A Euro per Install

By Georg Malahov

Just before New Year's, Google finally verified my ad account — a pleasant email arrived, I can buy ads now. My goal for the year I put in one short line: validate the idea and get the product to sales. And with the new year I launched the first ad campaign of my life, in search.

First thing, I had to surround myself with tools, which I'd never done before. Set up a CRM in HubSpot — liked it overall, no complaints, still on the free tier. Hooked email into it through Resend, wrote to the first people who registered. Set up conversion tracking, a stats spreadsheet. Lots of fiddling, and all of it unfamiliar — this isn't code; here I take every step slowly and with doubt.

Result of the first full week: spent sixty euros, got roughly sixty installs. So a euro per install — more or less tolerable for a start. CTR is holding somewhere around ten percent: at the lower edge, but livable. The page converts to installs decently too. It all seems to be coming together. But then the problems start, and they worry me.

People install the extension and do nothing with it. They register — and don't open the emails. Or they install and delete it right away. At one point I turned on the extra Google networks on top of search, and I didn't like the result at all: lots of clicks, few installs, junk traffic of some kind. When I turned those networks back off, things got better. Also, at some point my analytics events disappeared — data about session starts and recordings stopped coming in, and for a while I was counting completely blind. My hypothesis: it looks like a noticeable share of the installs comes from India, from not-so-powerful machines, and people simply don't get what the landing page promised them, so they drop off. But that's a hypothesis for now; it needs checking.

In parallel I'm fussing with emails to the people who left a request. My first lead was created in the CRM automatically — nice, the mechanics work — but it was created in a status I can't send emails from, and I had to sort it out by hand. A small thing, but telling: I supposedly automated everything, yet in practice every one of these things needs manual finishing. And so far there are more of those than I'd budgeted for.

I signed up for a tracking session with the course mentor — a personal review of the project: homework, the launch, the funnel, questions about your own product. But the earliest I can get in is in a week, on Monday, January 19. So this week I'm letting the ads run as-is, no changes, one more full week; then I go in for the review, gather my thoughts, and do a new iteration. By the mentor's rough estimate, to get one paying user you need at least a hundred people — at a conversion rate around one to six percent. Which means I just have to grow the base and, in parallel, fix what's breaking inside.

The main thing I need to do this week, technically, is bring back all the analytics events so I can once again see what happens to a user after the install. Without that I'm simply blind and won't understand where I'm losing people.

The plan for the coming week: bring the analytics back and at least understand which step most people drop off at — session start, first recording, or the paywall. Not "improve conversion", it's too early for that — just start seeing the whole funnel again. If by Monday's session I have a clear picture of where it leaks, the week wasn't wasted. If I'm still guessing about Indian machines — then it was.

Let's see what the numbers say. For now I'm letting it run as-is.