The Signal Repair Playbook
How to Stop Training Smart Bidding on the Wrong People and Start Getting Google Ads to Optimise Toward Buyers
ps - if you want this done for you, book a Lead-Leak Analysis with me.
Most B2B Google Ads accounts are not broken at the campaign level.
They are broken at the signal level. Smart Bidding is doing exactly what it was told. It is just chasing the wrong outcome.
The algorithm is powerful. But it learns from whatever conversion data you feed it. If that data is "someone submitted a form," it will find more people who submit forms. Not more people who become clients. The signal shapes the outcome, and most accounts have a weak signal.
This playbook gives you the exact framework to identify where your signal breaks, fix the inputs, and close the loop between what Google learns and what your sales team actually accepts.
Here is what you are about to read:
Why "form submitted" is almost always the wrong primary conversion signal in B2B
How to audit whether your current setup is feeding Smart Bidding useful or garbage data
The three inputs that improve bid signal quality without rebuilding your account
How to push qualified lead feedback downstream so the algorithm learns from buyers
What conversion volume your account needs before Smart Bidding can pattern-match reliably
How to structure conversion actions so Google optimises for commercial outcomes
PART 1: Why Your Conversion Signal Is Probably Training the Algorithm Toward the Wrong People
The "Form Submitted" Problem
When this happens: You set up a single conversion event for form submissions, hand it to Smart Bidding, and assume the algorithm takes it from there.
It does. It just does not know the difference between a qualified buyer and someone who clicked by accident.
"Smart Bidding optimises toward whoever converts. If your conversion is 'form submitted,' the algorithm finds whoever submits forms. That includes students, competitors, job seekers, and anyone who misread your ad."
Smart Bidding is not the problem. The conversion definition is.
In B2B, a form fill is not a commercial outcome. It is the start of a qualification process. Treating it as the primary signal tells Google to optimise for the top of your funnel, not the bottom.
❌ Bad: One conversion action called "Contact Form Submitted" or "Generate Lead" set as the primary goal. Smart Bidding trains on every submission regardless of quality.
✅ Good: A layered conversion structure where the primary signal is a qualified enquiry or downstream action, and form fills are tracked as a secondary (non-optimised) event.
Your conversion definition is a set of instructions. Make sure you are giving the right ones.
The Volume Threshold Problem
Before fixing the signal, check whether you have enough data to work with.
Smart Bidding requires a minimum conversion volume before it can pattern-match reliably. Google's general guidance is 30-50 conversions per month at the campaign level for Target CPA, and higher for Target ROAS.
Below that threshold, the algorithm does not have enough signal to learn from. It guesses.
Below 30 conversions per month: While the algorithm can sometimes run on 15–20 monthly conversions, campaigns generating fewer than 30 are generally operating blind.
30-50 conversions per month: Smart Bidding can begin learning properly, but the signal quality matters more, not less, at this stage.
50+ conversions per month: The algorithm has enough volume to pattern-match. Now the quality of each conversion becomes the critical variable.
Pro tip: If you are below the volume threshold, the fastest fix is not to increase budget. It is to add micro-conversion events (time on site, pricing page visits, phone clicks) as secondary signals so the algorithm has more data points to learn from.
Volume without quality is noise. Quality without volume is silence. You need both.
The Attribution Gap Problem
When to check this: If your Google Ads dashboard shows conversions, but your sales team says leads are low quality or low volume.
This is one of the most common signal failures in B2B accounts. The dashboard looks healthy. Sales says otherwise.
The gap usually lives in one of three places:
Consent Mode is misconfigured. Users decline tracking, conversions go unrecorded, and the algorithm underestimates performance in certain segments.
The conversion window is too short. B2B sales cycles can take weeks. If your window is set to 7 or 14 days, you are missing attributed conversions from longer-cycle buyers.
Offline conversions are not being imported. The sale happens in a CRM or a phone call, but that signal never reaches Google. The algorithm never learns which clicks became revenue.
If your reporting and your sales team tell different stories, the attribution setup is where to look first.
PART 2: The Three Inputs That Improve Bid Signal Quality Without Rebuilding Your Account
Input 1: Qualified Lead Feedback from Sales
This is the most impactful fix and the one most accounts skip entirely.
Smart Bidding can only learn from what it receives. If you never tell Google which form fills became qualified opportunities, it treats all form fills as equally valuable.
The fix is to push lead quality data back into Google Ads as offline conversion imports.
Here is the basic structure:
Assign a unique Google Click ID (GCLID) to every form submission.
Store that GCLID in your CRM alongside the lead record.
When sales qualifies or disqualifies a lead, export the qualified ones with the GCLID attached.
Import that file into Google Ads as an offline conversion event called something like "Sales Qualified Lead."
Set that event as your primary Smart Bidding signal.
"The goal is simple: every time a click leads to a lead your sales team accepts, Google should know about it. Every time it does not, Google should not get credit."
This closes the loop between what the algorithm learns and what your business actually values.
You do not need a perfect CRM setup to start. A spreadsheet with GCLIDs and a qualified/not qualified column is enough to begin.
Input 2: Micro-Conversions as Secondary Signals
When to use it: When your account does not have enough primary conversion volume to train Smart Bidding reliably.
Micro-conversions are meaningful on-site actions that indicate buying intent without being a form submission. They give the algorithm more data points to learn from without diluting your primary signal.
Examples that work in B2B:
Pricing page visit (more than 30 seconds)
Case study download
Phone number click
Return visit within 7 days
Scroll depth of 75% or more on a service page
Set these as secondary conversion actions in Google Ads. They are visible to the algorithm but do not directly influence Smart Bidding targets.
Conversion action setup (Google Ads):
- Primary (optimised): Sales Qualified Lead [offline import]
- Secondary (not optimised): Form Submitted
- Secondary (not optimised): Pricing Page Visit
- Secondary (not optimised): Case Study Download
- Secondary (not optimised): Phone ClickThis structure gives Smart Bidding a richer picture of what a high-intent user looks like, even before they convert.
More signal inputs mean more pattern-matching surface area. The algorithm learns faster and targets better.
Input 3: Audience Signals That Reflect Your Actual Buyers
When to use it: When you are running Performance Max or want to accelerate Smart Bidding's learning on new campaigns.
Google uses audience signals as a starting point before conversion data accumulates. If you do not provide them, it starts from scratch.
Useful audience signals to upload or connect:
Customer Match list (email addresses of existing clients)
CRM list of sales-qualified leads from the past 12 months
Remarketing audiences from high-intent pages (pricing, contact, case studies)
Similar segments based on converted customers
The algorithm uses these lists to identify patterns in who converts, then finds similar users in the broader market.
"Upload your best clients as a Customer Match list. You are telling Google: find more people who look like these. That is a far better starting instruction than 'find whoever clicks.'"
Your existing client list is a targeting asset. Most accounts never use it.
PART 3: How to Structure Conversion Actions So Google Learns From Commercial Outcomes
The Conversion Action Hierarchy
This is the structural fix that makes everything else work.
Most accounts have one or two conversion actions, all set to "primary," all weighted equally. Smart Bidding cannot distinguish between a qualified enquiry and a spam submission.
The correct structure uses a hierarchy:
Level 1 (Primary, optimised): Sales Qualified Lead or Offline Conversion. This is what Smart Bidding targets.
Level 2 (Secondary, not optimised): Form Submitted. Tracked for volume visibility, not for optimisation.
Level 3 (Secondary, not optimised): Micro-conversions. Tracked for signal richness, not for optimisation.
This means the algorithm is rewarded only when a click leads to something your sales team accepts. Everything else is data, not direction.
❌ Bad: Three conversion actions, all set to "primary," including "Form Submitted," "Thank You Page Visit," and "Phone Click." Smart Bidding optimises for all three equally.
✅ Good: One primary conversion action tied to a qualified commercial outcome. All other events set as secondary with "Don't use this conversion for bidding" selected.
The hierarchy is not a technical detail. It is the commercial instruction you give the algorithm.
The Conversion Window and Attribution Model
When to audit this: Before drawing any conclusions from your Smart Bidding performance data.
Two settings that most accounts leave at default and should not:
Conversion window. The default is often 30 days for form submissions. In B2B, a buyer might click an ad, visit your site, leave, come back three weeks later, and then convert. If your window is shorter than your average sales cycle, you are missing attributed conversions.
Check your CRM for average time from first touch to qualified lead.
Set your conversion window to match or exceed that number (up to 90 days for leads).
Attribution model. Last-click attribution gives 100% credit to the final ad interaction before conversion. In B2B, buyers often interact with multiple ads before converting. Data-driven attribution distributes credit across all touchpoints and gives the algorithm a more accurate picture of what is working.
Switch from last-click to data-driven attribution when you have sufficient conversion volume.
If volume is low, linear attribution is a more stable interim choice than last-click.
The algorithm cannot optimise for what it cannot see. Attribution settings determine what is visible.
The Diagnostic Checklist Before Touching Budget or Structure
Run this before making any campaign changes.
Is my primary conversion action tied to a qualified commercial outcome, not just a form fill?
Are offline conversions being imported with GCLIDs from my CRM?
Is my conversion window longer than my average time from click to qualified lead?
Is my attribution model set to data-driven (or at minimum, linear)?
Do I have at least 30 primary conversions per month at the campaign level?
Are micro-conversions set as secondary signals, not primary?
Is Consent Mode configured correctly so conversions are not silently dropped?
Have I uploaded a Customer Match list of existing qualified clients?
Can my sales team classify leads as qualified or not qualified in a format I can export?
Is there a documented process for importing that feedback into Google Ads monthly?
If you answered "no" or "I'm not sure" to more than three of these, your Smart Bidding signal has meaningful leaks.
Fix the signal before increasing the budget. More spend on a broken signal produces more of the wrong result.
Want Growth Junction to Run This Diagnostic for You?
We run a structured Lead-Leak Analysis that covers conversion setup, signal quality, attribution, campaign structure and landing page alignment.
You get a clear findings session showing where the leaks are, what they cost, and what to fix first.
Book a call with me here
You Have Two Choices
Choice 1: Take this playbook and implement it yourself.
Audit your conversion setup. Reconfigure your conversion hierarchy. Build the GCLID capture process. Set up offline conversion imports. Configure Consent Mode. Upload Customer Match lists. Run the diagnostic checklist. Give it three to four months to accumulate enough signal to draw reliable conclusions.
Choice 2: Let Growth Junction run the diagnosis and fix it for you.
We identify the first meaningful signal leak, repair the conversion setup, close the loop between Google Ads and your sales team, and connect it to the rest of your acquisition path. You make decisions from clean data. The algorithm learns from buyers, not browsers.
Work with Growth Junction
The signal repair framework in this playbook is what Growth Junction diagnoses and fixes as part of the Qualified Opportunity System.
We check whether your conversion tracking is recording the right actions, whether your attribution is giving the algorithm accurate instructions, and whether your sales team's feedback is reaching Google in a format it can learn from.
Most accounts we analyse have at least two of the signal leaks described in this playbook. Fixing them does not require rebuilding the account.
It requires connecting the pieces that are already there.
