Most case study pages are a highlight reel. This is an index — ten businesses on one page each, so that when you need the one that looks like yours, you’re not hunting for it.
Every one follows the same five steps: the situation, what the calculation said, what was done, what happened, and the point.
Where a business isn’t in the United States it’s marked. Prices don’t transfer across markets even when the arithmetic does, and a page about honest numbers shouldn’t blur where its numbers came from.
Two of these ten never ran a campaign. Those two saved the most money.
1. The company that fixed its sales process and no longer needed us
Another market.
The situation. A manufacturer of prefabricated storage units — the lockers people install in parking garages. Revenue was thin. They came to us for advertising, which is what businesses come for when revenue is thin.
What the calculation said. The numbers didn’t work, and the reason was specific: sales conversion was badly under where it needed to be. The calculation said something more useful too — the skills to fix it already existed inside the company. Nobody needed to be hired and nothing needed to be bought.
What was done. Nothing, by us. They would spend a month on their own sales process, run no advertising at all, and we’d talk again after.
What happened. A month later I asked whether the conversion had improved and whether we should launch. The conversion had improved. Revenue had tripled. And they no longer wanted advertising, because the demand they already had was now converting well enough to grow the business without it.
They’re still growing organically. We were never paid.
The point. When a business converts at half of what it should, advertising isn’t the missing piece. It’s a way of paying to avoid a conversation you could have had for free. The order of repair is where that logic lives.
2. The report that was accurate and still hid a loss
Hawaii. Residential cleaning.
The situation. A cleaning company with an agency already running Google Ads. Every month a report arrived saying how many leads had come in. The number was going up. Nobody was lying.
What the calculation said.
| Average job | $171 |
| Margin | 30% |
| Most they could pay per customer | $51.30 |
| What they were actually paying | ~$67 |
| Result per one-time job | −$15.75 |
The report had never put the cost of a lead next to the value of a customer, so nobody had to notice that one was larger than the other.
Worse: most of that traffic was buying move-out cleanings — the deep clean people order when they’re leaving a property. A move-out customer is, by definition, moving out. The “we make it back on repeat business” defense was structurally impossible for that channel.
What was done. Split the account by customer type — recurring, one-time, commercial. Changed the offer so one-time cleans could convert into recurring contracts, which raises the maximum rather than improving performance against it. Then one tight search campaign on high-intent local terms, a disciplined exclusion list, conversion tracking on calls and forms. Performance Max in month two, once search had conversion history to learn from. Local Services Ads in month three.
What happened. Ten weeks: $4,272 spent, 202 leads, blended cost per lead from roughly $31 down to $17 against a target of $21.15. Best weeks on Performance Max between $5.93 and $8.36.
The point. That’s a small budget by any standard, and it worked — which disposes of “Google Ads doesn’t work at our size.” Nothing that made it work was budget.
3. The account we refused to scale
Los Angeles. Appliance repair.
The situation. A new account, built from scratch, in a competitive metro.
What the calculation said. Average ticket $265, margin 50%, close rate 67%, website conversion 3%. Which gives $132.50 per customer, $88.78 per lead, $2.66 per click. Repeat factor about 1.5 jobs per customer.
What was done. Three channels: residential search, a separate campaign for commercial work, and Local Services Ads. Then a weekly cycle — review cost per lead by channel, prune the search terms bringing the wrong people, move budget toward whatever was producing.
What happened.
| Week | Spend | Leads | Cost per lead |
|---|---|---|---|
| 1 | $287 | 3 | $95.60 |
| 2 | $352 | 6 | $58.69 |
| 3 | $297 | 13 | $22.83 |
| 4 | $247 | 11 | $22.46 |
Week one came in above the $88.78 maximum — the calculation was refuted on day three, exactly as a first month should do. By week three: four times the volume for less money than week one. Over seven weeks, 55 leads on $2,477, averaging $45, running at around $22 — about a quarter of what the business could afford.
Then we stopped growing it.
Of the qualified leads reaching the dispatcher, roughly one in ten became a booked diagnostic visit. Nine out of ten people who called — people the business had paid for, with broken appliances, who wanted somebody to come — hung up without an appointment.
The point. Doubling the budget there doesn’t double revenue. It doubles the number of people who call and don’t book. Fixing the phone was worth more than any campaign work available, and recommending it capped our own invoice. It was still the only answer that made the client money.
4. Volume up, cost per lead down at the same time
HVAC.
The situation. An account running on Local Services Ads and not much else, with an owner who wanted real volume across residential repair, installation and commercial work — and the correct fear that paying several times as much would mean paying several times as much per lead.
What the calculation said. The maximum had room. The constraint wasn’t economics — it was that nobody had ever pushed the account. And critically, this is a licensed trade with a stable crew, so the capacity limit could be planned against.
What was done. Scaled Local Services Ads as the volume base, then separated residential and commercial search — one campaign at a time rather than all at once — and switched off the campaign types the demand had left.
What happened. Over five months: 745 leads on $33,968, finishing at 219 leads in July at $37 each. The first month of the scale-up cost $67 a lead and it came down after that — $49, $45, $45, $37 — without going back up.
And the first month of scaling went backwards. While budget was ramping, cost per lead sat at $67 — the most expensive month of the run. Two ordinary causes: new campaigns in their learning period with no conversion history, and search terms not yet pruned. It took the following month to pull it down — $49, then $45, $45 and $37.
The point. Anyone can buy a handful of leads cheaply at low volume. The job is keeping the math intact while spend grows five or ten times. The dip during ramp-up is normal and predictable — decide in advance how bad you’ll let it get, and for how long.
5. The service line that failed inside a healthy company
HVAC, with appliance repair as its second line.
What the calculation said.
| Service line | Average job | Margin | Max per click | Market price | |
|---|---|---|---|---|---|
| HVAC service & repair | ~$750 | ~50% | ~$4.70 | ~$4.50 | Green, barely |
| HVAC installation | ~$11,000 | ~10% | ~$11.50 | ~$7 | Green |
| Appliance repair | ~$300 | ~36% | ~$1.90 | ~$5.30 | Red |
Three things in that table. Margin runs from about 10% to about 50% inside one company — anyone quoting “our margin” for a business like this is quoting nothing. Margin and ticket move in opposite directions: installation keeps a tenth of the money and is the most profitable line, because a tenth of eleven thousand beats half of seven hundred and fifty. And one line is red: appliance repair cannot be bought at market price here.
The point. That’s what a “no” looks like in practice — not a doomed business, but a healthy one with a single service line it shouldn’t advertise, which would quietly drain the other two if it shared their budget.
6. Three maximums, and a budget that follows the crews
Hawaii. Remodeling.
What the calculation said. Not one maximum. Three.
| Service line | Average job | Margin | Max per customer | Max per lead | Max per click |
|---|---|---|---|---|---|
| Kitchens | $26,000 | 32% | $8,320 | $728 | $21.84 |
| Bathrooms | $22,000 | 35% | $7,700 | $674 | $20.21 |
| Decks & outdoor | $15,000 | 41% | $6,150 | $538 | $16.14 |
About 35% of leads become booked estimates and about 25% of estimates become jobs — roughly one lead in eleven turns into work, which is why a lead here can be worth several hundred dollars. Notice that the biggest job has the worst margin.
What was done. Separate campaigns per service line. Offline conversion tracking — forms and dynamic call tracking with click identifiers, matched to the CRM by phone number, with qualified and estimate stages pushed back so the platform optimizes toward revenue rather than ringing phones. Storm-triggered deck campaigns, built and paused in advance.
And a weekly rule worth copying: budget is redistributed by qualified leads, booked estimates, jobs won, and which crews are free next month — not by which channel produced the cheapest lead.
What happened. Search ran at $82.58 per lead in its best week with click-through above 16%; Facebook between $29 and $56; Yelp between $20 and $25. Once the model was proven on one island the same structure was deployed on another — a second market on a known playbook rather than a new experiment.
The point. If the cheapest leads are kitchens and the kitchen crew is booked into November, the money should go to decks. Buying cheap leads for work you can’t perform is the most efficient way to lose money there is.
7. The trade where advertising doesn’t buy leads
Roofing.
The situation. The largest money comes from insurance claims. A storm goes through, the companies good at this know precisely which streets it hit, and they walk the neighborhood: we’ll inspect it free, we’ll file the claim, we’ll do the repair. The marketing is a person standing on a doorstep.
What the calculation said. Average job around $15,000, margin in the mid-teens — still over $2,000 per customer. Close rate about a quarter, website conversion about 2.5%. Most they could pay per click: around $14. The market was charging around $5. One of the healthiest gaps we’ve measured.
What was done — and what it’s for. Not lead generation. Reach campaigns into the affected ZIP codes, so the company is familiar by the time someone from the team is on the doorstep.
What happened — and how you’d know. Cost per lead is undefined for a campaign like this; the click never becomes a lead. You measure the close rate of the door-knocking crews with air cover against without. Two adjacent ZIPs, same storm, same crew, one with a campaign and one without. Almost nobody runs it.
The point, and it isn’t about roofs. The depth of optimization available to you is set by your conversion volume, not your budget. Optimizing toward sales in a business that makes eight a month isn’t ambitious — it’s training an algorithm on noise.
8. Where paid advertising was impossible, and we said so
Another market. Fulfillment services for online sellers.
The situation. A startup wanting to run paid search. No established unit economics, processes not yet settled, small budget.
What the calculation said. Before quoting anything we counted the demand. Across every commercial term describing what they did, in their region: 335 searches a month. Not clicks — 335 times a month that anybody typed anything relevant at all. At competitive bids and achievable conversion rates, the projected cost of acquiring one customer came out to several times what a customer was worth.
What was done. We didn’t run it. The budget went into one round of search optimization on a single landing page — deliberately a one-off rather than an ongoing engagement, because the risk that the startup wouldn’t survive was real and we said so.
What happened. Sixteen target queries, twelve into the top ten. Visibility from zero to 42. Organic share of traffic from 0% to 23%. Cost per click 47 against the 400 paid would have cost, and cost per conversion 631 against a projected 9,000+ — roughly a tenth. (Local currency, another market; the ratio is the transferable part.)
The point. Demand capacity gets checked first, before economics, because if nobody is searching then the most profitable business in the county still can’t buy customers who don’t exist. It takes an afternoon and costs nothing.
9. The month the sales team was replaced
Another market.
The situation. A business wanting to launch with nothing to calculate on — barely a sales function, no history worth the name. Every number was going to be a guess.
What the calculation said. We built the model on assumptions, wrote each one down, and launched a deliberately small campaign to test them. One of them was the close rate. We assumed 65%.
What happened. The real figure came back under 30%. Not a miss — less than half. And notice which side of the table it came from: the market behaved as expected. It was the business’s own number that fell apart. Calls unanswered. Slow callbacks. No follow-up. No second attempt on anyone who didn’t buy immediately. The people weren’t lazy — they simply weren’t selling. They were taking orders.
What was done. The advertising was paused. The sales team was replaced. Not retrained. Replaced.
What happened next. The close rate came back at just over 50% — below the 65% assumed, short of the plan. And it was enough. The economics closed at 50%.
The point. Your assumption is a hypothesis. The threshold is the decision. Missing your assumption is not failing; falling below the threshold is.
10. The one where the product moved the economics
Remodeling. Bathrooms.
The situation. In the US a tiled bathroom is a premium product — not because of the tile, but because installing it is slow and skilled labor is expensive. Which meant every customer with a real budget constraint was being turned away.
What the calculation said. No amount of campaign work reaches a customer who can’t afford the product. The constraint wasn’t the ad account. It was the price point, and the price point was set by labor hours.
What was done. Finished bathrooms built on PVC panels, sourced in bulk, in designs that actually look good. Cheaper for the customer. Much faster to install. No tile setter required.
What happened. The average job fell. The margin percentage held. The maximum per click barely moved — because the close rate rose to meet it. So what did they gain? A market that used to be turned away (not cheaper customers — customers who previously bought nothing from anyone), and production capacity, because more projects fit into the same crew-month.
The point. A change like that hands you a strategic choice: pass the gain to the customer as a lower price and take volume, or hold the price and take margin. Both are legitimate. Drifting into one without deciding is not.
What the ten have in common
Read them together and the same shape appears.
In eight of the ten, the thing that decided the outcome was not in the ad account. It was a close rate, a dispatcher, a price point, a product, a service mix, or the absence of demand.
In two of them the correct action was to spend nothing at all, and both of those businesses grew.
In one, the numbers came back wrong in the first month and the correct response was to stop advertising and rebuild a department.
And in the one case where the advertising itself carried the result — $96 a lead to $22 in three weeks — the work that produced it wasn’t a clever setup. It was somebody looking, cutting and reallocating every single week.
Run your own seven numbers through the maximums calculator and find out which of these ten your business is. The full write-ups are in the book; the published client case studies are here.
Frequently asked questions
- Does Google Ads work for small home service businesses?
- It can, and budget is rarely what decides it. A Hawaii cleaning company spent $4,272 over ten weeks and got 202 leads, with blended cost per lead falling from about $31 to about $17. That's a small budget by any standard and it worked. But nothing that made it work was budget: the maximum was calculated before anything was switched on, the account was split by customer type instead of averaged, the offer changed so the maximum itself moved, search terms were reviewed weekly, and channels were added in sequence only after the previous one proved out.
- Would an agency ever tell me not to advertise?
- Two of these ten never ran a campaign, and both businesses grew. A storage-unit manufacturer came for advertising with revenue thin and a sales conversion far below where it needed to be — the fix already existed inside the company, so they spent a month on their own sales process instead. Revenue tripled and they no longer wanted advertising. A fulfillment startup had 335 relevant searches a month in its entire region; there was no version of that campaign that worked, so the budget went into one round of search optimization instead and produced customers at roughly a tenth of the projected paid cost.
- Why would an agency refuse to scale a profitable account?
- Because the bottleneck was downstream. A Los Angeles appliance repair account went from $95.60 per lead in week one to $22.46 by week four — running at about a quarter of what the business could afford, with obvious room to grow. But of the qualified leads reaching the dispatcher, roughly one in ten became a booked visit. Doubling the budget there doesn't double revenue; it doubles the number of people who call and don't book. Fixing the phone was worth more than any campaign work available, and recommending it capped our own invoice.
- What do these cases have in common?
- In eight of the ten, the thing that decided the outcome was not in the ad account: a close rate, a dispatcher, a price point, a product, a service mix, or the absence of demand. In two, the correct action was to spend nothing at all. In one, the numbers came back wrong in the first month and the right response was to stop advertising and rebuild a department. And in the single case where the advertising itself carried the result — $96 a lead down to $22 in three weeks — the work behind it wasn't a clever setup. It was somebody looking, cutting and reallocating every week.
- Are these numbers from US businesses?
- Most of them. Where a business is outside the United States it's marked as such, because prices don't transfer across markets even when the arithmetic does. The US cases are Hawaii (cleaning, remodeling), Los Angeles (appliance repair), Charlotte and roofing. The storage-unit manufacturer, the fulfillment startup and the sales-team rebuild are from other markets, and for those the ratios are the point rather than the absolute figures.