Your Dashboard Is Mostly Lying. Find the One Number That Isn't.
Most KPIs can be engineered without improving the business. The discipline that scales a company is identifying the one number that cannot.
I write for founders and executives navigating the inner game of scaling.
This issue was inspired by my recent conversation with Artin Bogdanov, serial founder, creator of Sun, the personalized AI audio app, and ex-Adobe designer. If you haven’t listened to the episode yet, it’s worth your time.
Which of the numbers on your dashboard could your team move next week without actually improving the product?
And which one could they not, no matter how hard they tried?
The Belief Worth Challenging
Many founders treat their dashboard as the company’s nervous system: when KPIs go up, things are working; when they go down, something is broken. The implicit assumption is that the numbers, taken together, give an accurate map of what is happening.
This belief has a problem. Most KPIs can be engineered. You can buy installations, incentivize referrals, pump activation with a tighter onboarding flow, prompt return sessions with a notification. None of it tells you whether the product is doing the thing the product is supposed to do.
In management accounting this has a name: surrogation. The metric becomes a stand-in for the strategy, people forget the difference over time, and what was meant to track the goal ends up replacing it.
The cost is not abstract. Wells Fargo measured cross-selling as a proxy for long-term customer relationships, and the cross-sales number went up. It went up because employees, under internal pressure on the metric, opened roughly 3.5 million accounts without customer consent. The metric did exactly what it was supposed to do; the strategy died.
The founder version is quieter, and quieter is more dangerous. The line moves as designed while the thing the line was supposed to represent quietly stops moving with it, and by the time the gap is visible from inside the business, the dashboard has been telling you everything is fine for two quarters. Artin Bogdanov ran an exercise designed to surface that gap before it gets that far.
What It Actually Looks Like
Artin is on his third company. His first sold high-precision night vision and thermal optics through a chain of e-commerce stores, ran for nearly five years, and ended in a lawsuit against Google over advertising restrictions. His second, Anolo, was a social-media automation tool that 20,000 e-commerce companies eventually signed up for, almost got acquired by Canva, and unraveled in early 2022 when Russia invaded Ukraine and his Moscow-based team became un-investable.
He says the first was driven by needing to apply his energy somewhere. The second was driven by money. The third, Sun, is the one he says he loves.
That admission matters because it explains an exercise he ran with his team at Sun. The exercise was simple: list every KPI the company currently tracks, then ask one question of each one.
“We ask ourselves truthfully, like which metric we cannot manipulate, like which metric we cannot engineer. And there was the only one metric and it was listen time.”
Everything else, Artin said, was solvable with budget or solvable with cleverness. Installation he could buy through performance marketing, activation he could engineer with onboarding, K-factor he could juice with referral incentives, return sessions he could prompt with notifications. Listen time was the only number where the user’s actual behavior had to align with the product’s actual quality.
“No matter what they do, if our content sucks, you’re not going to listen to it.”
What makes the exercise useful is the question itself. Most founders never sit down and ask which of their own numbers they could lie to themselves with, because the dashboard arrives pre-built with whatever the analytics package recommends, and the founder optimizes against it.
Artin made the question explicit, then made the answer the team’s North Star. He also pointed at something else: even the truth metric is not enough on its own. Listen time without new installs eventually starves, and you still need the top of the funnel to fill. But the order matters: hit the truth threshold first, then scale the funnel.
“In our case, we want to be equivalent for the social media apps, like 25 minutes a session, then you can increase the top funnel and scale it.”
The principle underneath is not new. Researchers in management accounting have documented the same pattern for over a decade under a different name. What Artin did at Sun was build a meeting around it before he had to recover from one.
If this resonates, forward it to a founder whose dashboard looks healthy and whose product is quietly drifting.
What the Research Actually Shows
The metric replaces the strategy when you tie pay to a single number.
Jongwoon Choi, Gary Hecht, and William Tayler ran two laboratory experiments with MBA-level participants assigned to manage a hypothetical company toward a stated strategic objective. The variable they manipulated was the compensation structure: participants were paid either on a single performance measure that served as a proxy for the strategy, or on multiple measures of the same underlying construct. Across both experiments, when compensation depended on a single proxy, participants made decisions as if the measure itself were the strategic objective, even when the measure and the strategy clearly diverged. The authors named this pattern “surrogation.”¹
The authors frame surrogation as a structural feature of how single-metric incentive systems get used. It shows up predictably across participants, independent of the integrity or attentiveness of the individuals inside the system. The cleanest mitigations they recommend are tying compensation to multiple measures of the construct and involving the people executing the strategy in defining it.
For a founder, this puts Artin’s exercise upstream of the failure mode the research describes. Before you tie any incentive (pay, OKR, board scorecard) to a number, you need to know whether the team will optimize the number toward the underlying construct or away from it. “Which number can we not engineer” is the pre-screen.
The cost of skipping the question shows up at scale, in public.
In the September-October 2019 issue of Harvard Business Review, Michael Harris and Bill Tayler applied the surrogation framework to Wells Fargo’s cross-selling collapse as a real-world case. The bank’s strategic objective, deepening long-term customer relationships, was operationalized through internal pressure on a single cross-sales metric. Between 2002 and 2016, employees opened approximately 3.5 million unauthorized customer accounts to hit cross-selling targets.²
Harris and Tayler conclude that Wells Fargo is the laboratory pattern at industrial scale: the same surrogation mechanism observed under controlled conditions, operating across a Fortune 50 bank’s incentive system. The cross-sales number moved exactly as designed, while the strategic objective the number was supposed to represent went the opposite direction. Their recommended mitigations match the underlying research: involve the people executing the strategy in defining it, avoid tying compensation tightly to a single proxy, and build review mechanisms that surface the divergence between metric and construct before it becomes a scandal.
The thesis of this issue follows directly. The metric that survives Artin’s “which number can we not engineer” test is, by definition, one where moving the number requires the strategic outcome to also move. Wells Fargo is what happens when no one in the room ever runs the test.
How to Find Your One Honest Number
Here are four moves to find your honest number and protect it from being optimized away.
Hold an engineering audit of your own dashboard.
Take an hour with your team and list every KPI you currently track. Next to each, write the cheapest way the team could move that number this quarter without improving the underlying business: a paid channel, an incentive, a notification, a re-segmentation. The numbers that come back with no good answer are what Eric Ries calls actionable metrics. The rest are vanity metrics, and they are subject to surrogation the moment you tie a goal to them.
Set the truth threshold before you set the growth target.
Most founders pick growth targets first and let the team optimize backward. Reverse the order. Define the threshold on the metric that cannot be faked (in Artin’s case, twenty-five minutes a session), and hold the team accountable to that threshold across the existing user base before anyone spends a dollar opening the top of the funnel. Scaling a leaky truth metric is how unit economics quietly break.
Compensate against process and multiple measures, not against a single proxy.
A 2022 rapid review of performance-management systems in healthcare identified the same pattern Artin caught in his meeting: when incentives ride on a single proxy metric, the most common unintended consequences are measure fixation, tunnel vision, and outright gaming.³ If a bonus depends on a number that can be gamed, the number will be gamed. Until the truth metric is reliable, tie compensation to multiple measures or to process discipline (interviews completed, root-cause documents written, weekly product reviews held). Once the truth metric is stable, you can lean on it harder.
Translate each top-line metric back into plain English every quarter.
Surrogation creeps in over time. Once a quarter, write one paragraph in plain English explaining what each top-line metric is supposed to be a proxy for, not just what it numerically measures. The discipline is close to what Eric Ries calls innovation accounting: regularly translating numbers back to the learning hypothesis they were originally hired to test. If the paragraph is hard to write, the metric has drifted from the goal it was originally hired to track, and the team has probably drifted with it.
The Bigger Picture
A founder will spend most of their attention on the dashboard. That is correct. The dashboard is the closest thing a company has to a contract with reality, and reality moves faster than memory.
What gets missed is the second-order question, the one Artin made explicit in a single meeting: what is the number a stand-in for, and could the number be lying to me about that thing?
The principle of pursuing truth in a metric, not just hitting it, is rarer than it sounds. It costs more to install, and it demands giving up the numbers that look good for the ones that say something. Most leadership teams pick the comfortable proxy and move on.
“We actually have a principle, pursuit of truth.”
The sharpest founders I work with do the opposite. They build the dashboard, then ask in writing what they would fall for if the dashboard were trying to fool them, and shrink the list to the line that cannot. That single line, defended, is what scaling actually looks like from the inside.
Founder Circle
I work with a small group of founders who want three things:
Clarity on what actually moves the needle past busyness.
A space to think through the hard decisions (co-founder dynamics, team restructuring, fundraising trade-offs) without having to perform with confidence they don’t yet feel.
A peer group that gets it: other founders who are doing the inner work alongside the operational work.
Our next circle meets soon. Grab your spot today.
P.S. What is the one number in your business that you genuinely could not fake your way into next quarter, no matter the budget? Reply and tell me. I read every response.
About the author
I’m Dar Patel, an ICF-certified executive coach (PCC) and founder of Little Pursuits. I partner with founders and executives through the leadership inflection points: the identity shifts, the hard conversations, the decisions you keep carrying alone. This newsletter is where that work meets the page.
If you want to go deeper, check out the references used in our research:
¹ Choi, J., Hecht, G. W., & Tayler, W. B. (2012). Lost in Translation: The Effects of Incentive Compensation on Strategy Surrogation. The Accounting Review, 87(4), 1135–1163. Two experiments demonstrated that compensating managers on a single performance measure tied to a strategic objective causes them to substitute the measure for the objective itself, a phenomenon the authors named surrogation. The effect was muted by using multiple measures.
² Harris, M., & Tayler, B. (2019). Don’t Let Metrics Undermine Your Business. Harvard Business Review, September-October 2019. Documented Wells Fargo’s cross-selling crisis as a real-world case of surrogation, in which approximately 3.5 million unauthorized accounts were opened in response to internal pressure on the cross-selling metric, destroying the long-term customer relationship strategy the metric was originally designed to represent.
³ Li, X., & Evans, J.M. (2022). Incentivizing performance in health care: a rapid review, typology and qualitative study of unintended consequences. BMC Health Services Research, 22, Article 690. A rapid review of 41 papers plus a qualitative study in Ontario cancer and renal care, identifying measure fixation, tunnel vision, and gaming as the most common unintended consequences of single-metric performance-management systems.
Read next