The Model That Couldn't Tell Me the One Thing I Wanted to Know

For years, part of my job was working with specialist companies to build models that proved whether our marketing had actually worked. It's the multi-million-dollar question every Finance Director asks, and the one very few of them believe the answer to. I've realised my retirement plan is the same kind of model, with the same blind spots. And the thing it's worst at measuring is the thing that matters most.

August 2026 : 5 min read - Part of the My Personal Journey to Early Retirement series at FreeBefore65.

A message last week from someone I used to work with sent me back to a piece of work we did together a few years ago. We wanted to know whether our marketing spend was actually earning its keep, so we used a specialist firm to build econometric models of it. Market mix modelling, if you want the term. In plain English, you feed in years of data, sales, advertising spend, prices, economic indicators, even the weather, and the model tries to untangle how much each thing you did actually drove the result, as against everything else going on at the same time.

It was genuinely useful. It could tell us, with reasonable confidence, that one campaign had paid back and another hadn't, which is a hard thing to know otherwise. But it came with objections we heard every single time, and I've been turning them over on the train back to the holiday cottage, because they're the same ones I'd now level at my own retirement plan.

The first was that it only knew the past. The model was built entirely on historic activity, and the future was full of things it couldn't see coming, or that were simply outside our control. A competitor doing something daft, or a shift no model trained on last year could have predicted. The second was that it was only ever as good as what you fed it. Coarse inputs gave coarse answers. And the third, the one that nagged most, was that the numbers we could measure were the easy part. The spend and the ratings and the price points were all quantifiable. The things that were hardest to put a value on, the strength of the brand, or whether the creative was any good, were often the ones doing the heavy lifting.

I built a retirement model too. A spreadsheet, tabs for assumptions and scenarios, the works. And it has every one of those flaws.

It knows only the past. It runs on historic investment returns, on what I've spent before, on the tax rules as they stand today, and the next thirty years will not politely follow any of it. It's only as good as my inputs, which means it's only as honest as I've been about what my life actually costs. And the biggest limitation is the same one. The money is the quantifiable part, and it turns out to be the tractable part. Whether I'll be content, and who I even am without a job to point at, are the inputs that decide whether the whole thing works, and they're exactly the ones the model can't hold.

There's a term for the gap. In the marketing models there was always a base, the level of sales that would have happened anyway, and an error term, the part the model simply couldn't explain. You optimised around the edges, the incremental bit you could actually influence, and made your peace with the base and the noise dwarfing it. My retirement is mostly base. Draw-down order and tax sequencing are the levers I can model and move. How long I live, what the markets do, and whether I end up happy. That's the base, and it dwarfs the rest.

For someone who spent a career wanting tight confidence intervals and clean attribution, that's the uncomfortable bit, and I doubt I'm alone in it. People trained to evaluate things quantitatively tend to want the number to be the answer. But we never actually treated those models as crystal balls. They were decision aids. I always told my teams the same thing; the numbers inform the decision, they don't make it. You looked at what they told you, understood how much they left out, and committed the spend anyway, because not deciding was also a decision.

I've ended up being my own Finance Director on this one, asking the million-dollar question and only half-believing the answer, and going ahead regardless. Retiring early was the same. I ran the numbers as well as anyone could, then had to act on a model that, by its nature, couldn't tell me the thing I most wanted to know.

The spreadsheet said I could afford to stop. It had no column for whether I'd be glad I did. I'm four weeks into finding that part out.

Further reading:

• The FreeBefore65 Tracker
• The Professional Stress-Test: Why I’m Taking My Spreadsheets to an IFA (And How You Can Too)

Part of the My Personal Journey to Early Retirement series at FreeBefore65.

Tony writes about his personal journey to early retirement at freebefore65.co.uk. He is not a financial adviser. All content reflects his own experience and research and should be taken as a starting point for your own thinking, not as professional advice.