Every board conversation that starts with ARR growth ends at retention. A strong acquisition number papers over a leaking retention number for exactly one quarter before the board learns to reweight. This page is for the CEO who wants to know what to ask about retention, what numbers to track, and when to escalate, without delegating the understanding to the VP CS.
20%
Save rate when at-risk identification happens inside 30 days of renewal
45%
Save rate when intervention begins 60+ days out, with scored triage
1:5
Cost ratio of retention vs acquisition. Every dollar of preserved ARR is worth roughly 5x its acquisition-cost equivalent
Four numbers every CEO should own directly.
Reporting
Most CEOs monitor ARR growth and churn rate. Those two numbers are necessary and insufficient. Four retention metrics belong on the CEO dashboard, reported monthly, not quarterly.
Gross Revenue Retention (GRR). Percentage of beginning-period ARR retained, excluding expansion. GRR directly measures the product-market fit of your installed base. A healthy B2B SaaS GRR lives between 88 and 95%. Below 85%, you have a product or ICP problem, not a CS execution problem.
Net Revenue Retention (NRR). GRR plus expansion. Healthy NRR ranges from 105% to 125%. NRR above 110% generates roughly 30% of your ARR growth from existing customers, which is the cheapest growth in the business.
Renewal Forecast Accuracy. What percentage of what you expected to renew actually renewed? A team that consistently forecasts renewals within 5% is operating on signal. A team whose forecast swings 15%+ quarter over quarter is operating on hope.
Early-Warning Lead Time. How many days before the renewal date do you first flag an account as at-risk? Below 30 days is late. 60-90 days is healthy. Above 120 days is aspirational unless the scoring model is sophisticated.
The CEO's one-line test
Can you, as CEO, state last month's GRR, NRR, forecast accuracy, and lead time from memory? If not, retention is not a CEO metric at your company yet. It is a VP CS metric you are hoping stays good.
Five questions to ask every month.
Governance
The pattern of questions matters more than any single one. These five, asked in the same order every month, force the retention system into transparency. If the answers are vague, the system is the problem.
QUESTION 01
What is our current early-warning lead time?
The right answer is a specific number of days, measured across the at-risk book. Below 30, the org is reactive. At 60-90, the scoring model is working. The question itself signals that you care about when the org identifies risk, not just how much of it exists.
QUESTION 02
Which three accounts in the T2 At-Risk queue have the highest ARR at stake?
A specific question that surfaces specific accounts. If the team cannot produce the three names with their current risk scores and active save plays, the triage queue is not operational. You are running on spreadsheet renewals.
QUESTION 03
What is our save-play win rate this quarter vs last quarter?
The flywheel question. If the team cannot produce the number, save plays are folklore. If the number is dropping, the playbook is decaying and needs retuning. If it is rising, the flywheel is working, and you should reinvest in the signal layer.
QUESTION 04
Which accounts did we walk away from and why?
Walk-away discipline is a CEO-level indicator. Teams that cannot name a walked-away account are spending CSM hours on unsaveable accounts. Teams that can name them and explain the reason are making real capital-allocation decisions.
QUESTION 05
What is the churn pattern telling us about ICP?
The Phase 1 question hiding inside the retention metric. If churned accounts cluster on a segment, a motion, a product configuration, or a funding profile, your ICP needs tuning upstream. Churn is a lagging indicator of ICP quality.
Four escalation triggers a CEO should not delegate.
Thresholds
Most retention losses are routine and belong to the CS team. Four patterns are not, and they are the CEO's to watch and escalate directly.
GRR drops more than three points quarter over quarter. Not a noise-level move. Indicates a structural shift: product, ICP, competitive, or organizational. Requires CEO-level diagnosis.
NRR inverts below 100%. Your existing base is shrinking faster than expansion can offset. Every acquisition dollar becomes less efficient. This is a board-level reporting threshold that you deliver to the board yourself, with your reading of the cause.
A strategic account above 5% of ARR shows a composite renewal risk score in the critical zone. Customer concentration combined with risk is an existential variable. CEO calls the executive sponsor personally, does not hand it to the CSM.
Forecast accuracy misses by more than 10%. The team's operational instrumentation is wrong. Not a renewal issue per se; a system-credibility issue. Intervention is on the scoring model and the forecasting process, not on the CS team.
Retention metrics become a CEO dashboard, not a buried CS report. GRR, NRR, forecast accuracy, and lead time are continuously computed and directly queryable. The five monthly questions above can be answered in under two minutes via the Board Report or the in-app Drafter agent, without a RevOps analyst building the deck.
Board Report
Live GRR, NRR, pipeline coverage, forecast health, and risk distribution. Generated from canonical data, not a reconciled spreadsheet. Ready when you are.
Renewal Forecast Accuracy
Tracked over time against actual outcomes. When the team's forecasts miss, the math gets visible, not the blame.
Drafter for the CEO Seat
Ask the five monthly questions in natural language. Drafter routes the 75 MCP tool surface to produce the specific three at-risk accounts, the current save-play win rate, the walked-away account list.
Escalation Signals
Automatic flags when a strategic account above 5% ARR crosses a risk threshold. You know before it hits the CS queue, not after.
Churn-to-ICP Correlation
Churned accounts' ICP scores at close are tracked and analyzed. The signal for when retention loss is actually an ICP definition problem upstream.
