Yoofix, 2018: we cut acquisition cost 18%. The business died anyway.
24 September 2026 · a go-to-market post-mortem
I made these GTM mistakes with my own money, so you don’t have to make them with yours.
In 2017 I founded Yoofix, an app for booking an AC technician in Jakarta and Yogyakarta, and ran its go-to-market end to end: acquisition, pricing, the city rollout, the technician side. We had almost no money. Our entire lifetime marketing budget was 41% of what the Yogyakarta operation alone cost us in a single quarter.
Acquisition worked. Monthly orders grew by more than 560% between February and May 2018, and in June our cost per acquired customer fell 18%. By the metric I reported every month, 2018 was going well.
The company still died. Only about half of all orders (51%) ever turned into a completed job. I recently went back through the original spreadsheets, removed the 6% of rows that were duplicates or test orders, and read them the way I should have read them then. Six signals were telling me something I wasn’t listening to.
1. Demand followed the promo, not the product
Orders peaked in June 2018. We stopped the promo. Within three months orders were down 80%. A cheaper acquisition cost during a promo month measured how efficiently we were giving money away, not whether anyone wanted the service at full price. I never ran a holdout, a city or a week without the discount, so I never learned what real demand looked like.
2. One price, two different markets
We charged one flat price per visit in both cities. The failure log shows what that did:
| Jan–Jun 2018 | Share of orders | Completed | Lost to “price doesn’t fit” |
|---|---|---|---|
| Jakarta | 56% | 65% | 1% |
| Yogyakarta | 41% | 55% | 9% |
The same price was nine times more likely to kill an order in Yogyakarta. It was probably also too cheap in Jakarta, where almost nobody objected. Pricing by area was the obvious lever and we pulled it too late.
3. Marketing sold jobs the network couldn’t do
In the core of Jakarta, orders mostly got done: 81% in West Jakarta, 78% in Central. At the edges they didn’t: 38% in Tangerang city, 36% in Tangerang regency, 22% in Bogor regency. “No technician available” caused as many failures as price did, 11% each. Only 38% of registered technicians were active, and by 2019 the average active technician got a job roughly once every six months. Nobody stays on a platform like that.
Acquisition was never the constraint. Supply density per area was, and every rupiah spent on demand outside a well-served area bought a cancellation.
4. Customers chose WhatsApp over the app
Two thirds of orders (66%) came through the app. The rest came through WhatsApp (20%), the phone (11%) and the website (3%). Almost a third of the business arrived through channels I treated as support, not sales. For a home-service purchase made once or twice a year, a chat with a person was the product.
5. Repeat was thin, but the window was short
6% of customers with a completed order ordered again within the roughly four months the log covers. AC servicing is a six-month cycle, so this doesn’t prove people left. It does mean I had no evidence they came back, and a marketplace without repeat has to buy every order twice.
6. Fixed cost ate the city
77% of what the Yogyakarta operation spent in its first quarter was salaries. That quarter alone cost roughly as much as the gross value of every order Yoofix handled in all of 2018. Next to the payroll, the acquisition numbers were a rounding error.
Was it a GTM success or a failure?
Both, and the difference is the lesson. Acquisition succeeded: cheap, fast, on a budget most startups would call nothing. The go-to-market strategy failed, because GTM is not getting people in the door, it is getting to revenue that repeats and makes money. And the company died mostly of operations and unit economics, which better GTM would have slowed but maybe not saved.
What I’d do now
The tactics of 2018 are gone. The questions aren’t.
- Measure contribution per completed order, not cost per acquired customer. A cheap customer who cancels because no technician exists is a loss.
- Launch area by area, supply first. Open a district for demand only when completion there clears a bar.
- Price by area, using local purchasing power, and test it before scaling spend.
- Run a holdout on every promo, so the demand you’re paying for is separated from the demand that exists.
- Treat WhatsApp as a sales channel with its own tracking, because that is where almost a third of the orders were already coming from.
None of this needs more budget than we had. It needs reading the numbers you already have before the next spend. That is the work I do now: I made these mistakes with my own money, so you don’t have to make them with yours.
Sources: Yoofix order logs (Oct 2017–Aug 2019), the January–June 2018 order recap, the June 2018 KPI report and the Q1 2018 Yogyakarta summary. Figures are shown as percentages; customer and company data stay private. · Another post-mortem: the emails that never arrived · bilhokista.web.app