For years, working an LSA account meant one gritty habit: disputing bad leads by hand. Wrong service, wrong city, a robocall, a wrong number — you flagged it, wrote a note, and asked Google to credit it back. Then, around July–August 2024, that entire workflow disappeared. Google's 2024 credit model replaced manual disputes with a machine-learning system, and the change reshaped how advertisers should think about lead quality. This piece explains what actually changed and why it matters more than it first appeared.
What the old model was and why Google killed it
The manual dispute system gave advertisers a lever they could pull case by case. It also gave Google a mountain of subjective, labor-intensive claims to adjudicate, and it created an adversarial rhythm where advertisers spent energy arguing over individual leads rather than improving their pipeline. Retiring it in 2024 was consistent with Google's broader direction: push repetitive judgment into automated systems and reduce the manual surface area advertisers touch.
How Google's 2024 credit model works now
The replacement is a machine-learning credit system. Leads are assessed automatically — often within about 72 hours — and qualifying non-bookable leads are credited back within roughly 30 days. Advertisers also see a "Rate this lead" survey, which feeds signal back into the model. Crucially, some categories of bad lead are simply not creditable: job-type mismatches and geographic mismatches fall outside the system, and healthcare and tax are excluded verticals entirely. Third-party estimates put recoverable spend at around 6–7% of total LSA spend.
| Aspect | Before (pre-2024) | After (post-2024) |
|---|---|---|
| Who decides | Human review of your dispute | Machine-learning assessment |
| How you act | File each dispute manually | Rate this lead survey feeds the model |
| Timing | Variable | Assessed ~72h, credited within ~30 days |
| Not creditable | Case by case | Job-type and geo mismatches, excluded verticals |
Why this changed the game, not just the interface
It is tempting to read this as a UI change — one workflow swapped for another. It is bigger than that. When credit was manual, an advertiser with time and discipline could recover a meaningful slice of wasted spend through sheer persistence. Under the machine-learning model, you cannot out-argue the system; you can only feed it accurate signals and, more importantly, improve the quality of leads coming in so there is less to recover in the first place. The leverage moved upstream, from disputing bad leads to preventing them.
That shift is decisive because of the underlying math. A large share of raw LSA leads — third-party estimates suggest around 45% — are unbookable, and average cost per lead is often cited around $53, ranging roughly $12–$180 by trade and metro. If only 6–7% of spend is recoverable as credit, then recovery alone cannot save a poorly-managed account. The real savings live in targeting, geography, scheduling, and speed-to-lead — the levers that stop you from paying for bad leads rather than clawing money back afterward.
What advertisers should actually do
Three habits follow directly from the new model.
- Catch every eligible credit. Recovery is bounded, but 6–7% of a real budget is not trivial. Every creditable non-qualifying lead you miss is money left on the table. Consistent, prompt "Rate this lead" responses matter because they feed the system that decides.
- Move the fight upstream. Tighten job-type and geographic settings so mismatches — which are not creditable — stop entering your funnel. This is where the largest savings hide.
- Track cost per booked job. Because credit is automated and partial, the only honest measure of channel health is what a booked job actually costs you after unbookable leads and recovered credits are accounted for.
The bigger signal in the 2024 change
Step back and the credit model reads as a preview of where local advertising is going. Google is steadily replacing human, case-by-case advertiser labor with machine-learning systems that reward clean signals and good upstream operations. The advertisers who thrive under this regime are not the ones who fight hardest over individual leads; they are the ones whose accounts are configured so precisely that there is little to fight about, and whose response speed converts the good leads before a competitor does. Google's 2024 credit model did not just change how you get money back — it changed what winning an LSA account requires.
Frequently asked questions
How does Google's LSA credit model work after 2024?
Manual lead disputes ended around July to August 2024. Google now uses a machine-learning system that assesses leads automatically, typically within about 72 hours, and credits qualifying non-bookable leads within roughly 30 days. Advertisers also see a Rate this lead survey. Job-type and geographic mismatches are not creditable.
Can you still dispute a bad LSA lead?
Not through the old manual dispute flow, which was retired in 2024. Credit is now handled automatically by Google's machine-learning system, supplemented by the Rate this lead survey. Because you cannot argue each case by hand, the leverage shifts to feeding accurate signals and improving upstream lead quality.
How much LSA spend is typically recoverable as credit?
Third-party estimates put recoverable spend around 6 to 7 percent. Healthcare and tax are excluded verticals, and mismatches in job type or geography are not creditable, so recovery is meaningful but bounded and depends on catching every eligible lead.