150 mistakes as a CMO

In basketball there's a kind of coach who spends the timeout drawing plays on the clipboard while his team is down by 20. Well, that was me.

For two weeks straight I didn't close a single task. A few days later, in 48 hours, I created four new AI agents.

I rebuilt a campaign log on Notion from scratch, decision log and procedures included, and threw it away a month later, because it was copying by hand what Meta already records on its own. I rebuilt the same report from scratch four times. In September half the time I spent with Claude went into maintaining systems, and only a fifth went into the lever I'm measured on.

Building a system gives you the feeling of moving forward without the risk of a number that doesn't move. That's exactly why it's so comfortable.

Since I started working with AI, writing a task costs nothing. The result: from August I was creating about 80 a month. Before, I created 8.

Guess how many I closed? Just over half.

On my weekly plans I hit 8 commitments out of 34. The “High” priority had become the default for whoever wrote the task: it closed one task in three. I wrote the rule “every task has a deadline” in June, then again in my role agreement, and in ten weeks I didn't respect it once.

We launched a new city one weekend in October. The press release went out four days later.

As in: after.

It wasn't bad luck. It was a commitment I'd made at the start of the month and kept pushing back. Same story with reviews: almost every task on that topic was weeks overdue. Ads, landing pages, CRM, community: there I know where to put my hands. PR and reviews, no, and in fact they were always at the bottom of the list.

This is the one that stings the most.

Creatives marked to be switched off, left running for three weeks. Not because nobody knew, but because nobody had said who pressed the button. As someone on the team wrote: “we need to say out loud who presses the button”.

A project that changed owner five times in just over three months, and in one of those handovers I was the one passing the ball back.

A city's ads switched off for a test, to be switched back on after a week. They stayed off because we were waiting for a confirmation that nobody had the job of giving. In those two weeks the number of new people coming to play halved.

And then the messages: after a launch, the retro showed that information wasn't flowing between areas. I was sending six messages with urgent to-dos to someone who'd had their plan scheduled for days.

In September I changed an important target. Rightly, in my view. Except I wrote it in my own files, and the confirmation in the document everyone read arrived almost six weeks later: in the meantime two different numbers were going around for the same thing.

A metric definition agreed in May didn't make it into the right document until September: for three months we measured it the old way. And I changed a choice about our retargeting campaigns four times in seven weeks, so much that the automated report kept asking me to do something I'd already cancelled.

The app store conversion rate we kept quoting was off by about three times.

We spent two months chasing an SEO problem that didn't exist: the pages were there, we were just looking at the wrong dashboard.

For ten days we tracked an event against 64 spots, when the real spots were 48. Alarms, recovery plans, worried messages... over a gap that was partly non-existent.

One afternoon I made 173 changes to our Meta campaigns in three and a half hours. The next day some campaigns showed as active with not a single ad inside, and stayed that way for a day and a half.

One city's campaigns spent in another city for days, because when we duplicated an ad set the location stayed the old one.

A countdown for a launch event: of more than a thousand people who were supposed to enter the automation, 12 got in.

An email sent to more than 13,000 people in other cities for a launch: zero bookings, 87 unsubscribes. 87 people from the crew who no longer want our emails, lost for nothing.

In June, in the team's sprint recap, it was written plainly: I was spending half my time on events. I'm the CMO, my job is the acquisition engine.

Four months later the numbers say this: over ten weeks the lever I'm measured on got on average less than a fifth of my hours, and in the week of a launch events took 42%.

Events are the best thing about play2match, which is exactly why it's so hard to say no to them 💛 The problem is that at every event we made the same mistakes again. At the Rome launch four problems were identical to the ones in Milan, three months earlier. We'd been talking about a playbook for months, but we didn't write it.

This is the most CMO mistake of them all.

In August we were bringing in about 1.7 times the people our courts could get playing. In one city we signed up more than 1,500 people in five weeks with a single sports centre. We were paying today for people who would play in January.

And before that I was opening courts even the day before because between 7 and 10 people were on the waiting list. Then those people didn't book.