The history AI can't give you
Last time I wrote about what AI doesn’t bring with it: years of tested experience. There’s a second thing it can’t give you, and it’s closer to home. It can’t give you your own history.
Here’s a simple test. Could your last three years be pulled up side by side, month against the same month, with the numbers meaning the same thing each year? Leads, booked estimates, closed jobs, what was spent to get them.
For a lot of companies the honest answer is no. Not because anyone was careless. The data was just never kept with comparison in mind.
What the gap looks like
Here’s a hypothetical. It’s April, and calls are down from last April.
Is that the weather? A slower spring across the whole market? The ad that was changed in March? A competitor who started bidding on your name? Or is it nothing, just a normal wobble?
Without history, there’s no way to tell. And it gets worse when you try. Last year a “lead” might have counted every call, including the spam and the wrong numbers. This year the new CRM only counts the ones someone qualified. The ad account was rebuilt in between. The website form changed. So even where last April’s number exists, it isn’t measuring the same thing as this April’s.
That’s the real gap. Not missing numbers, but numbers that can’t be compared.
AI can’t fill it. Asked why April is down, it will give you a thoughtful list of possible reasons. It can’t tell you which one is true for your company, because the record it would need was never kept.
Why history matters so much in this industry
Green industry work is seasonal, sharply so. A slow week in February and a slow week in May mean completely different things. The only fair comparison for this spring is last spring, and the spring before that.
That’s what history buys you: the ability to look at a number and know whether it’s good. Without it, every decision is made against a feeling.
History is not the same as testing
One thing worth being clear about. Building seasonal history takes time, but learning doesn’t have to wait for it.
A controlled test, two versions of an offer or a page run side by side, under the same conditions, can tell you something useful within a season. You don’t need three years of records to find out which of two landing pages gets more calls. History tells you whether the whole year went well. Tests tell you what to change while it’s happening. Both matter, and neither replaces the other.
How we build usable history
When we start running a company’s marketing, part of the work is making sure next year can be compared with this one. That means a few unglamorous things:
- Consistent definitions. A lead, a booked estimate and a closed job mean the same thing every month, written down, and they don’t quietly change when a tool does.
- Seasonal comparisons. Every month is read against the same month before, not against the month next to it.
- Controls. When something changes, the ad, the offer, the page, we note it, so the effect can be separated from the season.
While a new client’s own history builds up, we don’t start from nothing. We start from what we’ve seen work for other clients in similar situations. But that comes with care. Another company’s numbers are only a useful guide where the work is truly comparable: the same kind of service, a similar market, a similar season. So we say what the baseline is based on and where it stops applying. What carries over from one client to the next is the lesson, never the records.
Then, as the company’s own history grows, it takes over. Borrowed experience gets you started. Your own record is what you steer by.
The quiet advantage
Nobody gets excited about keeping definitions consistent. But a few years from now, the companies that did it will be able to look at any month and know what it means. The ones that didn’t will still be guessing, with better tools to guess with.
AI makes the guessing more articulate. History is what makes it unnecessary.
Chris