I built my first serious workbench from Tasmanian oak.
It took a weekend. It was solid, flat, well-jointed. One leg was perhaps half a millimetre short.
You’d never notice standing next to it. You’d only notice when you placed a long piece of timber across the surface and watched it rock.
A swaying workbench still lets you build things. Most things, in fact, turn out perfectly fine.
But the rocking is always there. And occasionally — when precision matters — it makes things go wrong in ways that are very hard to trace back to the bench.
A slightly-off financial model works exactly the same way.
Here’s what ‘slightly off’ actually looks like in a business:
A margin formula that rounds at the wrong step. A revenue projection that hardcodes last year’s growth rate. A cost model that forgets to include contractor loadings. A cashflow sheet where one input links to an older version of the file.
None of these feel catastrophic. Most of the time, the outputs still look reasonable.
That’s what makes them dangerous.
The most damaging financial model errors aren’t the ones that produce an obviously wrong number. They’re the ones that produce a plausible number — that happens to be wrong.
I’ve seen this play out in real decisions:
→ A tender priced at 8% margin because the model omitted a cost category. Actual margin: 1.2%.
→ A hiring decision made on projected cashflow that assumed a payment timeline that never materialised.
→ A contract renewed because the reporting showed it was profitable. It wasn’t. The model was pulling the wrong line.
In each case, the person using the model trusted it. Why wouldn’t they? It was the model they’d always used.
There’s a line between ‘close enough’ and ‘accurate enough to trust’.
Close enough works fine for estimates and rough planning. It’s perfectly appropriate for a quick back-of-envelope.
But when a model is driving real decisions — pricing, hiring, investment, forecasting — it needs to be accurate enough to trust. And that requires it to be audited, not just used.
Most financial models I’m brought in to review have never been audited. They were built by someone smart, under time pressure, with good intentions. Then they were used. Then they were trusted. Then they were never questioned again.
The workbench leg doesn’t get fixed because nobody thinks to look at it. The model doesn’t get fixed because nobody thinks to question it.
Until it matters.
If your business makes decisions based on financial models — and most do — it’s worth knowing whether the bench is level.
“The changes Jürgen made resulted in a substantial time saving and improved the reliability of our process — allowing us to focus on the analysis rather than the collation and manipulation of data.” — David Barton, COO, Future Super
If you’d like a second pair of eyes on a model you’ve been trusting for a while, let’s take a look together.
A Model Audit takes 30–60 minutes. It either confirms your model is sound — or it finds the leg that’s been slightly off the whole time. Either way, you’ll know.
→ Book a free 10-minute discovery call — no obligations.