“Where Did These Numbers Come From?” — The Question That Decides Whether Your Data Survives

In 2019 I was sitting in on a quarterly review as an adviser, watching a capable divisional head present a growth plan built on a market-sizing figure. Twenty minutes in, the most senior person in the room — who had said almost nothing until then — leaned forward and asked, mildly, “Where did these numbers come from?” The divisional head paused, then said, “My team put the market data together — I can check the exact source and come back to you.” It was an honest answer. It was also fatal. I watched the senior person sit back, and I watched the temperature of the room drop. For the next forty minutes every figure was questioned, the plan was picked apart, and it was ultimately parked. The market-sizing number, I later learned, was perfectly sound — sourced from a reputable industry report the team had used for years. The data survived. The presenter did not, because in the second that mattered he had handed ownership of his own numbers to an absent team.

(This article was created with AI assistance; all stories and insights are based on 35 years of real client work.)

That moment is the clearest illustration I know of a truth that data presenters learn late: the attribution question — “where did these numbers come from?” — is almost never a request for a citation. It is a test of whether you own your data. This piece is about how to pass that test. It covers why the question is a trust probe rather than a data query, why the natural defensive answer is the one that sinks you, and the three-line method that converts the probe into a demonstration of command: the Provenance Answer, delivered in source-confidence-caveat order, in under forty-five seconds.

Why the question is a trust probe, not a data query

A senior person who genuinely wanted the citation for a number would usually ask for it offline, or after the meeting, because the citation itself is rarely decision-relevant in the room. When they ask it live, in front of the group, they are doing something else: they are checking whether the person presenting actually commands the material or is merely relaying it. The question is a quick, cheap way to find out a great deal. A presenter who owns their numbers answers instantly and specifically. A presenter who is relaying someone else’s work hesitates, deflects, or promises to come back — and that tell is the real information the senior person was fishing for.

This is why the accuracy of the number is almost beside the point. The senior person is not, in that moment, evaluating whether the figure is correct — they usually have no way to evaluate that live either. They are evaluating you. Can this person be trusted with the decision this data supports? The attribution question is a proxy for that larger judgement, and it is a good proxy, because ownership of one’s numbers correlates strongly with having done the thinking behind them. Someone who can tell you instantly where a figure came from, how confident they are in it, and what they would do to firm it up has almost certainly done the underlying work. Someone who cannot, may not have. The deeper executive question of why a board should believe your numbers sits underneath the attribution probe; the probe is just its fastest, most common surface form.

Understanding the question as a trust probe changes how you prepare for it. You stop preparing a citation and start preparing a demonstration of ownership. The goal of your answer is not to satisfy a footnote; it is to show, in forty-five seconds, that you stand behind this data with full knowledge of its strengths and its limits. That is a different answer from “it came from the industry report,” even when the industry report is the honest source. The honest source is necessary but not sufficient; what completes the answer is the confidence and the caveat that prove you have thought about the number, not just retrieved it.

Why the defensive answer kills the deck

The defensive answers all share one feature: they move ownership of the number away from the presenter. “My team pulled those together.” “That came from finance.” “I’d have to check the exact source.” “Those are the standard figures we always use.” Each is an attempt to reduce personal exposure, and each does the opposite, because the question was a test of ownership and the answer just failed it by disowning. The senior person hears: this person does not actually stand behind these numbers, so I should not stand behind their plan. The deck does not die because the data was challenged; it dies because the presenter declined to own it under a mild test, and a board cannot back a recommendation that its own author will not visibly back.

The “I’ll come back to you” answer is the most seductive and the most damaging, because it feels responsible — it sounds like diligence. But in the context of a trust probe it reads as a confession that you do not currently know your own data, and the offer to check later does not repair the present-moment failure; it confirms it. By the time you come back with the citation, the judgement has already been formed and the meeting has moved on. The damage is done in the live second, and a follow-up email cannot reach back into the room and undo it.

There is a subtler failure too: the over-defensive answer that fights the question. “These numbers are absolutely correct, I checked them myself, I’m very confident in them” — delivered with heat — also fails, because it treats a trust probe as an attack and responds with defensiveness rather than command. Heat signals that the question landed somewhere tender, which tells the senior person they have found a soft spot, which invites more probing. The composed answer neither disowns the number nor fights for it; it simply demonstrates ownership, calmly, and moves on. Handling the range of questions a data presentation attracts almost always comes down to this temperature control: answer from command, not from defence.

The Provenance Answer: source, confidence, caveat

The Provenance Answer is three lines, always in the same order, because the order does specific work. Line one is the source: where the number actually came from and how it was reconciled or validated. “The market-sizing figure is from the [sector] industry report we’ve used for three years, cross-checked against our own bookings data for the segments we operate in.” This answers the literal question and demonstrates that you know the provenance cold. Line two is the confidence: which part of the number you would stake a decision on and which part is directional. “I’m confident in the total addressable figure; the segment-level split is directional, based on the report’s allocation rather than our own.” This is the line that builds the most trust, and we will come back to why. Line three is the caveat: what you would do to firm up the soft part, and by when. “If the segment split needs to be decision-grade, I can have it rebuilt from our own data by Friday.”

The three lines together do something the bare citation cannot. They show that you know not just where the number came from but how good it is — which parts are solid and which are soft — and that you already have a plan for the soft parts. That is the signature of someone who owns their data rather than relays it. A relayer can sometimes produce the source; only an owner can produce the confidence line, because the confidence line requires having thought about the number’s limits, not just its origin. When a senior person hears all three lines delivered calmly, the trust probe is answered in full: yes, this person commands the material.

The order matters because it front-loads the answer to the literal question (source) before moving to the more sophisticated content (confidence and caveat). If you led with the caveat — “well, the segment split is a bit rough” — you would sound unsure before you sounded authoritative, and the room would latch onto the weakness. Leading with the source establishes command first; the confidence line then demonstrates self-awareness from a position of strength; and the caveat closes with a forward action that shows you are already ahead of the concern. Source, confidence, caveat. The same three lines in a different order land far less well.

The forty-five-second rule

The Provenance Answer must fit in about forty-five seconds, and the time limit is not a stylistic preference — it is part of how the answer signals command. A short, complete, three-line answer says: I know this so well that I can give you the source, the confidence, and the plan in three sentences without reaching for notes. A long, winding answer says the opposite, even if every word is accurate, because length under a trust probe reads as the sound of someone assembling an answer rather than retrieving one. The senior person is reading your fluency as much as your content, and forty-five seconds of clean structure is more fluent than three minutes of thorough explanation.

The forty-five-second discipline also protects you from the trap of over-answering, which is its own failure mode. When you keep talking past the three lines — explaining the methodology, justifying the report’s credibility, walking through the cross-check in detail — you signal anxiety, and you give the room more surface to question. Every extra sentence past the Provenance Answer is an invitation to a follow-up. The disciplined presenter delivers the three lines, stops, and lets the silence sit. Stopping is itself a demonstration of command: I have answered your question completely, and I am comfortable waiting for your next one.

This is where the connection to the broader skill of handling questions becomes practical. The forty-five-second Provenance Answer is one instance of a general structure for decision-safe answers under pressure: a bounded, structured response delivered calmly and then closed off, rather than an open-ended explanation that bleeds anxiety into the room. The related challenge of “walk me through the numbers” uses the same discipline — a structured walk with a clear endpoint, not a meander through the whole model — because the underlying principle is identical: command shows in structure and boundedness, not in exhaustiveness.

Why volunteering the soft number builds trust

The most counterintuitive part of the Provenance Answer is the confidence line, because it volunteers a weakness. Most presenters instinctively hide the soft parts of their data, reasoning that admitting a number is “only directional” hands the room ammunition. In a trust probe, the opposite is true. Volunteering which parts of your number are soft is the single most powerful trust signal available to you, because it proves you are not selling — you are informing. A presenter who flags their own data’s limits before being asked is demonstrating that they will tell the room the truth even when the truth is inconvenient, and that demonstration is worth more to a senior audience than a number that is one degree firmer.

I learned this from the other side of the table, in my banking years, sitting on committees where people came to present. The presenters I trusted were not the ones with the cleanest numbers; they were the ones who told me, unprompted, which of their numbers I should lean on hard and which I should treat as indicative. Their willingness to mark their own data’s confidence levels made me trust the parts they were confident about more, not less, because I knew they were not overselling. The presenters who claimed everything was solid made me suspicious of all of it, because nobody’s data is all solid, and a presenter who would not admit the soft spots was either naive about their own work or hoping I would not notice. Both are worse than honest calibration.

There is a discipline to the confidence line that keeps it from sounding like an apology. It is stated as a fact about the data, not a confession about the work: “the segment split is directional” rather than “I’m afraid the segment split isn’t very good.” The first is calibration; the second is self-criticism, and self-criticism under a trust probe reads as weakness. Calibration reads as command. The same fact — that the segment split is less reliable than the total — can be delivered as a strength (I know exactly how reliable each part is) or a weakness (I’m not sure about this), and the difference is entirely in the framing. State the confidence level as something you know, not something you fear.

Preparing the answer before the question comes

The Provenance Answer cannot be improvised well under pressure; it has to be prepared, because the calm three-line structure is exactly what falls apart when a senior person leans forward and the adrenaline rises. Preparation is straightforward. Before any data presentation, take the handful of numbers the recommendation rests on — the same headline figures that carry your decision slides — and for each one write the three lines: source, confidence, caveat. Writing them does two things. It readies the answer so it is retrieval rather than improvisation, and it surfaces any number whose confidence line you cannot write honestly, which is your signal to firm that number up before the meeting rather than discover its softness live.

Rehearsing the delivery matters as much as writing the content, because the forty-five-second, calm-then-stop delivery is a performance that needs reps. Say each Provenance Answer aloud, time it, and practise stopping at the end of the third line instead of trailing into justification. The goal is that when the attribution question comes — and on a data-heavy presentation to senior people, it reliably comes — you are not constructing an answer, you are delivering one you have already built and rehearsed. That difference is visible to the room. A rehearsed Provenance Answer looks like command because it is command: it is the outward form of having done the work and knowing your data’s strengths and limits cold.

The contrast with the 2019 divisional head is exact. He had the source — a reputable report his team had used for years — but he had not prepared to own it, so under the probe he reached for the team instead of the answer. Had he prepared the three lines, the same question would have produced: “It’s from the [sector] report we’ve used for three years, cross-checked against our bookings; I’m confident in the total, the segment split is directional, and I can firm that up from our own data by Friday if it’s decision-relevant.” Same number, same source, same truth — but ownership instead of deflection, and the room would have read command instead of doubt. The plan would have survived its own author.

One thing to do before your next data presentation

Before your next data presentation, take the three or four numbers your recommendation actually rests on, and for each one write the Provenance Answer on a single card: line one the source and how it was reconciled, line two which part you would stake a decision on and which part is directional, line three what you would do to firm up the soft part and by when. Then say each one aloud and time it — aim for under forty-five seconds — and practise stopping at the end of the third line. When someone leans forward and asks where the numbers came from, you will not reach for your team or your notes. You will deliver three calm lines that prove you own the data, and then you will stop and let the room come to you.

Frequently asked questions

What if I genuinely don’t know where a number came from in the moment?

Then you have found the real problem, and it is not a Q&A problem — it is a preparation gap. If a number is load-bearing enough to be on your slide, you should be able to give its Provenance Answer before you present, which is exactly why you write the three lines in advance. If the question still catches you on a number you cannot source, the least damaging response is an honest, owned one that does not deflect to others: “I want to give you a precise answer rather than a rough one — let me confirm the exact source and the confidence level and come straight back to you in the session if I can, or by end of day.” That owns the gap without disowning the number to a team, and it commits to a fast, specific follow-up. But the real fix is upstream: never carry a load-bearing number you cannot immediately source. Build the Provenance Answers for the headline figures and the gap closes before the meeting.

Doesn’t admitting a number is “only directional” give the room a reason to reject the whole plan?

It does the opposite, when it is done as calibration rather than apology. A senior audience knows that no real-world dataset is uniformly solid, so a presenter who claims everything is rock-solid reads as either naive or overselling, and that suspicion spreads to the numbers that genuinely are solid. Volunteering which parts are directional makes the room trust the parts you are confident about more, because you have shown you will flag the soft spots honestly. The key is framing it as a fact you command — “the segment split is directional” — not a weakness you fear. And you pair it with the caveat: the plan to firm it up. A directional number with a credible plan to make it decision-grade is rarely a deal-breaker; an undisclosed soft number that surfaces later, after the plan was approved on the assumption it was solid, is a far bigger problem for you.

How is this different from just citing my source properly on the slide?

A source line on the slide answers the literal question but not the real one. The attribution question is a trust probe, and the trust it is testing is not “does a source exist” but “does this person own their data.” A footnote citation proves a source exists; it does not prove you have thought about the number’s strengths and limits, which is what the confidence and caveat lines demonstrate. You should put the source on the slide — it pre-answers the routine version of the question and signals diligence — but when a senior person asks live, they are usually probing past the footnote to see whether you command the material. The Provenance Answer is what you say out loud in that moment; the slide citation is the quiet baseline underneath it. They work together: the citation handles the routine, the spoken Provenance Answer handles the probe.

What if the person asking is genuinely hostile, not just probing?

The Provenance Answer works for both, because calm command is the correct response to a hostile attribution question and a neutral one alike. A hostile questioner is hoping to provoke a defensive reaction that confirms you are rattled; the three-line answer denies them that by staying structured and unhurried. Deliver the source, the confidence, and the caveat exactly as you would for a friendly probe, and then stop — the stop is especially important with a hostile questioner, because trailing into justification is what they are trying to bait you into. If the hostility continues past a complete, calm answer, that is no longer about your data; it is a dynamic in the room, and the way you hold your composure becomes the thing the rest of the audience reads. A presenter who answers a hostile probe with three calm lines and a steady stop usually wins the room regardless of the questioner, because everyone else can see who is in command.

The Winning Edge — weekly newsletter

The Winning Edge is a weekly (Thursday) newsletter for senior professionals who present and field tough questions from senior rooms. One short email a week on the structural moves that keep you in command when the questioning starts. Subscribe to The Winning Edge →

For the full set of skills behind a high-stakes data presentation — slides, storytelling, confidence, and handling the questions — the Complete Presenter library of seven products brings them together for £99, with lifetime access.

About the author

Mary Beth Hazeldine is Owner & Managing Director of Winning Presentations Ltd. With 24 years of corporate banking experience at JPMorgan Chase, PwC, Royal Bank of Scotland, and Commerzbank — including years spent on the asking side of investment and credit committees — she helps senior professionals hold their composure and command the room when their data is challenged.