Your Company Cannot Remember Last Year

Try this in your next planning meeting. Ask what a paid social opportunity cost you, all in, in the third quarter of 2024.

Watch what happens. Someone will open a spreadsheet. Someone else will say the numbers moved when the attribution setting changed. Then a third person says the words that end the conversation: “That was before we moved off the old CRM.”

And that is it. The question dies. Not because the answer is complicated, but because your company does not have one. The record is gone. You are running a business that cannot remember what it did.

Your data has expiration dates you never agreed to

Nobody at your company decided to delete two years of marketing history. It happened anyway, in about six places at once, because every tool in your stack ships with a retention policy and none of them asked your permission.

Google Analytics 4 is the loudest example. On a standard property, event level data caps out at 14 months, and the default when you set it up is 2 months. If you never touched the setting, and most people never touched the setting, your granular web history is a rolling window that quietly drops the far end every month.

The Universal Analytics shutdown was the same story with a firmer deadline. Standard properties stopped collecting in July 2023, and Google began deleting the historical data the following year. A lot of companies learned what “deleting” meant at the exact moment somebody asked for a five year trend line.

Then there are the migrations. When you move CRMs, contacts come over. Companies come over. Open deals mostly come over. Activity history, the emails and calls and stage changes that actually explain why a deal moved, is the first thing cut, because it is the hardest thing to map and the person doing the migration is already behind schedule. Nobody announces it. It surfaces eighteen months later, when someone wants to know how long the sales cycle used to be.

And then the quiet one. Your ad platforms keep reporting, but they change the model underneath you. A conversion counted one way in 2024 and a different way now. The old number is still sitting there in the interface. It just no longer means what the new number means. That is worse than deletion, because deletion at least announces itself.

“We exported everything” is not a backup

Every company that has ever churned off a tool had the same plan. Export before we cancel. And they did it, honestly. There is a folder somewhere, usually on the drive of somebody who has since left, with a dozen CSVs in it.

Go open that folder sometime. You will find a table of activity records keyed to contact IDs. The contacts are in a different file, keyed to account IDs. The accounts reference a set of stage names that only ever existed inside the CRM you stopped paying for. Half the columns are internal identifiers. One is called custom_field_17 and nobody alive knows what it held.

The problem is not the file format. It is that the meaning lived in the system, and you only took the rows.

That is what makes leaving a CRM or an analytics platform so much worse than it looks. The stage names, the scoring logic, the join keys, the reason a field existed at all: none of it is in the download. The export satisfies a checkbox on the offboarding doc and then sits there unreadable, which was the entire reason to keep it.

Nobody is going to fight for this

Institutional memory has no owner. It is not on anyone’s goals. No marketing leader has ever been promoted for keeping 2023 queryable. There is no dashboard tile called “our past.” The work is invisible when it is done right, and it only becomes visible as a problem years later, to whoever happens to be sitting in the chair.

So the default wins. Every time. Retention settings stay where the vendor put them. Migrations drop whatever is expensive to carry. The old system gets switched off on the last day of the contract, the history goes with it, and nobody connects those two facts for years.

What it actually costs you

Losing your history does not feel like losing anything. It feels like being bad at your job.

• You cannot tell seasonality from a trend. Without three years of the same month, every August dip is a crisis and every September bump is a strategy.
• You re-run experiments that already failed. The test feels new to everyone in the room, because the people who remember it left and the data that settled it is gone.
• Your benchmarks turn into vibes. “That is about normal for us” becomes one person’s recollection, which is a nice thing to have and a terrible thing to plan against.
• You cannot answer the board’s favorite question. Boards ask about multi-year trends. Multi-year trends require multi-year data.

None of that shows up as a line item. It shows up as a team that argues in circles, and a finance leader who slowly stops believing marketing numbers, because marketing numbers keep changing shape.

The boring fix, which is also the only fix

This is not a big project. It is an afternoon and a spreadsheet, and almost nobody has done it.

Start with the systems that actually collect: analytics, your CRM, the ad platforms, whatever runs your store. For each one, write down a single number, which is how long it keeps event level history before it rolls off. Not what you assume. The real setting, in the admin panel nobody has opened. Most teams cannot produce that number for one system of record they run on, and the shortest clock is usually in the system they worry about least.

Then fix the free stuff. Extending GA4 retention to its maximum is a setting, not a project, and it costs nothing. Do it today and it still will not recover what already rolled off. That is the point. Retention decisions only work forward.

Then write your definitions down somewhere that is not a tool. What counts as a qualified lead. What counts as a customer. When the fiscal year starts, and what changed the last time someone redefined any of it. Definitions rot faster than data does, and when they go, old numbers become unreadable even when you still have every row.

And before the next migration, decide out loud what history matters and where the raw record is going to sit afterward. Not “we will export it.” Name the place: a warehouse, a database, somewhere the rows keep living once the old system is switched off. If you cannot name it in one sentence, you are not migrating your history. You are ending it.

One more thing

The reason this stays broken is not technical. It is that your history is scattered across a dozen products you rent, each running its own clock, and there is no single place where the question “what happened” can even be asked.

That is the part we cared about when we built THE DASHBOARD at I Hate Marketing. It connects the tools you already pay for and puts the question in one place, so asking is one question instead of a scavenger hunt across six logins and somebody’s memory. The clocks are still the clocks. But what your systems are holding right now should at least be something you can ask about in a meeting, and for most teams it currently is not.

Either way, go look at your retention settings this week. We are around if the looking turns up something ugly.