He put it on the desk between us. A3 paper, landscape, colour-coded. Forty-seven rows, twelve columns. The footer read ‘v6.2 FINAL.’ Margaret sat beside him with her arms folded—not hostile, but guarded. The way people sit when they already know the conversation won’t go the way the spreadsheet suggests.
They’d come to me because they couldn’t agree on when to retire. David was 61. Margaret was 58. He wanted to stop at 62. She wanted to keep working until 63, maybe 64. The spreadsheet was supposed to settle it. The spreadsheet was supposed to settle me.
I want to be clear about what happened next, because it’s the part that matters and it doesn’t show up in the cells. I didn’t tell David his spreadsheet was wrong. I didn’t tell Margaret she was right. I didn’t open my laptop and produce a competing model with shinier assumptions and more decimal places. What I did was ask them both to walk me through it, line by line, while I took notes on what they said between the numbers.
What the Spreadsheet Was Actually Doing
The model was perfectly competent. David had used a 5% net annual return, 2.5% inflation, their current combined pension pots, projected State Pension at their respective State Pension ages, and a spending assumption of £42,000 per year in retirement. He’d built in a buffer. He’d accounted for the mortgage ending in 2027. He’d even included an estimate for one long-haul holiday per year, which Margaret later told me she had never once agreed to.
The model proved David could retire at 62. Final pot balance: £380,000 at age 90. The model was also, in a very real sense, beside the point.
Over the first twenty minutes of walking through it, I noticed three things. First, David had specified his return assumption to one decimal place—5.0%—as though the extra precision made the projection more reliable. It doesn’t. It just makes the output look more authoritative. Second, he’d left the column for Margaret’s pension contributions between ages 58 and 62 entirely blank, because filling it in would have meant acknowledging that her working longer changed the math in ways that pushed his retirement date backward. Third, the spending assumption of £42,000 wasn’t a number they’d arrived at together. It was the number David needed the model to produce for his retirement date to work.
This is what spreadsheets do when they’re built under emotional pressure. They don’t model the future. They manufacture certainty about a decision the builder has already made but hasn’t admitted to themselves. The precision of the inputs is doing the emotional work of a conversation that hasn’t happened yet.
How a Good Advisor Reads a Client’s Model
When a client brings their own projections, the temptation for an advisor is to treat it as a technical document to be audited. Check the formulas. Question the return assumption. Produce a corrected version. That approach assumes the spreadsheet is the argument. It isn’t. The spreadsheet is the client telling you what they’re afraid of, formatted as a cash flow forecast.
What I actually did with David and Margaret’s model was slower and less impressive-looking. I asked David to talk me through how he’d chosen 5% as his return assumption. He said he’d looked at long-run equity market returns and deducted a bit for conservatism. I asked him what ‘a bit’ meant. He said he wasn’t sure. I asked Margaret what she thought the spending number should be. She said £48,000, because she wanted to help their daughter with a house deposit and David had left that out entirely. I asked David why Margaret’s contribution column was blank. He went quiet.
The real planning work happened in the silences between those questions, not in the cells. A spreadsheet can tell you whether a number works. It can’t tell you whether both partners agree the number is real.
This is why the discipline of stress-testing assumptions matters more than the discipline of building models. Any model that assumes a fixed 5% annual return for thirty years is a fiction, and not a particularly useful one. The FCA’s consumer guidance on investing is blunt about this: all investments involve risk, markets fluctuate over time, and knowing your time horizon and personal tolerance for risk is foundational. A deterministic spreadsheet that ignores both isn’t conservative. It’s a story dressed up as arithmetic.
What I told David and Margaret was that the model was a starting point, not a verdict. We would rebuild it together, with inputs they both recognised, and we would test it against historical data rather than comfortable round numbers. UK economic time series for inflation, interest rates, and market returns are publicly available through resources like the ONS inflation and economic data series, and the point of referencing them isn’t to forecast the future. It’s to show clients what actually happened across thirty-year windows in the past, because the range of outcomes is always wider than a single column of projected values suggests.
The Variables People Leave Blank on Purpose
In my experience, the blank cells in a client-built model are more diagnostic than the filled ones. People don’t leave things out randomly. They leave them out because filling them in requires a decision they’re not ready to make or a conversation they’re not ready to have.
David left Margaret’s contributions blank because acknowledging them would have delayed his retirement date. Margaret hadn’t built her own version because building one would have meant saying out loud that she didn’t want to retire at the same time as David, and she wasn’t sure how that conversation would land. Neither of them had modelled what happens if one of them dies first, because that’s the column nobody wants to populate.
A good advisor’s job at this stage isn’t to fill in the blanks for the client. It’s to name them. To say, gently and clearly: ‘You’ve left this column empty. I think that’s because it involves a conversation you haven’t had yet. Shall we have it now, or shall we come back to it?’ The spreadsheet becomes a map of what the client is avoiding, not just what they’ve calculated.
This is also why I pay close attention to which inputs a client has over-specified. A return assumption of 5.0% to one decimal place, when the client can’t explain how they arrived at it, isn’t diligence. It’s a person reaching for certainty in a place where certainty isn’t available. An inflation assumption of exactly 2.5%, when CPI has ranged from near-zero to over 11% within the last fifteen years, is a similar signal. The precision of the input is inversely proportional to the confidence the builder actually feels. People who are genuinely uncertain specify to two decimal places. People who are comfortable with ambiguity round to the nearest whole number and say ‘let’s test what happens if we’re wrong.’
Why the Documentation Matters After the Meeting
Here’s something that surprised me when I first noticed it, and still surprises me now: clients re-read their planning documents between meetings. Not all of them, and not always, but enough that it changes how I write them. The cash flow forecast, the suitability report, the meeting notes—these aren’t compliance artefacts that get filed and forgotten. They’re the record a client returns to at 11pm on a Sunday when doubt has crept back in and the spreadsheet from the kitchen table is starting to look optimistic.
This means the quality of that documentation matters in a way that goes well beyond regulatory obligation. A planning report that’s vague, hedged, or full of generic caveats won’t help a client who is sitting alone with their uncertainty. A report that’s specific about what was assumed, what was tested, what was left as an open question, and what the client and advisor agreed to revisit—that report does something useful. It gives the client something to re-read that accurately reflects where they actually are, rather than where they wish they were.
That same discipline applies to editorial structure: before publishing, editors need a way to test scattered notes become an argument readers can follow, which is where an AI writing software that fits the project can function as a planning aid rather than a substitute for domain evidence.
I think about this the same way I think about any writing that has to carry weight under pressure. The discipline of writing clearly—whether in a planning report, a client-facing summary, or any structured document someone will return to when they’re worried—forces the writer to confront which parts of their argument are evidence and which are comfort. When I’m drafting a report that has to separate a tested assumption from a hopeful guess while keeping the structure legible enough for a client to follow at 11pm, tools like AI writing software can help with the mechanics of keeping that structure visible. But the judgment about what belongs in the document and what doesn’t—what is a tested assumption versus what is a hopeful guess—still has to come from the person who sat across the desk and listened to the silences.
The Conversation That Wasn’t About Money
David and Margaret came back for a second meeting two weeks later. They had talked. Not about the spreadsheet, but about what the spreadsheet was standing in for.
David wanted to retire at 62 because he was tired. Not in the physical sense—he was still fit, still engaged with his work—but in the way that accumulates when you’ve been the primary earner for thirty-five years and you’ve started to feel your identity shrinking to the size of your payslip. He wanted to stop because he wanted to be something other than the person who earned the money. He’d never said this to Margaret because he didn’t have the language for it, so he built a spreadsheet instead.
Margaret wanted to keep working because she was good at her job and she’d started it later than David started his. She wasn’t protecting the income. She was protecting the part of her life that felt like hers. She hadn’t said this to David because the spreadsheet had already decided things, and she couldn’t find a column in it for ‘I’m not ready.’
These aren’t financial planning problems. They’re marriage problems that arrived disguised as pension timing problems. But they’re also exactly the kind of problems that walk through an advisor’s door, because money is the language people use when they can’t find the other words. The spreadsheet was David’s way of saying ‘I want to stop’ without having to admit he was scared of what stopping meant. Margaret’s silence was her way of saying ‘I’m not done yet’ without having to justify why her timeline was different from his.
What we did—what a good advisor does in that moment—was give them a structure for a conversation that didn’t require either of them to abandon their position. We rebuilt the model with both of their inputs, including the £48,000 spending figure and Margaret’s contributions through to 63. We modelled David retiring at 62 with Margaret continuing to work, and we modelled both retiring at 63. We looked at what changed. We looked at what didn’t.
The math said David could retire at 62 and Margaret could continue to 63, and the plan would still hold. But that wasn’t the point. The point was that they’d finally had the conversation the spreadsheet was trying to replace, and the numbers were now a tool in that conversation rather than a substitute for it.
What I Wish More People Understood About Financial Models
A financial model isn’t a prediction. It’s a structured way of asking ‘what would have to be true for this to work?’ and then checking whether those conditions are reasonable. The moment a model becomes a prediction—when the client treats the output as a fact about the future rather than a question about it—it has stopped being useful and started being a source of false confidence.
This is why I’m suspicious of models that produce a single answer. ‘You can retire at 62’ isn’t an answer. It’s one outcome among many, and the question is whether the other outcomes are survivable. A model that shows a range—with a lower bound the client can live with and an upper bound they’d be content with—is a model that’s doing its job. A model that shows a single number with two decimal places is a model that’s doing something else entirely.
The clients who benefit most from financial modelling aren’t the ones who build the most elaborate spreadsheets. They’re the ones who use the model to identify what they don’t know, what they’re avoiding, and what they need to discuss with the person they’re planning alongside. The model is a prompt, not a conclusion. When clients understand that, the spreadsheet becomes a conversation starter instead of a conversation ender.
And when they don’t—when the model becomes a weapon in a debate about timing, or a shield against the uncertainty of a decision that feels too big to make without numerical cover—it’s the advisor’s job to notice that, name it, and gently redirect. Not by tearing up the spreadsheet, but by asking the questions the spreadsheet can’t answer. What are you afraid of? What haven’t you said to each other? What would change if the number were different?
What David and Margaret Decided
David retired at 62. Margaret continued to 63. They revised their spending assumption to £46,000, which was a genuine compromise—higher than David’s original, lower than Margaret’s, arrived at after they actually looked at twelve months of bank statements together, which is an exercise I recommend to every couple I work with and which very few of them have done.
The spreadsheet sits in a folder now. David told me, six months later, that he hadn’t opened it since the second meeting. ‘It did its job,’ he said. ‘It got us in the room.’ Margaret said something I’ve thought about often since: ‘I wish I’d built my own version. I think I would have learned something about myself.’ I told her she probably would have, and that learning something about yourself isn’t the same as winning the argument.
The model didn’t solve their problem. The conversation did. But the model is what got them to the conversation, and that’s worth something. It’s just not worth what most people think it’s worth. It’s not a verdict. It’s a door.
The Question to Sit With
If you’ve built a spreadsheet to settle a financial decision—retirement timing, pension contributions, whether to pay down the mortgage or invest more—ask yourself which inputs you’ve specified with suspicious precision and which you’ve left blank. The blank cells are where the real planning work is. The over-specified cells are where your anxiety lives. A good advisor won’t fill in the blanks for you. They’ll help you notice them, sit with them, and eventually have the conversation the spreadsheet was trying to replace.
What column have you left empty, and why?