95 Claims From One Conference Stage, 19 Held Up as Stated: How Numbers Drift Before They Reach Your Deck
This autumn, a Canadian startup conference brought founders, investors and defence buyers together for one day of talks. I had 95 factual claims from five of its sessions checked against two independent sources each, primary source first. 19 held up exactly as stated. 54 were right in substance but needed a corrected number, unit, date or name. 22 were refuted in whole or in part, or no source could be found Sagentix Evidence Gate, 2026.
This is a story about how numbers travel. Most of the 76 claims that needed work started from a real figure and moved a short distance in the retelling — a word swapped, a year skipped, a company's own line repeated as market data. The same drift happens in every pitch deck, proposal and board paper that borrows a number from a talk, a podcast or a colleague's slide. This piece shows the patterns, with the corrected wording and the primary source for each, and ends with a check you can run in ten minutes.
How I counted. A claim counts as holding up only when two sources agree with it as spoken, with no word changed. A claim that was right in one part and wrong in another counts as needing correction. One row recorded during a break, not on stage, was left out. The tables behind the count are in the Sagentix library as five evidence tables, one per session.
95 claims from one conference day, by what the check found
- Held up exactly as stated19
- Right in substance, needed a fix54
- Refuted in whole or part, or no source22
Most Drift Is a Category Swap
The largest group of fixes changed no number at all. The number was right — the thing it counted had changed. Six swaps came up — on one stage, in one day.
Exports became production. Canada was said to rank 10th in world seaweed production. In the Food and Agriculture Organization (FAO) data, Canada ranked 10th in seaweed exports (US$18 million in 2019), while its harvest of 12,655 tonnes was 0.04% of world production — China produced 56.75% (Cai, 2021).
One drug maker became a drug class. The figure of US$36.5 billion in 2025 sales was given for the whole family of weight-loss drugs. It is Lilly's tirzepatide franchise (Mounjaro and Zepbound) alone — Ozempic, Wegovy and Mounjaro together made about US$55.6 billion (PharmaCompass, 2026).
Share of weight became share of people. Sixty to 70% of people who stop regain the weight sounds like a count of patients. The trial finding is different: a year after stopping semaglutide, participants regained two-thirds of the weight they had lost (Wilding et al., 2022). Two-thirds of the weight — not two-thirds of the people.
Deals became dollars. Chief executives aged 20 to 29 "represent only 3% of investment deals" in the Angel Capital Association's data (Angel Capital Association, 2026). On stage, it became 3% of angel money — a different measure. Deal counts and dollar totals move separately, and a deck that mixes them invites the first hard question in diligence.
Stock became purchases, and production became buying. In September 2026 the Prime Minister said the armed forces currently have about 2,000 drones — a stock, not a year of purchases. The national objective is producing two million drones a year, and the timeline for it is unclear (The Globe and Mail, 2026).
The right fact moved to the wrong name. Under oath before the French Senate on 10 June 2025, Anton Carniaux of Microsoft France said he could not guarantee that French citizens' data would never be passed to United States authorities (Sénat de la République française, 2025). The stage version moved the testimony to Microsoft Canada, which no record supports.
A swapped unit survives every spell-check and most reviews, because the number itself is correct. In the 95 claims, the unit, the scope or the name moved far more often than the figure. Read the noun next to the number before you read the number.
Numbers Age, and Retold Research Grows
The second pattern is time — figures age on a publication schedule. About 12% of Canadian businesses use artificial intelligence (AI) was true for the second quarter of 2025. A year later, Statistics Canada measured 19.2% of businesses using AI to produce goods or deliver services, up from 12.2% (Statistics Canada, 2026). The figure on stage was the one printed in the national AI strategy, which is a sound source; it had simply been overtaken by a newer release — one published before the conference.
Some figures grow each time they are told. Ukraine's forces were on track to receive about three million first-person-view (FPV) drones in 2025 (Kyiv Independent, 2025); on stage it was five million. The finding that repeat founders do better is real, but it is not four times better. In the working-paper version of the study, founders whose last venture succeeded had a 34% chance of succeeding in the next one, against 22% for first-timers (Gilbert, 2009); the published version reports 30% against 21% (Gompers et al., 2010). That is about 1.4 to 1.5 times — not four. A careful investor knows the paper, and four times is the number that ends the conversation.
A Company's Framing Is Not Market Data
A $100 billion agricultural plastics market was cited as a market size. The check traced the figure to one company's own website, where it frames the problem that company sells against; no market study at that size was found Sagentix Evidence Gate, 2026. The same applies to revenue, margins and customer counts that only the company states: they are the company's position, and they must be attributed as such.
The fix is short — one clause of attribution. Write "the company describes the market as $100 billion" or find the independent estimate. A press release or a home page must not become a market fact two slides later.
A Commitment Is Not a Law, and a Plan Is Not a Fund
Status drifts toward "done" — and it drifts in one direction. The European Union (EU) was said to be banning petroleum-based agricultural plastic. What is in force is the EU restriction on microplastics intentionally added to products, which began applying on 17 October 2023 (European Commission, n.d.); the check found no instrument banning conventional farm film Sagentix Evidence Gate, 2026.
Closer to home, on 29 October 2026 the Prime Minister announced the Government of Canada's "commitment to making an equity investment of up to $1.6 billion in Cohere" (Prime Minister of Canada, 2026b). A commitment two days old is not a completed stake — the same release calls it a planned investment. The $500 million Canadian Tech Growth Fund in the national AI strategy is still described as planned (BetaKit, 2026).
For a public buyer this is the pattern that matters most — it is the one an evaluator can check in a minute. A bill tabled is not a bill passed, a program announced is not a program open for applications, and a proposal that treats them as the same reads as careless to the person who will score it.
Four patterns cover most of the 76 fixes: a category swap, an outdated or inflated figure, a company's framing repeated as market data, and a status stated as done. Each one is caught by a question, not by more research.
Investors and Public Buyers Now Check
Mark Lawrence, executive chair of Angel Investors Ontario, said it on stage on the conference's investor panel: a founder who overstates in the first minute of a pitch is out (Lawrence, 2026). His closing advice was the other half of the same rule: know what goes in the due-diligence room, and have it ready before you talk to investors.
Public buyers check as well — and in writing. Canada's first drone contract awards in September 2026 followed an evaluation that included Foreign Ownership, Control and Influence (FOCI) assessments before any award (Prime Minister of Canada, 2026a). A buyer who checks the ownership of a supplier will also check the market size in its proposal.
Therefore the cost of a drifted number is not the number. It is the credibility of every other page in the same document.
Six Questions Before a Number Goes Into a Deck, a Proposal or a Board Paper
I run this check on every figure before it reaches a client — it takes about ten minutes for a slide's worth of numbers.
- Where is the sentence it came from? Open the original document and read the line, not a summary of it.
- What does it count? Exports or production, one product or a class, people or weight, deals or dollars, inventory or purchases.
- When was it measured, and is there a newer release? Statistics agencies publish on a schedule; check it.
- Who published it? If the only source is the company that benefits from it, attribute it to that company.
- What is its status today? For laws, funds and programs: proposed, tabled, in committee, committed, open, in force.
- Does a second, independent source agree? If not, attribute the figure to its single source, or cut it.
Run questions 2 and 5 on the three numbers in your current deck that carry the most weight. If any of them changes, you found it before an investor or an evaluator did, and the corrected figure is usually still strong enough to make your case.
Where Sagentix Fits
This check is the last of the ten phases in the Sagentix method, and the same two-source rule runs through every deliverable. The library behind it holds 1,464 cataloged intellectual property artifacts, including the five evidence tables behind this article. Phase 1 Market Intelligence is CA$4,500, delivered in 5–7 business days, with a money-back guarantee (subject to terms); a full engagement runs from CA$3,750 a phase to CA$45,000 for all ten phases (the same numbers in US$ outside Canada).
You have three ways to act on this. Run the six questions yourself on your next deck; it costs an afternoon. Give the job to one person on your team, with the rule that no figure ships without its source sentence and a second source. Or ask Sagentix to check the evidence base of a deck or proposal before it goes out.
Which of the six questions would have changed a number in the last deck you presented?
References
- Angel Capital Association. (2026, August 26). The role of founder age in early-stage investing, part 1.
- BetaKit. (2026, October 29). Canada to invest up to $1.6 billion in AI firm Cohere.
- Cai, J. (2021). Global status of seaweed production, trade and utilization [Presentation]. Food and Agriculture Organization of the United Nations.
- European Commission. (n.d.). Commission Regulation (EU) 2023/2055: Restriction of microplastics intentionally added to products. Retrieved November 1, 2026.
- Gilbert, S. (2009, February 2). The success of persistent entrepreneurs. Harvard Business School Working Knowledge.
- Gompers, P., Kovner, A., Lerner, J., & Scharfstein, D. (2010). Performance persistence in entrepreneurship. Journal of Financial Economics, 96(1), 18–32. https://doi.org/10.1016/j.jfineco.2009.11.001
- Kyiv Independent. (2025, December 24). Ukraine on track to receive total of 3 million FPV drones in 2025.
- Lawrence, M. (2026). Remarks on an investor panel [Panel remarks]. Canadian startup conference, autumn 2026.
- PharmaCompass. (2026). Top pharma companies and drugs in 2025: Lilly vaults eight spots to emerge at the top; GLP-1 drugs dominate list.
- Prime Minister of Canada. (2026a, September 10). Canada and Ukraine scale drone production to build Canada's defence [Backgrounder].
- Prime Minister of Canada. (2026b, October 29). Canada announces landmark investment in Cohere to scale secure, sovereign AI [News release].
- Sagentix Evidence Gate. (2026). Two-source fact-check of 95 claims made on stage at a Canadian startup conference, autumn 2026 [Five evidence tables]. Sagentix Advisors Inc.
- Sénat de la République française. (2025, June 10). Commission d'enquête sur la commande publique : audition de Microsoft France.
- Statistics Canada. (2026, May 27). Canadian Survey on Business Conditions, second quarter 2026. The Daily.
- The Globe and Mail. (2026, September 11). Ottawa defence drone initiative contracts.
- Wilding, J. P. H., Batterham, R. L., Davies, M., Van Gaal, L. F., Kandler, K., Konakli, K., Lingvay, I., McGowan, B. M., Oral, T. K., Rosenstock, J., Wadden, T. A., Wharton, S., Yokote, K., & Kushner, R. F. (2022). Weight regain and cardiometabolic effects after withdrawal of semaglutide: The STEP 1 trial extension. Diabetes, Obesity and Metabolism, 24(8), 1553–1564. https://pubmed.ncbi.nlm.nih.gov/35441470/
Adapted from Statistics Canada, Canadian Survey on Business Conditions, second quarter 2026, 27 May 2026. This does not constitute an endorsement by Statistics Canada of this product.
Subscribe + get the workbook
The Bottom-Up TAM / SAM / SOM Workbook — free with your subscription
An 11-page tactical workbook with fillable worksheets — NAICS lookup, three-filter SAM test, Bull/Base/Bear SOM, and the diligence cross-checks. Not published anywhere else. Then get evidence-backed analysis every other Tuesday. No spam. Unsubscribe anytime. See past issues.

Stéphane Raby, CISSP, CMC, P.Eng., MBA
Founder & Principal — Sagentix Advisors
CMC | CISSP | P.Eng. | uOttawa Telfer Executive MBA — ranked #1 globally by CEO Magazine, 2023. 25+ years in technology strategy, cybersecurity, and management consulting.
Want This Evidence Applied to Your Market?
Phase 1 Market Intelligence starts at CA$4,500 with a money-back guarantee.