Running forecast reviews that change decisions

0:00 Where numbers become decisions

Most forecast reviews I've attended fall into one of two traps. Either they're a ritual, where each manager reads out their numbers and nobody challenges anything. Or they're a trial, where challenge feels personal and people learn to hide bad news. Neither produces an honest forecast. In this lecture you'll learn how to run a forecast review that turns numbers into decisions: a tight agenda, a consistent set of challenge questions, the difference between symptoms and root causes, variance thresholds that tell everyone what must be explained, and practical techniques for countering optimism bias. By the end you'll be able to chair a review that produces an owned forecast and a clear action list in about an hour.

0:51 Why it matters

Why does the review matter so much? Because a forecast is only as good as the conversation that produces it. Formulas give you a range. Bottom-up estimates give you the team's view. But the review is where someone asks, 'what would have to go wrong for this to be ten per cent worse?' and the room answers honestly. It's also where culture is set. If people are punished for bringing bad news, they stop bringing it. And a controls system that can't see bad news is worse than useless, because it gives leaders false confidence. So think of the review as the moment the whole control cycle either works or quietly fails.

1:40 The agenda

Here's the agenda I use, for sixty to ninety minutes. First, the headline, five minutes: EAC versus budget and contingency, forecast finish versus baseline, and what's changed since last month. Second, top variances, twenty minutes: the three to five control accounts with the largest adverse cost variance or slip. Third, forecast challenge, twenty minutes: compare each manager's forecast with the formula range and the TCPI. Fourth, risk and change, fifteen minutes: top risks, pending changes, contingency position. And fifth, decisions and actions, ten minutes: owners, dates and any escalation. Notice the last segment. If you run out of time, it's the one that gets squeezed, and it's the only one that changes the future. So protect it.

2:31 The concept: fair challenge

Here's the key idea: challenge feels fair when it's consistent. Ask the same questions of every manager, every month, and it stops being personal. Think of it like a pilot's checklist. Nobody's insulted when the co-pilot reads it out. Here are my questions. What's the root cause of this variance? Keep asking why until you reach something controllable. Is it one-off or systemic? What does your estimate to complete assume about productivity, prices and quantities, and what evidence supports that? What's the TCPI implied by your forecast, and how does it compare with your CPI trend? What pending changes or claims aren't in the forecast yet? And finally, what would have to go wrong for this forecast to be ten per cent worse? That last question surfaces the risks people are quietly hoping won't happen.

3:30 Worked example one: symptom versus root cause

Let's look at a simple example of what good looks like. A manager explains a variance: 'Labour costs are higher.' That's a symptom. It tells you nothing you can act on. Ask why. Because cable-pulling crews are less productive than estimated. Why? Because access to the risers was delayed by the ceiling contractor. Now we have something. The strong version reads: crew productivity on cable pulling is thirty per cent below estimate because riser access was delayed by the ceiling contractor. The impact is so much cost and so many days. The owner is the building services manager, and the response is to agree a revised access sequence by Friday. Cause, quantified impact, owner, response. That's the standard every explanation should meet.

4:23 Worked example two: UK rail signalling

Now a realistic scenario from the lesson. A fictional UK rail signalling upgrade presented EAC equal to budget for six months running. A new controls lead introduced TCPI into the review. For the largest control account, CPI was nought point eight four, and the TCPI implied by the manager's forecast was one point two one. So the question was asked, calmly: what will change to deliver a forty-four per cent improvement in efficiency? The manager couldn't name anything specific. The forecast was revised to a range of eight to twelve per cent over budget. The sponsor released management reserve for one specific risk, and a scope phasing option was agreed with the client. None of those decisions would have happened with a green report. That's the value of a good review.

5:20 Watch me do it: preparing the review pack

Let me show you how I prepare the pack the day before. I start with the EVM table by control account and sort by cost variance, then by forecast slip. I take the worst three to five. For each, I add the formula EAC range and the TCPI implied by the manager's current forecast, and I flag any where TCPI is more than ten per cent above achieved CPI. Then I write two or three specific questions for each account in advance, so the challenge is prepared, not improvised. This is a place where an approved AI assistant helps: give it the structured table and ask it to list the top movers, flag TCPI gaps and draft challenge questions. Then I check every number against the source before the meeting, and I own what goes in the pack.

6:20 Thresholds, optimism bias and mistakes

Two more ingredients. Thresholds tell everyone what must be explained. For example, and this is illustrative, a control account with cost or schedule variance worse than five per cent and twenty-five thousand needs a written explanation and action. CPI or SPI below nought point nine five for two periods needs a recovery plan to the steering committee. And EAC above budget plus remaining contingency triggers formal escalation. Then, optimism bias. People systematically underestimate cost and duration. Counter it with reference-class thinking, comparing with outcomes of similar past projects. Use pre-mortems. Keep controls' view independent. And track forecast accuracy month by month against the final outcome. The classic mistakes: reviews that present numbers but make no decisions, letting the loudest person set the forecast, and punishing honest bad news.

7:16 Recap and try this now

Let's recap. A forecast review is where numbers become decisions. Use a tight agenda and protect the decisions segment. Ask the same challenge questions of everyone, every month, so challenge feels fair. Insist on root causes, not symptoms: cause, impact, owner, response. Set thresholds so people know what must be explained and what escalates. And counter optimism with reference classes, pre-mortems, an independent controls view and a record of forecast accuracy. Close every review by logging actions, and open the next one by closing out last month's. Your try-this-now: draft variance thresholds for a project you know, at control account and project level, and write the three challenge questions you'd always ask.