A premium local service business was getting hundreds of calls a month from Google Ads — and most of them were the wrong calls. We did not touch the budget. We changed what the algorithm was learning from, and only after months of preparing the data. Here is the before and after, step by step.
Both periods are two months long, with practically the same spend. The only thing that changed is the bid strategy — and the months of data work that made the change possible.
| Metric | Before (2 months) | After (2 months) |
|---|---|---|
| Ad spend | $12,127 | $12,435 |
| Tracked conversion value | $3,128 | $23,464 |
| ROAS (return on ad spend) | 0.26 | 1.89 |
| CTR (click-through rate) | 0.65% | 1.13% |
| Average CPC (cost per click) | $2.15 | $1.50 |
| Phone calls | 672 | 236 |
| Store visits | 318 | 60 |
Tracked conversion value is the sum of deal amounts the client confirmed in the CRM and we sent back to Google Ads, plus store visits counted at $1 each. ROAS = conversion value ÷ ad spend; it is a sales figure, not profit.
The client offers a premium version of a common local service in one of the largest US cities. Google Ads was bringing in a lot of calls, but the owner was frustrated: most callers were looking for the budget option, had dialed the wrong business, or were trying to reach a competitor.
On paper the campaign looked healthy: hundreds of conversions a month, a low cost per lead. In reality only a small share of those calls produced money, and the bulk of the budget bought traffic the business could not — and did not want to — serve.
The campaign ran on Maximize conversions. For the algorithm every conversion weighed the same: a call that led nowhere and a call that closed a large deal were equal.
So the algorithm did exactly what it was asked: it hunted for the cheapest, most frequent conversions. The cheapest ones come from broad, budget-driven searches. The better it "optimized", the further it drifted from the target customer.
The problem was not the targeting settings. It was what the algorithm was learning from.
We set up a CRM → Google Ads loop. Every lead from ads lands in the CRM; the client confirms whether it is a real prospect and enters the deal amount. That value goes back to Google Ads automatically and is attached to the original click. If a deal closes later, the value is updated retroactively.
For the first time Google saw not just "a call happened", but "this call brought $1,950 — and this one $0".
Without this step nothing else works: value-based bidding can only learn from values that actually reach Google Ads — and most accounts never send any.
We went through every search term the ads had shown for and matched it against confirmed deals. Budget-driven and irrelevant demand was cut — but not blindly: some broad terms that looked like junk were actually producing target sales. Negative keywords were built from data, not intuition.
The least obvious step: we waited. Switching to value-based bidding right away would have been a mistake — the campaign history did not yet have enough deals with real amounts, and the algorithm would have learned from an incomplete picture.
We kept the old strategy for months, until the account had enough confirmed conversions with correct values and the average conversion value in Google Ads matched the business's real average ticket. That was the readiness signal.
Only then did the campaign move to optimizing for value. Now the algorithm is not looking for people who will call — it is looking for people who will bring in a deal. The budget was deliberately left unchanged so the result would be comparable.
Tracked value grew from $3,128 to $23,464. ROAS went from 0.26 to 1.89: every dollar spent now brings back almost two in tracked sales.
CTR rose from 0.65% to 1.13% and the average click fell from $2.15 to $1.50. The ads are being shown to a more interested audience, and Google charges less for it.
Calls dropped from 672 to 236. The calls that never produced money are gone — which is exactly what the owner wanted.
Store visits dropped from 318 to 60. This matters: these are people who saw the ad and then walked into the location, which Google detects from device location. For a walk-in business that is real demand, and it also feeds the algorithm indirectly. But the campaign was measured on calls and leads with confirmed amounts, and store visits were counted at a nominal $1 — so this drop sits outside the value figures above and is worth watching separately.
Value-based bidding works, but not by flipping the setting. Switch a campaign to conversion value without preparation and the algorithm learns from empty or distorted data — and the result is worse than before.
The sequence that works: real deal amounts flowing from the CRM → clean incoming traffic → months of accumulated data → only then the change of strategy. Stable, long-term results in Google Ads — the kind many accounts never reach — take that much work. Probably like anything else.
Whether it fits your business depends on the niche, the average ticket, and how many deals the account can accumulate. A business with a handful of deals a month will need longer; a business where every lead is worth about the same gains little from it. What we can promise is what we work on, not the outcome.
And fewer calls is not always bad news. Sometimes it means the ads have finally stopped paying for people who don't buy.
Maximize conversions treats every conversion as equal and looks for the cheapest, most frequent ones — for a local business that usually means broad, budget-driven searches. Maximize conversion value optimizes for the total value of conversions, so a $1,950 deal weighs more than a $0 call. The second strategy only works if Google actually receives real deal values for each conversion.
Calls went from 672 to 236 over comparable two-month periods because the algorithm stopped paying for the cheap, off-profile calls that never turned into deals. That was the owner's main complaint. Tracked value from the campaign went from $3,128 to $23,464 on the same budget, so fewer calls meant more business, not less.
Every call and form from ads lands in the CRM. The client marks whether the lead is a real prospect and enters the deal amount. That value is sent back to Google Ads and attached to the original click, using Google's offline conversion import. If a deal closes later, the value is updated retroactively. Google then sees not just "a call happened" but "this call was worth $1,950 and this one $0".
You can, but without preparation it usually makes things worse. If the account has no real values — or values on only a handful of conversions — the algorithm learns from an empty or distorted picture. In this case we kept the old strategy for months while confirmed deals with correct amounts accumulated, and switched only when the average conversion value in Google Ads matched the business's real average ticket.
ROAS (return on ad spend) is tracked conversion value divided by ad cost. 1.89 means that for every dollar spent on ads, the campaign brought back $1.89 in tracked sales value. Before the change the same campaign was at 0.26. ROAS is revenue-side, not profit — the margin on those sales is the business's own number.
It depends on the niche, the average ticket, and how many deals the account can accumulate. Value-based bidding needs enough confirmed conversions with real amounts to learn from; a business with a handful of deals a month will take longer, and a business where every lead is worth about the same gains little from it. What we can say is what we work on: CRM verification, value import, clean search terms, and the switch only when the data is ready.
Abbreviations: ROAS — return on ad spend; CTR — click-through rate; CPC — cost per click; CRM — customer relationship management system; PMax — Performance Max campaign type.
Start with a $35 audit: we review your campaigns and conversion tracking, and tell you whether the account is ready for value-based bidding — or what has to be built first.
How we manage Google Ads →