives, from day one, conversions with different values according to estimated quality, and later the truth. With value-based bidding, it optimises towards leads that look like the ones that are worth something.
Side effects to expect
- "Conversions" go down (the bad ones stop counting the same): it is not a drop, it is honesty.
- CTR may fall slightly (fewer clicks from browsers): good.
- CPA rises at first (fewer things are counted) and the cost per useful lead falls within weeks.
- The "cheap" campaigns that brought in rubbish lose budget: that is the correct reallocation.
Requirements
- The GCLID (or email/phone with enhanced conversions) captured on every lead.
- A CRM (or a spreadsheet) with each lead's outcome.
- An upload mechanism (a scheduled sheet, the API or a script).
- An agreement with the sales team: marking each lead's status is compulsory — without it there is no recalibration.
💡 Ninja trick: the Suite's Lead Rating implements this whole lesson: a snippet on the site captures layer A and the GCLID; the positive actions (B) arrive by webhook from the CRM or from email; the campaign ranking (C) is calculated from history; the score is uploaded to Google as the raw conversion the same day, with qualified and sale following later, by bulk upload. The leads never leave the client's own spreadsheet (GDPR); all Google receives is the GCLID with its value. And the sales rep's manual rating checkbox is the ground truth that recalibrates everything.
What you should remember
- The form is not the conversion; the customer is. Score every lead.
- Three layers: behaviour (A), positive actions (B), source (C); monthly recalibration with the CRM and manual ratings.
- Send Google raw with its score, qualified, and sale with its amount by GCLID.
- Expect fewer "conversions" and more customers; cost per useful lead is the metric.