The Science of Scaling: Why Retention Starts in Sales, Not Customer Success

🎙️ Listen to the audio version of the podcast.

Most founders think retention is a product problem.

Fix the onboarding. Hire a customer success manager. Build a better dashboard. Reduce friction in the UX.

Wrong department.

Mark Roberge was the fourth employee at HubSpot. He took them from zero revenue to IPO as their founding Chief Revenue Officer.

Then he spent a decade at Harvard Business School and as co-founder of Stage 2 Capital, studying hundreds of startups. IPOs. Billion-dollar exits. And a graveyard of companies that scaled themselves into the dirt.

The pattern he found is uncomfortable: retention starts with sales. Not after the sale. Not during onboarding. At the point of sale. Who you chose to sell to and the expectations you set on the way in determine whether that customer stays or bleeds out twelve months later.

His new book, The Science of Scaling, puts data behind the decision most founders still make with their gut: when to scale, and how fast.

100% of the proceeds go to mental health through McLean Hospital. That matters.

...and it tells you something about the man.

About Mark Roberge

Mark Roberge is the Co-Founder and Managing Partner of Stage 2 Capital, a venture firm backed by 1,000+ sales and marketing leaders from companies including OpenAI, Anthropic, Zoom, Databricks, Salesforce, and Workday.

He's a Senior Lecturer at Harvard Business School and the Founding Chief Revenue Officer at HubSpot, where he was the 4th employee and led the company from $0 to IPO.

His new book, The Science of Scaling: Using Data to Decide When — and How Fast — to Scale Revenue, is available now. 100% of proceeds are donated to McLean Hospital for mental health research.

📕 Get Mark's book — The Science of Scaling: https://amzn.to/4uS6Z0s
🔗 Mark on LinkedIn: https://www.linkedin.com/in/markroberge/
🔗 Stage 2 Capital: https://www.stage2.capital

The Mythology of "Scale at All Costs"

The internet sells a version of entrepreneurship that looks like a Red Bull commercial. Scale fast. Break things. Hire ten reps on January 1st. Burn cash until you hit orbit.

Mark's seen the other side.

Go back a few decades and entrepreneurship was unglamorous. Business school kids thought startup founders were idiots. Mark was at MIT when the dot-com bubble popped. There were 500 students in his class. Three of them were doing startups. Everyone else called them morons.

"B to B and B to C meant back to banking and back to consulting."

The pendulum swung. Venture capital exploded. Blitzscaling became gospel. And suddenly every founder with a million in ARR was handed eight million dollars and told to go hire.

The problem is context. The companies that made blitzscaling famous (Databricks, OpenAI, early Google) had product-market fit so absurd they could have hired chimpanzees and hit quota. That's Mark's line. Not mine. But I'm stealing it because it's true.

Most startups don't have that. They have decent traction, a pile of VC money, and a board that's seen exactly one moonshot work. So they copy the playbook from the 1% and wonder why they're bleeding cash in month eight.

Half the founders Mark meets are going too fast. Half are going too slow. The answer isn't speed or caution. It's measurement.

WATCH NEXT: Why Your Passion is Destroying Your Life (And How to Fix It)

The Leading Indicator of Retention

This is where the conversation shifted for me.

I spent 20 years in the property casualty insurance industry. Retention is the entire business model. The reason insurance agencies can operate on 1992 technology and still print money is because once a customer is on the books, they tend to stay. A C-player in insurance can be a millionaire. The model is that good.

But I had never thought about a leading indicator of retention.

Mark's framework: What can you observe in the first 30 days of a customer's engagement that predicts whether they'll stay forever or churn?

At Slack, it was 2,000 team messages in the first month. Hit that number and the customer sticks. Miss it and they're gone.

At HubSpot, it was using five or more features on their 25-feature platform. Five features and you're locked in. Four or fewer and you're a ghost by renewal time.

In insurance, I can map the same principle. Industry data is dead simple: one policy, 36% retention. Two policies, 72%. Three policies, 93%. The leading indicator of retention in insurance is policy count. Three policies and you've got them for life.

The LIR concept does something powerful. It connects the sales conversation directly to long-term business health. Once you know the number, you can build compensation, coaching, process, and accountability around making sure every new customer hits that threshold.

Retention Starts with Sales (The Contrarian Truth)

Mark told a story about HubSpot's Series D investment from Sequoia. The partner pushed back: "Sales' job is to drive revenue. Customer success and product own retention."

Mark's response: that statement is incorrect in most contexts.

When he goes into companies that are flatlined on retention, the root cause isn't product. It isn't the onboarding team. It isn't the customer success manager.

It's sales. It's who the rep chose to sell to. It's the expectations they set (or didn't set) on the way in. It's whether they got IT involved pre-sale when setup required IT. It's whether they sold a lifetime value customer or just got a wire and a contract to hit quota.

"I can get a contract and a wire without talking to IT. You want me to sell? Fine. I'll sell ice to Eskimos. But that's not a good business."

The objective of sales is not to generate a signature. It's to generate a lifetime value customer. Every compensation plan, every quota, every territory decision should trace back to that truth.

The Gated Commission Plan

This is the framework I want insurance agency owners and SaaS founders to steal today.

The standard comp structure in insurance (and a lot of industries) creates a predictable lifecycle: grind for five years, build a book of business, then coast. Feet up. Umbrella in the drink. Still pulling $250K on renewals while doing the absolute minimum to maintain accounts.

Mark's not against earning that rest. He's against wasting a hunter's skill set.

The gated commission plan works like this:

— If your new sales in 2026 are under $200K, your renewal commission drops to 15%. — If your new sales hit $200K-$500K, your renewal commission stays at 20%. — If your new sales hit $500K-$750K, your renewal commission bumps to 25%. — Over $750K? You're at 30% on renewals.

The floor drops if you coast. The ceiling rises if you hunt. It self-selects. Grinders love it. Coasters leave. And the ones who leave? Let them coast at the competition.

The concern is obvious: "Won't this scare off talent?" Mark's counter: word gets around. The top performers, the ones addicted to quota, the ones who want to wake up every morning at 22 and still be hunting at 65... they're coming to you. Because you pay hunters like hunters.

The Leveled Promotion Path (How HubSpot Held 6.5-Year Tenure)

Tech sales has an average tenure of 2.2 years. Mark pulled off 6.5 at HubSpot. Here's why.

Top reps kept coming to him saying they wanted to be managers. He'd ask why. "It's the only way to grow in sales." Every study shows top reps make the worst managers. They hate it. They call it adult daycare. They lose personal independence and make less money.

So Mark built an alternative: a leveled career path inside sales.

Level 1: $50K base, $50K commission, $800K quota. You prove yourself.

Level 2: Once you hit an $800K install base and your leading indicator of retention metrics are solid, you get a 10K OTE bump and 1,000 stock options.

Level 3, Level 4, Level 5: each tier raises the bar on install base, trailing average new sales, and LIR performance. Each tier bumps compensation and equity.

The result is a game. Reps aren't grinding toward a vague "someday manager" promotion. They're chasing Level 5. It's specific. It's measurable. And it keeps your best closers closing instead of sitting in HR meetings learning how to write performance reviews.

Translate this to insurance, to services, to any business with a recurring revenue model. The frame transfers directly.

Coachability: The #1 Hiring Attribute

After hiring 200 salespeople at HubSpot, Mark quantified every interview assessment. Scored candidates 1-10 on eight attributes. Correlated those scores to actual performance over two years.

Coachability was number one.

Not hustle. Not grit. Not "closing ability." Coachability.

Some attributes, if you hire wrong, take a psychology degree to fix. Call reluctance is wired into childhood. You can't train that out in a sales meeting. But product knowledge? Territory expertise? Those are teachable.

Coachability determines whether someone can learn everything else.

Mark's interview technique is elegant:

1. Send the candidate your training manual and a prospect's LinkedIn profile before the interview.

2. Run a role play.

3. After the role play, ask: "How do you think you did?"

Low coachability candidates think they crushed it. High coachability candidates dissect their own performance with precision. The ability to self-assess is the tell.

4. Give one piece of positive feedback and one piece of constructive feedback.

5. Watch how they receive it.

6. Run a second role play in the next round and see if the coaching landed.

If someone sneaks through and turns out to be uncoachable? Build a coaching culture so visible that they feel it.

First day of every month is coaching setup day. Every rep gets a diagnosis, a plan, and a measurement cadence. Most of the time, the uncoachable rep watches everyone else improving and comes back with their tail between their legs.

And if they don't? The performance plan catches it. "Welcome to the company. Here's how you get fired." No ambiguity. Two quarters below 80% of quota, you're on a plan.

Next quarter below 90%, you're gone. No hard feelings. Set it day one. The coasters self-select out before they ever accept the offer.

The Post-Acquisition Lesson I Learned the Hard Way

I built Rogue Risk into the fastest-growing small commercial insurance agency outside the top 200 in the entire industry. Wholly inbound. 35+ leads a day. A one-call close process that hit 63% first-touch close rate on qualified leads. Near-zero CAC because the engine was content marketing: YouTube and SEO.

I sold it. And I sold it to the wrong people.

The playbook looked perfect on paper. Roll equity into a PE-backed company. Inject capital. Go to the moon.

Three months in, friction. Then more friction. Then 17-person board meetings where I was told my CRM "wasn't professional enough." An $8,000 enterprise IT line item appeared on my budget for the privilege of calling tech support about my Mac. The team that bought a lean, mean selling machine wanted me to migrate to Salesforce at $275 per user seat.

I'm ADHD. Irish Catholic. I curse like a sailor. And I'm sitting in a boardroom with people I don't think have the brain capacity to keep up with what I was doing, being told to slow down and professionalize.

The mistake wasn't selling. The mistake was not asking one question: What do you think my day-to-day looks like after I sign this paper?

Not what I think it looks like. What they think it looks like.

I help other founders with acquisitions and raises now. That question is number one on the list. Not the valuation. Not the exit triggers. Not the equity structure. What does the buyer think your daily operating reality becomes post-close?

Mark's abstract takeaway is clean: during any investment or acquisition, 90% of the work goes into making the transaction happen. It should be 50/50. Half on the deal. Half on post-transaction life.

Most founders go through this one, two, maybe three times. You don't get ten reps to figure it out. Ask the question before you sign.

Athletes vs. Specialists

Mark dropped a frame I hadn't heard before: athletes vs. specialists.

The first salesperson at a five-person startup is an athlete. They're setting meetings, building scripts, setting up the CRM, running the first demo, and telling engineers what to build. No playbook. No territory. No coach. Just hustle and adaptability.

The hundredth salesperson hired into a 10,000-person company is a specialist. They run the San Francisco healthcare territory for mid-market. They have a pitch deck, a coach, and a defined lane.

This maps to the four stages Mark has lived across:

Private: Hustle. Athletes. 10% strategy, 90% execution. You have an idea at 9 AM and you've tested it by lunch.

Public: Polished. Specialists. 90% strategy, 10% execution. You have an idea and it takes 18 months to test it because the organization moves like a tanker.

Academia: Slow. Rigorous. One question (like "should you specialize your sales reps?") becomes a five-year tenure-track research study. But this is where truth gets pressure-tested across contexts: North America, Africa, Asia. 10-billion-dollar companies and 10-dollar ones. Healthcare and tech.

VC: Pattern recognition. 500 companies reviewed per investment. Founder-picking. Market-sizing. And the uncomfortable reality that just because you need money and a VC has money doesn't mean it's a fit.

The lesson for operators: know which mode you're in. Know which mode your people need to be in. And stop hiring specialists for athlete jobs (or vice versa).

One Last Thing on AI

I asked Mark about AI with a few minutes left. His answer was concise and worth repeating.

"It's a huge bubble. Lean in anyway."

A lot of what you're reading right now is more Pets.com and Webvan than it is Google. But the technology is going to define everything. Start playing with it. Be patient when it slows you down at first. Become AI-enabled.

And if you want a reality check: go into ChatGPT and ask whether first movers usually win or fast followers do. Read the studies. Then decide how fast you really need to move.

Do This Today

1. Define your Leading Indicator of Retention. What can you observe in the first 30 days that predicts a customer stays? Policy count. Feature adoption. Message volume. Usage milestones. Find the number. Build your sales process around making sure every customer hits it.

2. Audit your comp plan against coasting. If your top reps can hit a revenue threshold and downshift into maintenance mode indefinitely, you're wasting their talent. Explore a gated commission structure that rewards continued hunting.

3. Add a coachability test to your next sales interview. Role play. Self-assessment. One piece of feedback. Second role play. You'll know in 30 minutes whether this person can grow or if they've already peaked.

Subscribe to Finding Peak for frameworks that work and zero fluff: ryanhanley.com/subscribe

This is the way.

Hanley.