Health, Metrics, Renewals & Expansion

A dashboard full of green numbers can sit on top of a dying account. The skill is knowing which number, if it moved the wrong way last week, should ruin your morning.

Start with the one metric that gates everything else: activation rate, the share of provisioned seats that are actually active. A 2,000-seat contract with 500 active seats is a renewal that's already losing. Unused seats aren't neutral, they are churn in waiting, because at renewal the buyer divides spend by real usage and the math turns against you.

Activation gets a seat in the door. Depth tells you whether the developer stays. A seat that opens Cursor to read a file once a week is technically active and commercially worthless. What you want is evidence that Cursor is woven into how the developer works.

Sort every signal into leading or lagging. Renewal and expansion are lagging: by the time they move, the outcome is already decided. Activation and depth are leading: they move months earlier and you can still change them.

A metric a CTO doesn't care about is noise, however clean it looks. Acceptance rate means nothing to a VP of Engineering until you connect it to something they already track, like review load or how fast a team ships. Translate before you present or you'll get the polite nod that precedes a downsell.

A health score isn't a grade you assign at quarter-end. It's a tripwire system that tells you which account to call today and what to say when you do.

Usage alone is a thin signal. An account can have strong weekly-active seats and still be one resignation away from churn. A real health view blends three inputs, because each one fails in a way the others catch.

Relationship strength deserves its own weight. Single-threaded accounts feel easy because one champion smooths every meeting and that comfort is the danger. Champion departure is one of the most common ways an enterprise account dies and when your only thread walks out you have no economic buyer and no power-user base to carry the renewal.

The point of a score is to fire the right play. Define thresholds and attach a different motion to each band, so the color tells you what to do, not just how to feel.

Renewal risk is won or lost months before the contract date. The score exists to surface that risk while you can still act, so put a clock on it.

1. 1Score weekly. Refresh the three inputs on every key account so a drop surfaces in days, not at QBR.
2. 2Trigger on change, not just level. A green account sliding toward yellow is more urgent than a stable yellow.
3. 3Attach an owner and a date. Every red and yellow gets a named recovery action with a deadline.
4. 4Open the renewal window early. Begin the renewal narrative roughly 90–120 days out, longer for the largest accounts.
5. 5Close the loop. Re-score after the intervention to confirm the play worked, then move the account band.

Tell a room of senior engineers that Cursor makes them "55% faster" and you've lost them. They've read the same studies you have and they know that number doesn't survive contact with a real codebase.

You need the productivity vocabulary and its limits in equal measure. Know what each term means, where it's measured and where it breaks, because a skeptical VP of Engineering will test exactly that.

The trap is attribution. AI-assisted coding lives inside a system with reviews, meetings, flaky tests and scope churn, so isolating Cursor's contribution to a clean percentage is something an honest engineer knows you can't do. Reach for an unprovable figure and you forfeit the credibility the whole renewal rests on.

Anchor instead on outcomes the customer already measures, then layer developer-experience signals on top. You're not inventing a new scoreboard, you're showing movement on theirs.

• Outcomes they already track: review turnaround, time-to-first-PR for new hires, toil tickets closed, on-call load.
• Adoption telemetry as evidence of value, not as the value itself: weekly-active depth, acceptance, feature breadth.
• Developer-experience signals from surveys and EBRs: "I'd hate to give this up," perceived friction removed, satisfaction.
• Concrete workflow stories: the migration a team did in two days, the legacy module a new hire understood with @-symbols and Chat.

The strongest narrative pairs hard usage data with one specific story. Numbers earn the meeting; the story is what the CTO repeats to their own boss.

The renewal is decided long before the contract date. If you start building the case the month it's due, you're not running a renewal, you're hoping for one.

Begin the narrative months early. The renewal case is an evidence file you assemble over a quarter, not a deck you write the week before. By the time you sit down with the economic buyer, the proof of realized value should already be obvious to both of you.

1. 1Re-open the goals. Re-confirm the original business objectives the customer bought against, in their own words.
2. 2Show progress against them. Map activation, depth and outcomes to each goal - value realized, not features shipped.
3. 3Surface and name the risks. Budget pressure, champion change, a competing tool, a stalled team. Write them down.
4. 4Build the forward plan. Bring a next-phase rollout and expansion view, so the conversation looks ahead, not back.
5. 5Align with Sales on commercials. Agree timing and pricing with the AE before the customer ever sees a number.

Re-confirming goals matters because the customer's memory drifts and their org changes. The pain that justified the purchase may have a new owner or a new name. Restate the goals in their language and show movement against them, so the value is framed on their terms rather than yours.

Then hunt the risks before they hunt you. Each common churn driver has a distinct counter-move and naming it early is what gives you time to act.

Walk into the executive conversation with a plan, not an invoice. A backward-looking renewal asks the buyer to pay again for what's already done. A forward-looking one shows them the next phase of value and makes the renewal the obvious step toward it.

Expansion isn't a separate sales motion you bolt on at quarter-end. It's the natural next step of a rollout that's actually working and your job is to see it before the customer asks.

Expansion follows adoption. A team running at high activation, with deep daily usage and a vocal power user, is the beachhead for the next team. The proof already exists inside the account, so the pitch becomes "do what that team did," which is far easier than selling from zero.

Keep a live map of customer fit, because the product gives you new expansion triggers constantly. At a company shipping features monthly, every release is a re-engagement moment: a new capability is a reason to call the account, demo it against a known need and open a deeper or wider footprint.

Net revenue retention is the metric that compounds. Landing and expanding inside an account is the ADM's growth engine, because an existing happy account is cheaper and faster to grow than a new logo and NRR above 100% means your book grows even if you never add a customer.

Engineering leaders don't buy "productivity." They buy the ability to ship faster without the defect rate creeping up. The whole ROIReturn on Investment. The value gained versus what it cost, the language an economic buyer funds deals in. story fits in those two axes: velocity and quality, held together.

Velocity is one axis, quality is the other and a credible case has to move on both at once. A team that ships twice as fast while customer-facing defects climb hasn't gained anything a CTO will pay to keep. Frame the conversation as the pair, then go one level deeper on each.

The velocity pillar

Velocity has three layers and they trade off against each other. The easiest one to measure sits furthest from the outcome the business cares about; the one closest to the outcome is the hardest to capture. Pick the layer deliberately and say which trade-off you're making.

Reach past PR count when you can. Feature-completion velocity is harder to instrument, yet it's the only velocity layer that maps cleanly to value the customer experiences. A wall of merged PRs can sit on top of zero shipped features; a completed feature can't be faked the same way.

The quality pillar

Quality is what keeps velocity honest. Lead with the number that matters most to the business and treat the rest as supporting evidence.

• Customer-facing defects - the headline. Flat or declining is the win; if shipping faster pushes this up, the velocity gain was an illusion.
• Test coverage - whether the safety net grew with the code, so faster shipping isn't quietly raising risk.
• Code half-life - how long a line of code survives before it's changed again. Longer-lived code points to fewer thrash-and-rewrite cycles.
• Developer sentiment - whether engineers trust the codebase and the workflow, an early signal before defects ever surface.

The instrumentation Cursor gives you

Three product capabilities turn this frame from a whiteboard model into something you can actually measure inside an account.

Cursor BlameAn augmented git blame that records line-level human and agent co-authorship, so you can trace which code was written by AI versus a person. is worth a careful pitch, because attribution is exactly where ROIReturn on Investment. The value gained versus what it cost, the language an economic buyer funds deals in. conversations get stuck. Line-level co-authorship through the AI-code-tracking API lets a customer answer "how much of our shipped code did the agent write?" against their own definition of done - which is far more defensible than any percentage you'd quote them.