We found every problem below before changing a thing.
Read-only access, no commitment. You keep the findings either way.
Harkla sells sensory swings, weighted blankets, compression products and therapist-led courses for children with sensory and developmental needs. We took over their Google Ads in July. Against the same two months a year earlier, the account made two thirds more profit and brought in half as many orders again.
Google Ads results for July and August 2025, under the previous setup, against our first two months on the account.
Read-only access, no commitment. You keep the findings either way.
Harkla's goal is profit after product costs, shipping and fees, not revenue. Before changing a bid, we audited the account, the Shopping search terms and every product's return against that goal. Four problems stood out.
The summer before, 14 campaigns had spent money and 8 of them lost it: $2,205, or 16% of the budget, on campaigns that returned less gross profit than they cost.
Competitors bid on Harkla's name, yet there had been no brand search last summer. When we arrived, the brand campaign was capped at $30 a day and losing 58% of its impressions to budget, while returning more than 8x its cost.
Shopping was set to a 3.0x return. No price tier in the catalogue reached it, the best topped out at 2.5x, so the campaign was losing 56% of its impressions to ad rank.
All physical products sat in one catch-all group. Sensory swing variants alone ranged from 0.86x to 3.27x, and one target cannot price both correctly.
We set the goal to profit, gave each campaign one job, and protected the cheapest sales in the account before chasing new ones.
We worked out Harkla's real break-even from product margin, shipping and fees, then judged every campaign, product and week on gross profit after ad spend. A campaign with a good return on ad spend that still loses money after costs is a loss, and we treated it as one.
Moved brand search to a target impression share strategy with a controlled click ceiling, so Harkla shows up whenever someone searches its name, without overpaying for clicks it would win anyway. Over the two months, brand search cost $1,536 and returned $6,932 in profit.
Seven campaigns instead of fourteen: brand, physical products in Shopping, courses in Performance Max, and a dedicated campaign for the new weighted vest launch. Each one has its own target and budget, so brand, courses and the new launch no longer compete with the main range for the same money.
Ran an n-gram analysis on 16,637 Shopping search terms, added negatives only where the catalogue confirmed the search was off-product, paused the variants that kept spending without selling, and managed the target against what each product line actually returns.
Harkla's therapist-led courses sell to a different buyer than the physical range, at a different margin. They ran in their own Performance Max campaign with their own creative and return target, and in July courses made more profit than a year earlier on less spend.
A weekly report on conversion-time numbers, cross-checked against Triple Whale, so every change was judged on what it did to profit and adjusted within days rather than months.
July, our first month on the account, made 82% more profit than July 2025. August, with back-to-school competition pushing click prices up, still made 48% more.
Clicks cost 17% more. Profit still grew faster than spend.
Average cost per click rose from $1.35 to $1.58 across the year, so the gain did not come from cheaper traffic. It came from where the money went. Brand search turned $1,536 into $6,932 of profit, and the rest of the account found 50 more new customers than the year before at close to the same profit:
| Jul and Aug, 2025 vs 2026 | Ad spend | Orders | New customers | Profit after ad spend |
|---|---|---|---|---|
| Brand search | $0 → $1,536 | 0 → 106 | 0 → 85 | $0 → $6,932 |
| Shopping, Performance Max and seasonal | $13,789 → $16,907 | 335 → 393 | 291 → 341 | $9,692 → $9,149 |
| Account | $13,789 → $18,443 | 335 → 499 | 291 → 427 | $9,692 → $16,081 |
With 23% of Harkla's customers coming back to buy again, those extra first orders keep paying after the reporting window closes.
Read access and forty-five minutes. You keep the findings either way.
Book a free audit