Who moves first, and what it costs you
Every company runs one of two sequences. Either you sell first and build to the promise, or you build first and go find the buyer. Which sequence dominates is set by the team you're looking at. What that sequence costs is set by the kind of company you are.
Product × Engineering
The Asset
You build the thing once and sell copies of it. The buyer can arrive years later.
- What's really happening
- Engineering is manufacturing an asset. Cost is paid upfront and spread across every future customer, so the marginal cost of customer number 10,000 is close to zero. This is the only quadrant where you can genuinely build something before the demand exists and survive the wait.
- What the team optimizes
- Throughput, architecture that holds at 10× load, defect rate, time-to-ship. The roadmap is a portfolio of bets on a future that hasn't happened.
- Where it breaks
- Build-first with no market thesis. Engineering starts mistaking hard for valuable and ships something beautiful that nobody asked for. The amortization only works if someone eventually shows up.
- Looks like
- Nvidia shipping CUDA a decade before generative AI created the demand · AWS, which was internal plumbing before it was a market · Apple designing its own silicon · 3M's Post-it, a failed adhesive looking for a job
Product × Sales & Marketing
The Funnel
You're selling a fixed artifact you can't bend for one buyer. So the job is reach, not tailoring.
- What's really happening
- Marketing can move first here because demand can be measured before a line of code is written. Research the segment, write the spec, hand it to engineering. And distribution scales without headcount - one campaign reaches a million people, which is impossible in the row below.
- What the team optimizes
- Positioning against a defined segment, CAC and payback period, win rate, share of category. Losing deals are treated as data: lost-deal analysis becomes roadmap input.
- Where it breaks
- Sell-first curdles into promising features to close one deal. Every bespoke commitment on a product roadmap is a tax paid by every other customer, and it accumulates as debt nobody can see until velocity dies.
- Looks like
- Procter & Gamble running consumer research → product for a century · Salesforce's sales-led enterprise motion · HubSpot inventing the demand it then served · Coca-Cola, where the product barely changes and the whole game is meaning
Service × Engineering
The Bench
What you sell is skill and hours. Capability is inventory - and it spoils while it waits.
- What's really happening
- You're building capability ahead of the contract that pays for it. Nothing amortizes here - every new client consumes new hours, so cost scales linearly with revenue. An expert with no project on Monday still costs full salary on Monday.
- What the team optimizes
- Utilization, bill rate, margin per head, ramp time for a new hire. Reusable accelerators and IP matter enormously because they're the only lever that breaks the linearity.
- Where it breaks
- This is the most expensive quadrant to be wrong in. You capitalize a skill and the market moves; the bench burns cash the whole time. Capability built without a named buyer isn't a practice, it's a résumé.
- Looks like
- Thoughtworks building a genuine XP/agile practice, then charging for it · Accenture and Infosys standing up cloud and AI practices before client demand arrived · specialist boutiques that get world-class at one narrow thing and price accordingly
Service × Sales & Marketing
The Promise
You sell an outcome that doesn't exist yet, on credentials and trust. Then you make it true.
- What's really happening
- Sales moves first, almost always - and this is the default operating mode of the entire services industry. It's capital-efficient by design: you never pay for a bench you haven't already sold. What's actually being sold is trust, references and case studies, because there's no artifact to demo.
- What the team optimizes
- Pipeline, win rate, deal size - and sold margin, which is the quiet one. The price is committed before the true cost is known, so scoping discipline is the whole business.
- Where it breaks
- The gap between what was sold and what can be delivered. Overpromise once and you get the classic death-march project: margin erosion, burnt-out teams, and a marquee client you can't use as a reference.
- Looks like
- TCS, Infosys and Cognizant winning large deals, then staffing and training into them · Big Four consultancies competing on RFPs · ad agencies pitching the campaign, then hiring for it · law firms selling the partner and delivering with associates
Two sequences, and the price of each
The four quadrants aren't four independent stories. They're two sequences, each running under two very different cost structures.
Sell first - market pull
Demand is proven before cost is committed. Somebody has said out loud that they want this, and often signed for it. Low risk, fast feedback, strong capital efficiency.
The ceiling is the problem: you can only build what a customer is already able to articulate. Nobody asked for the smartphone. Run this exclusively and you end up excellent at serving a market someone else defined.
Failure mode → promising what you can't deliver
Build first - capability push
You create the capability before the demand exists. Higher variance, and the only route to real differentiation - if you can only build what customers already ask for, so can every competitor.
You're funding the gap between capability and revenue out of your own pocket. The question isn't whether that gap exists, it's how long you can hold it open.
Failure mode → building what nobody wanted
Build-first is cheap in a product company and brutally expensive in a service company. In the top row the gap is an investment - you build once, and every customer who eventually arrives helps pay for it. In the bottom row the same gap is pure burn - idle capacity, nothing to amortize against, the meter running every day.
That single difference explains why product companies can afford research labs and most service companies can't. It's also why services that do invest in capability ahead of demand tend to convert it into a product, a platform or reusable IP as fast as they can - moving themselves up a row.
The same six questions, four different answers
| The Asset Product × Eng | The Funnel Product × S&M | The Bench Service × Eng | The Promise Service × S&M | |
|---|---|---|---|---|
| What you sell | A repeatable artifact | Access to that artifact, and its meaning | Proven capability, priced per hour | An outcome that doesn't exist yet |
| Moves first | Build - capability can precede the buyer | Sell - demand is measured, then specced | Build - the practice precedes the pipeline | Sell - sign, then staff, then train |
| Unit economics | Upfront cost, near-zero marginal cost | Reach scales without headcount | Cost scales linearly with revenue | Price fixed before cost is known |
| Core metric | Velocity, reliability, scalability | CAC, payback, win rate, share | Utilization, bill rate, margin per head | Pipeline, win rate, sold margin |
| Cost of being wrong | Sunk R&D - recoverable if the market arrives late | Roadmap debt that every customer pays | Idle bench - unrecoverable, burns daily | Death-march delivery and a dead reference |
| Scaling lever | Leverage - one build, many customers | Distribution and brand | Reusable IP and accelerators | Repeatable playbooks and pricing discipline |
Three things the grid hides
01 - Most companies are hybrids
Product companies run professional-services arms. Service companies spin out platforms. These are archetypes, not boxes. The useful question is which logic a given business unit is actually running, not which cell the company's logo belongs in.
02 - This is where cross-functional fights come from
Engineering thinks sales overpromises. Sales thinks engineering builds what nobody asked for. Both are correct, because they're optimizing different sequences. Naming the sequence usually defuses the argument faster than arguing the facts.
03 - Healthy companies oscillate
Pure sell-first drifts into commodity work. Pure build-first drifts into expensive hobbies. The real skill is knowing when to switch - and the switch usually comes from a capability you can no longer sell by the hour.