The Business Is an Equation
Model the whole business as one equation before optimizing any funnel leaf node.
- Difficulty
- Advanced
- Time to result
- ~months to results
- Steps
- 5
- Confidence
- 95%
Tom Conrad's core lesson from Quibi's $2B collapse and Zero's profitable growth: a company is not only the product you build, it is a mathematical formula that converts investment into returns over a time horizon. Product craft only moves variables inside that equation. If the equation itself is broken, no amount of iteration, funnel optimization, or beautiful design can rescue the output. The practice is to build an explicit model of the entire business first, find the high-leverage variables, and only then let product teams optimize.
Origin
Developed by Tom Conrad across a 30-year career: he entered it as a pure product/engineering craftsman (Apple, Pandora, Snap), watched Quibi's foundational math fail despite excellent execution, and only fully internalized it at age 53 as CEO of Zero, where he took on the FP&A modeling himself. He credits investors and board members — whose 'probing questions' at Pandora board meetings never touched the product — as having been operating at this layer all along.
Core principles
- 01A company is an equation that turns investment into returns on a time horizon; the product is a set of variables inside it, not the equation itself.
- 02Execution excellence cannot repair a broken foundational equation — Quibi made 70 shows in 18 months and still failed.
- 03The cost structure determines how long you are allowed to iterate: a company burning a billion a year on content cannot afford two years of funnel grinding.
- 04Start at the top of the equation (LTV, CAC, cost structure), not at the leaf nodes (install-to-register conversion).
- 05Model first, then find which dimensions have the most leverage, then task the product team.
How to run it
- 1
Write down the equation before the roadmap
Before evaluating a product plan, state explicitly how investment turns into returns: what you spend, on what, to acquire whom, at what lifetime value, over what horizon. If you cannot write it, you do not yet understand the company you are working in.
Pro tip Ask what the model REQUIRES to be true — Quibi's model required landing in the top 10 most downloaded apps from day one and staying there.
Watch out A model that only works under a heroic distribution assumption is not a model, it is a bet.
- 2
Stress-test the required assumptions and name their improbability out loud
Separate 'structurally impossible' from 'highly improbable'. Impossible assumptions get killed easily; improbable ones survive because nobody can prove them wrong. Quantify the improbable ones against comparables (Uber has unlimited paid-acquisition budget and is not consistently top-10) and present the data to leadership before launch, not after.
Pro tip Conrad's own self-critique: 'maybe I should have beat the drum a little harder about just how unlikely it was that we were going to land the kind of distribution in month one the model sort of required.'
Watch out The dangerous assumptions are the ones that are merely improbable, not impossible — those are the ones a room full of optimists will approve.
- 3
Build one model that aggregates the whole business
Construct a single spreadsheet/model that pulls BI data, raw subscription transactions, historical results, and all expenses, and lets you manipulate variables to predict future scenarios. At Zero, Conrad — the CEO — is the full-time FP&A guy who built exactly this.
Pro tip The point of the model is not forecasting accuracy; it is identifying which variable has the most leverage.
- 4
Only then optimize the leaf nodes
Once you know which dimension carries the leverage, hand product teams the specific funnel steps (install to registered, registered to trial, trial to paid, retention) that actually move the top-line equation. Do not start here.
Pro tip If a leaf-node optimization cannot be traced back to a high-leverage variable in the model, it is busywork.
Watch out Conrad's natural instinct was to jump straight to leaf nodes — this is the default failure mode of product-first leaders.
- 5
Re-run the equation the moment reality diverges
When launch metrics come in (downloads, retention), do not automatically assume you get the startup's standard grace period of iterate-and-grind. Recompute the equation with real numbers and ask whether the cost structure grants you the time to iterate at all.
Pro tip Quibi retained at 'a good industry starting point' — the retention was normal; the equation was not.
Watch out 'It wasn't going to take two billion it was going to take six or eight or 10 billion' — the recomputed number is the decision, not the vibes.
In the wild
Quibi raised over $2B, produced 70 bespoke shows in 18 months (more than all major broadcast networks combined make in a year), recruited the world's most famous celebrities, and launched a beautifully engineered mobile-native product. But the foundational equation required a bespoke library large enough to drive subscription and retention for ~$2B, plus a top-10 app-store landing from month one sustained indefinitely, plus ~a third to half of content spend going to daily shows produced day-of in purpose-built studios. COVID killed the daily studio content two weeks after launch, and the recomputed cost was $6-10B, not $2B.
→ Quibi shut down less than a year after launch. Conrad's conclusion: the answer to the foundational question was no from the beginning, and the risk/reward of betting $10B on the format was 'more than anyone can stomach'.
At Zero (longevity/fasting app), Conrad stopped starting at conversion percentages and built an ever more sophisticated model aggregating BI tools, raw subscription transactions, historical results, and expenses to predict scenarios and locate leverage. The company then focused two full years on unit economics — LTV, CAC, funnel efficiency, retention — driven by what the model said mattered.
→ 27 employees, over a million monthly users, 100,000+ paying subscribers, tens of millions in revenue, double-digit growth coming out of the pandemic while peers retreated, and zero paid user acquisition.
Common mistakes
Believing product craft alone will save you
Conrad's lifelong belief was: find an interesting problem, solve it 10x more elegantly than anyone else, listen and iterate, and word of mouth handles the rest. That works only when the equation permits it. When the equation is broken, elegance is irrelevant.
Starting at the leaf nodes
Jumping straight to 'what percentage convert from install to registered' feels productive and measurable, but without the whole-business model you cannot know whether that funnel step has any leverage at all.
Assuming you'll get time to grind on the funnel
Product leaders assume the standard startup pattern: launch, retain badly, then spend six months to two years optimizing your way to fit. A high-fixed-cost business does not grant that time — 'in a world where you have to spend you know a billion dollars a year making content you just can't afford to not be a hit.'
Is it for you?
Best for
Product and engineering leaders (CPO, VP Product, first-time CEO) at venture-funded companies who came up through craft rather than finance, and who sit in board meetings baffled by why investors never ask about the roadmap.
Not ideal for
Very early pre-product-market-fit exploration where no meaningful data exists to model yet, or hobby/mission projects where returns are not the objective function.
From the transcript
“companies are also kind of like a they're kind of a math problem that describes how you take you know investment and pour them into…”
“if the equation is fundamentally broken or a big swing in and of itself no amount of like iteration and execution can like kind of…”
“my natural instinct was to go immediately to sort of the leaf nodes of the equation”
“it is really really powerful to spend the time to create the model that describes the whole thing so you can identify like what are…”
From the episode
Billion dollar failures, and billion dollar success
Tom Conrad (Quibi, Pandora, Pets.com, Snap, Zero)