You Get What You Measure: The Four Metrics That Kill the Feature Factory

From Customer-Led Development: How to Build What Your Customers Want When AI Can Build Anything — the operating manual for building what customers want when AI can build anything. See every chapter.

If you measure features shipped, you'll ship features. Nobody will ask whether anyone uses them.

If you measure velocity, you'll get velocity. Nobody will ask whether you're running in the right direction.

If you measure story points, you'll get story points. Nobody will ask whether the stories matter.

You get what you measure. And most product organizations are measuring the wrong things.

The feature factory, by the numbers

This stat should make you a little sick. According to Pendo's 2019 Feature Adoption Report, 80% of features are rarely or never used.

Picture that on the ground. Your team builds five features. Each one gets the late nights, the sprint planning, the code reviews, the QA cycles, the docs, the release notes. Customers ignore four of them.

Pendo extrapolates that public cloud companies spent up to $29.5 billion building features nobody uses, and that a $50 million revenue company might waste $8.4 million a year the same way. The full Pendo report adds a sharper cut: only 12% of features generate 80% of product usage. That's Pareto, applied brutally to your roadmap.

This is the feature factory. Ship features, count features, celebrate features. Never ask whether anyone used them.

It's the incentives, not the people

Feature factories aren't staffed by lazy or stupid teams. They're staffed by smart people doing exactly what they're rewarded for.

• Measure PMs on features shipped and they ship features. "Shipped 47 features this year" goes on the review. Gold star.
• Measure PMs on velocity and the machine goes faster. Nobody asks where it's going.
• Measure engineers on output (lines committed, PRs merged, tickets closed) and the codebase grows whether customers need it or not.
• Measure sales on deals closed and they'll promise the moon. Product will figure it out. (Product won't.)

Rule of thumb: The metric defines the behavior. If you want customer outcomes, measure customer outcomes.

The four metrics that matter

For a customer-led organization, four metrics matter. Everything else is noise.

A warning first: most benchmarks are junk, and even good medians move with who gets surveyed. Median net revenue retention sits near 101% in the SaaS Capital 2025 B2B SaaS Retention Benchmarks for private companies, and at 110% among public companies, per Clouded Judgement. The numbers below are the ones I trust enough to print.

1. Retention: are customers staying?

The most important metric. If customers leave, nothing else matters.

SaaS Capital's survey of 1,000+ private B2B SaaS companies puts median revenue retention at 91%, with contracts over $250K a year keeping 95%. The top quartile across most size bands also keeps 95%, per High Alpha's 2025 SaaS Benchmarks Report. The Benchmarkit & Aleph 2026 SaaS & AI Performance Benchmarks are bleaker: median gross retention fell to 84%, from 88% a year earlier.

Every company I've seen keep 95% or more operates customer-led. The ones losing 10% or more a year have decided that's normal. No survey measures customer-led yet, so use the top quartile as your bar.

Experience tracks renewal closely. In "The Value of Customer Experience, Quantified", Peter Kriss found that at one $1 billion-plus subscription business, members who rated their experience worst had a 43% chance of renewing. Those who rated it best: 74%.

And churn compounds. Lose 5% of customers a month and you lose 46% of them in a year. You're refilling a leaky bucket, buying new customers to replace the ones quietly walking out.

2. Engagement: are they using it?

Customers can stay subscribed without using your product. That isn't retention. It's delayed churn.

Across 62 B2B companies in Userpilot's customer engagement benchmarks, average activation is 37.5% and core feature adoption is 24.5%. Read that second number again: three-quarters of users aren't using the thing you built. In my own work, the customer-led companies I've been inside clear 45-50% activation. Most sales-led products I've audited sit at 30-40% or worse.

Engagement predicts retention. People who use the product stay. People who don't, leave.

3. Expansion: are they buying more?

The clearest sign customers love you: they give you more money without being asked. At the median private SaaS company, expansion supplies about 40% of new ARR, and about half at companies above $50M ARR, according to Benchmarkit and the KeyBanc Capital Markets & Sapphire Ventures 2024 SaaS Survey.

It shows up in growth. In SaaS Capital's survey, companies with NRR above 130% grew at a median 50% a year. Companies below 90% grew at 15%. That's a 3.3x gap, driven largely by whether existing customers stay and expand.

4. Conversion: are prospects becoming customers?

Leads qualified by actual product usage convert at 8-15%, about 2.8x the rate of traditional leads with no product engagement, per Gainsight's PQL conversion benchmark. SlashExperts' PLG conversion guide puts customer-led free trials at 18-30%, against 2.8-3.5% for traditional leads. Higher conversion means lower CAC, which means more room to grow.

Why the four compound

These aren't four scoreboards. They're multipliers.

• Profit: In Bain's original study, "Zero Defections: Quality Comes to Services", cutting defections by 5 percentage points raised profits per customer by 25% to 85%.
• Firm value: In "Valuing Customers", a 1% improvement in retention raised firm value about 5%, versus about 0.1% for the same improvement in acquisition cost.
• Valuation: McKinsey's "SaaS and the Rule of 40" found public B2B SaaS companies with NRR of 120% or more traded at a median 21x revenue, versus 9x for the rest. "The Net Revenue Retention Advantage" found top-NRR companies held that premium through bull and bear markets.
• Revenue: Forrester compared five pairs of direct competitors from 2010 to 2015, and in every pair the better experience grew faster. Southwest grew 7.3% a year to United's 2.2%; Schwab grew 10.7% to E-Trade's 1.6%. McKinsey's "Experience-Led Growth" found the top half of each industry by NPS grew revenue more than twice as fast as the bottom half.

A meta-analysis of 96 studies spanning 26 years finds the link between satisfaction and financial performance is strongest where customers can switch easily. AI makes switching easier every year.

Then there's plain arithmetic. Take a fixed customer base, add no new customers, and compound for five years:

Same starting point. Radically different companies.

The vanity metrics graveyard

Rewrite the PM review

The old review reads: shipped 47 features, hit velocity targets, completed the roadmap. Exceeds expectations. That PM may have shipped nothing anyone used.

The new review reads: retention for owned features at 94% against a 92% target, weekly engagement at 67% against 60%, expansion from feature adoption at 18% against 15%, and requests addressed for 34 named customers. Same grade. Completely different job.

1. Assign ownership by customer outcome, not feature area. PMs own retention for a segment, not "the settings page."
2. Track adoption by requester. Did the customers who asked for it actually use it?
3. Measure time-to-value, not time-to-ship. When did customers get value, not when did code merge?
4. Put customer names in reviews. "This improved retention for Acme, TechCo, and BigBrand." Names, not numbers.

Would you rather ship 50 features with 20% adoption, or 10 features with 80% adoption?

The feature factory says 50. Customer-led says 10. Or come at it from the other direction: ship all 50 to the customers who asked, keep the 10 they use, and delete the other 40.

Your business model is your incentive

You can't be customer-led if your business model punishes customer focus. With seat-based pricing or annual contracts, the customer commits the money the day they sign, whether or not anyone logs in. Your pressure to keep serving them is real but distant. Usage-based pricing flips it: you get paid when customers succeed, every day.

But you don't need usage pricing. You need to run the business around the renewal. Costco shows how.

What Costco knows about renewals

Costco members prepay a fee once a year, like an annual software contract. In fiscal 2025, those fees brought in $5.3 billion, about half of Costco's $10.4 billion operating income, per its Form 10-K. Costco sees that money again only if members renew, so it runs every warehouse to earn the renewal.

The markup cap is how. As "How Costco Became the Anti-Wal-Mart" reported, no brand-name item could be marked up more than 14% and no Kirkland item more than 15%, when supermarkets generally marked up 25% and department stores 50% or more. Beating your margin target was a problem, not a win. A produce buyer told Fortune that if you did better than 10%, "Jim will say, 'Well, why didn't you lower prices?'"

The hot dog and soda combo has cost $1.50 since 1985, per CNBC. When Craig Jelinek told founder Jim Sinegal they were losing money on it, Jelinek recalled to 425 Business, the answer was blunt:

"If you raise the effing hot dog, I will kill you. Figure it out."

Costco built its own hot dog plants and kept the price. Analysts groaned. "At Costco, it's better to be an employee or a customer than a shareholder," one told BusinessWeek. Then from 2004 to 2013, Costco's shares returned 180% against 40% for the S&P 500, per The Motley Fool. In fiscal 2025 it renewed 92.3% of U.S. and Canadian members.

Give everyone skin in the game

Most companies already share revenue. They just share it with one department and call it commission. The median SaaS account executive earns about 11.5% of first-year contract value at quota, per The Bridge Group's 2024 SaaS AE Metrics & Compensation report.

Ship something that brings in $100 million and sales feels it. The engineers who built it, the designers who made it usable, and the support team who made customers successful don't. Clawbacks and retention bonuses are band-aids.

Your sales team was never over-incentivized. Everyone else is under-incentivized.

At my company, we take 10% of all revenue, renewals included, and distribute it across the whole team. We tip the chefs, not just the waitstaff. Engineering cares about retention because they benefit from it. People stop delivering whatever the scope said and start hopping on calls with customers to solve the real problem.

The OKR problem

This is what killed product-led for most companies. They claimed the philosophy, handed product the authority, then left every other department's goals untouched. Sales chased sales numbers. Success chased revenue. Marketing chased acquisition. Departments stopped competing with competitors and started competing with each other.

The four metrics fix this because they reinforce each other. Give each department a primary metric and make it accountable for its neighbors:

• Product: retention, plus engagement and conversion.
• Customer success: expansion, plus retention.
• Sales: conversion, plus retention.
• Marketing: engagement, plus conversion.

The fights between departments aren't personal. They're built into the goals, and they only end when the goals change.

The roadmap becomes a list of outcomes

Measure the four metrics and the roadmap changes shape. The feature factory's year is "Build A, B, C" four times over, with success meaning 12 features shipped. The customer-led year reads: improve retention for Segment X by 5%, get Feature A to 60% adoption, grow enterprise expansion 20%, lift trial conversion to 25%.

One is a list of things to build. The other is a list of things that have to be true for customers. You get whichever one you measure.

For the full system behind measuring customer outcomes instead of output, see Customer-Led Development.

If this changed how you'll read your next roadmap review, subscribe to the Context Limit newsletter for more ideas that fit in your head.