Giving a number you are not sure of

The idea

When a client asks “what are our chances?”, they are not asking you to be right. They are asking how much to rely on you, and how to plan. So give a range, say out loud what makes it that wide, and say when it will narrow.

Two answers fail. A single confident figure hides the spread — and the day it is wrong, everything else you said loses value. Refusing to say anything sounds careful in your head, but it lands as evasion: the client is left holding nought to a hundred and no way to make a decision.

How it works

Say the estimate as three moving parts, in this order. It takes about twenty seconds out loud.

1. The range, with a centre. “Call it low fifties — say forty-six to sixty-four.” A centre gives them something to plan against; the range tells them how hard to lean on it.
2. What sets the width. Name the two or three specific unknowns doing the work. This is the part that converts a guess into a judgement.
3. When it narrows. A date and an event. “After the causation addendum on the fourteenth I expect that to tighten by roughly ten points.”
4. The floor. Say plainly what will never resolve — how a witness performs, how a tribunal exercises discretion. That is why the band never collapses to zero.

The width is not decoration; it is a book-keeping exercise you can actually do. In the model above:

base uncertainty  = ±30 pts   (pleadings only, nothing verified)
irreducible floor = ± 6 pts   (witness performance, judicial discretion)

fact value in points of width removed:
  records reviewed .......... 7      comparable outcomes ....... 4
  our breach expert ......... 8      causation opinion ......... 6
  claimant's expert seen .... 5

Suppose you hold records + comparators + breach expert:
  removed  = 7 + 4 + 8            = 19
  half-width = max(6, 30 - 19)      = ±11
  centre   = 50 + 3 + 1 + 6       = 60      (each fact also moves the mid)
  answer   = "about 60% — 49 to 71"
  outstanding: causation opinion, claimant's expert
  next narrowing: causation addendum, ~9 weeks

You do not need this arithmetic on a call. You need its shape: width comes from named gaps, gaps have dates, and some width never goes away.

When to use it

Where it does not fit: questions of fact you can simply look up, and moments where a decision is already made and the client needs a recommendation, not a distribution. And never quote a number you would not be willing to have read back to you in writing.

Watch out for

• False precision that ends the conversation. “Sixty-five per cent” invites no follow-up. Nobody asks what would change it, so nobody prepares for the other thirty-five. Precision beyond your evidence is a promise you did not mean to make.
• Hedging dressed as caution. “Too early to say” is only honest if you also say what would let you say it, and when. Without that, it is indistinguishable from not having thought about it — and the more you clearly know, the more evasive it reads.
• A range so wide it is a refusal in disguise. “Somewhere between ten and ninety” transfers the whole problem back to the client. If your range spans most of the scale, say so explicitly and give the one fact that would halve it.
• Silent revision. Moving from “fifty-fifty” to “we should settle” without ever saying which fact moved you destroys the value of every earlier number. Update out loud: old figure, new figure, the reason, in one sentence.
• Confusing a base rate with your judgement. Published success rates are about a population; your client wants a figure about this case. Say whose number it is: “the general run of these claims sits around X; this one I put lower, because…”.
• Numbers where the person wanted meaning. Especially in prognosis. Check what they are trying to decide before you answer with a percentage — sometimes the useful answer is a range plus “and here is what that means for the next three months”.

Worked example

An interviewer says: “Six weeks before the expert report, the trust’s claims manager calls and asks whether you’ll win. What do you say?”

You start with the number, because withholding it is the trap: “Today I’d put us a little either side of even — call it fifty-five, with a real range of about forty to seventy.” Then the reasoning that sets the width: the records are complete and support our chronology, but we have no breach opinion yet, and causation is the live question — whether earlier action would actually have changed the outcome. “Those two are why it’s a thirty-point band rather than a ten-point one.”

Then the date and the movement: “The breach report is due on the sixth; if it reads as I expect, I’d narrow to roughly fifty to sixty-five. The causation addendum three weeks after that is the one that could move the centre, in either direction.” Finally the floor: “Even fully evidenced I won’t get below about a ten-point band — how the treating clinician gives evidence is not something I can price.”

That answer is four sentences and gives the claims manager everything they need: a reserve figure, the two things to watch, a diary date, and an honest limit on your own precision. It is also unfalsifiable in the good sense — when the number moves, you have already told them why it would.

Check yourself