She came in near closing, holding her phone the way you hold something that has betrayed you. It had stopped showing her the photographs of her grandchildren. She had already been told, by two different people, that the problem was probably her. That she had pressed something.
It wasn't a complicated fix. Most of them aren't. What took the time was the part before the fix, where she explained, carefully and slightly apologetically, why those particular photographs mattered, in case I needed convincing.
Then, before she left, she asked whether we could take a photo together.
Not of the phone. Of us.
She held it up afterwards and checked it, the way people check a picture they intend to keep. Then she thanked me and walked out, and I stayed standing there a second longer than I needed to.
I've thought about that photograph more than any campaign I've worked on.
What the data could see
By then I had spent a long time inside a CRM. I could tell you, with reasonable confidence, which accounts were about to leave. Usage falls off a shelf. A complaint sits open two days longer than it should. The same question arrives three times in a fortnight.
Data is very good at this. It will point at a person and say, with something close to certainty: this one is going.
What it won't tell you is why. It records that she called three times. It doesn't record that nobody used her name. It logs the resolution and closes the ticket, and the ticket is closed, and she is still leaving.
That gap is the whole job. Everything I find interesting about marketing lives inside it.
The year I was the one who needed noticing
I came to Australia from Bangladesh with a suitcase and a university offer, which at the time felt like an enormous amount of preparation and turned out to be almost none.
Nobody warns you about the quiet part. Not the weather, or the accent, or the paperwork. The quiet part is how many days you can go without anyone saying your name out loud.
Then someone does. Someone in a corridor asks whether you've found the right building yet, and waits for the answer. It's such a small thing that it feels embarrassing to admit how much it holds you together.
I signed up as a peer mentor not long after that, and spent orientation weeks walking new students around a campus I had only recently stopped getting lost in. I told myself it was for them. Most of it was, and some of it was me trying to return something.
What I do with it now
Retention isn't a number. It's a person deciding, quietly and usually without telling you, whether this is worth the trouble.
So I read the data twice. Once for what it says, which is usually what. Then again for what it's circling, which is usually why. The first pass gives you a list. The second gives you something to actually say.
And when a plan comes back from that second pass looking smaller and less impressive than the one I started with, I've learned to take that as a good sign rather than a failure. The oversold version wins the meeting. The honest one tends to win the year.
- Data tells you who is leaving. Only people tell you why.
- The fix is rarely the point. The ten minutes before the fix usually are.
- If a strategy can't survive being said plainly to one person, it isn't ready.
I never got her name. I have no idea whether she still has the photograph. But she is the reason I open a dashboard expecting to find a person in it, and it hasn't let me down yet.