Teaching Google Which Enquiries Became Clients
Ask a law firm owner what they want from Google Ads and the answer is always some version of “more clients.” Then look at what the account is actually optimising toward, and it is almost never clients. It is form submissions, or phone calls longer than sixty seconds, or in the worst cases, clicks on a phone number that nobody ever answered.
Automated bidding is extremely good at getting more of whatever you tell it to get. That is precisely the problem.
What the machine currently knows
Google can see everything up to the moment someone contacts your firm, and nothing after it. It knows a person searched, saw an ad, clicked, and submitted a form. It does not know that the person lived three states away, or wanted a practice area you do not handle, or was a paralegal shopping for a job.
It also does not know that the enquiry three positions down the same report became a $40,000 contested probate.
So it optimises for volume of the only event it can observe. Left alone for six months, an account like this reliably drifts toward cheap, plentiful, low-quality enquiries, because those are the ones the system is being rewarded for finding.
Closing the loop
Offline conversion import is the mechanism that sends the outcome back. Google’s own developer documentation covers the upload of offline click conversions keyed on the click identifier, with the conversion time, value and currency attached (Google Ads API).
In practice the loop has four moving parts, and most firms have only the first two:
One, the click is stamped. Auto-tagging appends an identifier to every ad click. Your landing page captures it into a hidden field and your CRM stores it against the lead record. If this step is missing, nothing downstream is possible, and it is missing more often than not. Auto-tagging is usually switched on. The capture field usually is not.
Two, the outcome is recorded. Not “contacted.” The actual milestones: qualified, consultation booked, matter opened, fee value. If your CRM records leads but not outcomes, this project becomes an intake-process project before it becomes a marketing project, and that is a healthy discovery to make early.
Three, the outcome is uploaded. Daily is ideal, weekly is acceptable, monthly is close to pointless. Bidding responds to recent data, and a feedback loop that lags by a month cannot steer a system making decisions this week.
Four, bidding retrains. Over the following weeks, budget shifts toward the clicks that look like the ones that became matters.
The half nobody runs: taking conversions back
Uploading wins is the obvious half. The other half is corrections, and it is where a lot of the improvement actually comes from.
- Restatement updates a conversion’s value after the fact, when a matter turns out to be worth considerably more or less than the estimate at intake (Google Ads API).
- Retraction removes a conversion entirely, when the lead turned out to be a duplicate, a conflict, or something that was never a real enquiry.
Retraction is the part that changes behaviour fastest. If you leave a bad lead counted as a conversion, you are actively teaching the system to find more people like that. Withdrawing it teaches the opposite. Including an order identifier on the original upload is what makes later adjustment straightforward, which is why we send one on every conversion even when nobody has asked for it yet.
What good reporting looks like after this
Once outcomes flow back, the reporting conversation changes completely. You stop comparing cost per lead between months and start looking at the whole ladder: impression share, leads, qualified leads, consultations booked, signed matters, fee value, and cost per signed matter.
It is a more uncomfortable report to receive, because it makes the weak parts of the funnel visible. A firm with an excellent cost per lead and a terrible consultation-to-signature rate can no longer hide behind the first number. In our experience owners want that discomfort. They have usually suspected it for a while.
The order of operations matters
We build this in sequence, and we resist client pressure to skip ahead:
1. Fix measurement first. Click capture, first-party serving, enhanced conversions, conversion action hygiene.
2. Get outcomes into the CRM consistently, with definitions the whole team uses the same way.
3. Upload counts, daily.
4. Add values once they can be trusted.
5. Move bidding onto value.
Jumping to step five on top of inconsistent CRM data produces worse results than staying at step one, because you have now taught the system something confidently wrong.
Law Ops Forge builds this loop because we already live on both ends of it: the intake operation that captures the outcome, and the systems team that moves the data. Talk to us about closing yours.
Sources
Every claim above comes from Google’s own documentation. These are the pages it came from, so you can check any of it yourself, or send it to whoever else needs convincing.
Talk to Law Ops Forge
Want a second set of eyes on your firm's intake?
Get a free intake and sales audit and see exactly where your firm is losing cases, and what fixing it would look like.
Get your free audit