Your Dashboard Measures Activity. Investors Will Ask About Causality.
A founder I work with opened a catch-up with the good news. “Our revenue has been growing steadily, month on month.”
I asked the only question that matters next. “That is great. What is driving it?”
A pause. “I am not sure. I cannot tell from the dashboard.”
That answer is honest and common. It also ends a diligence meeting early.
The dashboard had done exactly what dashboards do. It had counted. A dashboard is a ledger of activity: how much, how many, how often. Every question an investor asks in diligence is causal: what produced the growth, what happens when you add capital, what would have happened without the spend. A ledger cannot answer any of those. A causal answer needs a comparison with a world that did not happen, and no table holds one.
Five arguments follow. The ledger has no counterfactual. Investors decompose before they ask. Attribution does not close the gap. The turn-it-off test keeps winning. And contrast is expensive, so you have to choose.
1. A dashboard is a ledger, and a ledger has no counterfactual
Most founders believe a dashboard makes them data-driven. Count the tiles on yours: revenue, orders, sessions, sign-ups, spend, conversion rate. Every one is a count, or a ratio of counts.
A causal question has a different shape. “Did the campaign produce the lift?” means “would the lift have happened without the campaign?” Answering it requires a second series: the world where you did not run the campaign. That series is missing from your data because you ran the campaign. It exists only if you designed for it: a group that never saw the ad, a region where you held spend flat.
That is the structural fault. A ledger records one world. A causal answer needs two. No number of tiles turns one into two.
💡 Key Takeaway: Every number on your dashboard is a count. Every question in diligence is a comparison. The gap between them is the counterfactual, and it has to be designed in before the fact.
2. Investors take your growth line apart before they ask a single question
Tribe Capital published the method its partners built at Social Capital. They call it growth accounting. A revenue line splits into six movements each month: new, retained, expansion, resurrected, contraction and churned. Growth accounting, they write, “helps us understand how the business has operated to-date and guides the next set of questions we would ask.”
The decomposition comes before the questions. “Steady month-on-month growth” is the input, and the first act of diligence is to take it apart. Jonathan Hsu, who built the tools, describes what they exist to answer: “Is this startup working, and what will happen when we add more capital?” Both halves are causal, and the second decides the cheque.
Singapore makes this sharper. Venture funding here fell to USD 4.6 billion in 2025, and the reporting on that figure records that “investors are now placing far greater emphasis on due diligence and clear pathways to profitability before committing funds.” Unproven growth narratives, in its words, “have fallen out of favour.”
So the three questions are these. What produced the growth? What happens when doubles spend? What would have happened without it?
💡 Key Takeaway: Investors decompose your growth line before they read it. If you cannot say which of the six movements produced the growth, the meeting has already moved on.
3. Attribution on the dashboard is a model of the ledger, and the ledger is biased
The strongest objection: “our dashboard has attribution, so it tells us which channel drove which sale.”
Facebook’s own researchers tested that claim with two Kellogg economists. They took 15 large advertising experiments, 500 million user-experiment observations and 1.6 billion impressions. Then they compared the randomised results against what observational methods estimated from the same data. The observational methods “often fail to produce the same effects as the randomised experiments.” In half the studies, the estimated lift in purchases was “off by a factor of three”.
Same fault. An attribution model reads the ledger. The ledger records who saw an ad and who bought. It also records that likely buyers were the people most likely to be shown the ad, because the platform targeted them. The model cannot pull those two facts apart. Only a holdout can.
💡 Key Takeaway: Attribution is a story told about the ledger. A holdout is a second ledger. Diligence trusts the second.
4. The turn-it-off test keeps embarrassing the activity numbers
Two large companies ran the crude version in public. The results rhyme.
In 2017, JPMorgan Chase cut the number of sites carrying its programmatic ads from 400,000 to 5,000. Its chief marketing officer reported “little change in the cost of impressions or the visibility” of the ads. The activity number had fallen by 99%. The outcome had not moved.
The same year, Procter & Gamble cut more than USD 100 million of digital spend in a single quarter. Its chief financial officer told analysts there was “no negative impact on growth rate”. Organic sales grew 2% that quarter regardless.
Neither ran a clean experiment. Both ran a contrast: a before and an after, with one input changed. Every activity number had said the activity mattered. The contrast said it did not.
💡 Key Takeaway: A before-and-after with one input changed is the cheapest counterfactual there is. Run one before an investor asks why you never did.
5. Designed contrast is expensive, so choose the questions
I will concede the hard part: clean causal measurement is costly. Lewis and Rao studied 25 large advertising experiments with USD 2.8 million of spend behind them. They found the median confidence interval on return on investment was more than 100 percentage points wide. An informative experiment, they wrote, “can easily require more than 10 million person-weeks”. Most founders will never have that traffic.
So the job is to choose. Three moves cover most of a Series A conversation:
- Run growth accounting every month. It uses data you already have and costs nothing. It answers “what produced the growth”.
- Put one holdout behind each channel the round depends on. Five to ten percent of the audience, untouched, for a quarter. It answers “what would have happened without it”.
- Change one input on purpose and watch. Double one region’s spend, or pause one channel for four weeks. It answers “what happens when spend doubles”, crudely and honestly.
None of these lives on a dashboard. All three produce something a dashboard can then display.
💡 Key Takeaway: You cannot buy causality for every number. Pick the questions the round turns on, and build the contrast for those before the data room opens.
Final Thoughts: your dashboard says what happened, and the room will ask why
Steady growth is a fact about the past. Diligence is a question about the future: will more capital produce more of this? Answering it takes a comparison the dashboard was never designed to hold.
The founder I opened with had every right to be proud of the line. Where they were wrong was in what the line could defend. A cause defends itself. A cause has to be built before anyone asks.
If your growth is real, this is good news. The contrast will show it, and you will walk into the room with an answer instead of a pause.
If you want a second opinion on which three questions your next round turns on, book a discovery call or connect with me on LinkedIn.
A note before you close this tab. If your dashboard is full and your answers are thin, the problem is rarely the data. It is the absence of a designed comparison, and that is fixable in a quarter.
Mervyn Chua is a growth-transformation consultant helping founders and CEOs build the strategic clarity and systems to grow in an AI-first world. If this raises questions worth exploring for your brand, let’s talk.
