The Renewal Forecast: Calling Your Number 90 Days Out

The first time I had to call a quarterly renewal forecast in front of a CRO, I did what most CSMs do. I looked at my portfolio, color-coded the accounts red-yellow-green based on my gut, added up the green numbers, knocked a polite percentage off the yellow numbers, and presented it as the forecast. I was about eleven points off. On a portfolio in the low eight figures, eleven points is not a rounding error. It is a board-level miss.

The CRO did not yell. He asked a worse question: "What changed between when you called this and today?" I did not have a real answer, because I had not had a real model. I had had a feeling.

What follows is the framework I rebuilt to replace the feeling. It is not complicated. It does not require a data science team. You can run it in a spreadsheet on a Sunday afternoon. What it requires is the discipline to score each account against the same criteria every time, and the honesty to mark a green account yellow when the mechanics say it should be yellow.

The output is a portfolio-level forecast with a confidence interval you can defend. The byproduct — which is almost more valuable than the forecast itself — is a list of exactly which accounts need which interventions, and in what order, between today and the renewal call.

The problem with how most teams forecast

Most CS teams forecast renewals one of three ways. All three are broken in different directions.

The vibe forecast

The CSM owns a relationship, knows the account, and says a number. The number is usually optimistic by 8 to 15 points because the CSM has spent six months talking themselves into the renewal. There is no model behind it, so when reality lands differently, there is nothing to debug.

The dashboard forecast

The team uses a health score from their CS platform. The health score is built on activity data — logins, feature adoption, support ticket volume — and produces a green-yellow-red label. The team forecasts every green account at 100% and every red account at 0%. This is worse than the vibe forecast, because it laundered the gut feel through a dashboard and now everyone trusts it. Most health scores measure correlation with the past, not causation with the future, and they go green right up until the moment an account churns. We will fix this in the health scores piece.

The commit / best-case forecast

Borrowed from the sales team. Each CSM gives a "commit" number (very confident) and a "best case" number (everything goes right). This is better — at least it acknowledges uncertainty — but it has the same underlying problem: it does not break the account down into the actual variables that determine the renewal. You are just uncertainty-quantifying the gut feel.

The four-tier model

The model has two parts. First, you score each account on four dimensions. Second, you map the composite score to a renewal probability and a recommended action. That is it.

The four dimensions are deliberately chosen to capture the only things that actually move a renewal:

1. Commercial signal. Has the account taken any action that signals intent — budget approval, procurement engagement, multi-year discussion, expansion question?
2. Champion strength. Do you have a sponsor with enough political capital and tenure to actually drive the renewal? Or is your champion a senior IC who can advocate but cannot decide?
3. Adoption / outcome. Is the customer getting the outcome they bought you for? Not "are they logging in." Are they achieving the business result that justified the purchase?
4. Risk events. Has anything happened in the last 90 days that materially changes the picture — leadership change, acquisition, layoff, security incident, executive sponsor exit, major bug, missed SLA, public negative review?

Each dimension is scored 1 to 4. The definitions matter — not the numbers. The numbers are just a way to make the definitions add up.

Commercial signal — the scoring rubric

A score of 1 inside 90 days of a renewal is a five-alarm fire, regardless of how green the rest of the account looks. If nobody on the customer side has acknowledged the renewal exists, you do not have a renewal — you have a hope.

Champion strength — the scoring rubric

The reason champion strength matters more than most CSMs admit: a champion at score 1 or 2 cannot save a renewal that hits a procurement headwind. They can only advocate. When procurement squeezes you for a discount or a delay, you need a 3 or a 4 to absorb the pressure. If you are a 1 or 2 inside 90 days, the work of the next 60 days is to fix that, not to nail the proposal.

Adoption / outcome — the scoring rubric

The trap here is grading on activity instead of outcome. A score of 3 requires the customer to be using the product. A score of 4 requires them to be able to say what they got in business terms — dollars, hours, deflected tickets, faster time-to-resolution, whatever it is. If you cannot get them to say it, you do not have it, no matter what the dashboard shows.

Risk events — the scoring rubric

Risk events are where the model earns its keep against the dashboard. A health score will not catch that your champion's VP just left. You will, if you are looking for it. The discipline of forcing yourself to answer "what happened in the last 90 days" — every 90 days, on every account — is most of the value of this exercise.

From four scores to one forecast

You now have four scores per account, each from 1 to 4. The composite is the sum, ranging from 4 to 16. The composite maps to a renewal probability:

The probabilities are not magic — they come from looking at three years of renewal outcomes against the scoring rubric across multiple portfolios. You will recalibrate them for your business. A high-touch enterprise portfolio with strong onboarding will see A tier closer to 98%. A self-serve product with light services may see A tier closer to 88%. Run the model for one quarter, compare forecast to actual, adjust.

The portfolio-level math

Once each account has a probability, the forecast is mechanical:

Forecasted renewal ARR = sum of (account ARR × probability) for every account in the quarter.

And the confidence interval is the difference between forecasting every account at its probability versus forecasting at the next tier up and the next tier down. That spread is your honest range.

You can now walk into a CRO conversation with three numbers: a commit, a forecast, and a stretch. All three derive from the same model. When the CRO asks "what changed?" you can point at the accounts whose score moved, and which dimension moved, and what event caused it. The forecast becomes auditable.

A worked example

Consider a portfolio of 12 accounts up for renewal in the same quarter, with total ARR of $4.2M. Here is what the model produces:

Total ARR at risk: $4,205,000
Forecast (probability-weighted): $3,248,000
Implied gross renewal rate: 77%
Honest range: $2,940K (downside) to $3,560K (upside)

Two observations the model surfaces immediately. First, three accounts — Solace, Atlas, and Vanta — represent $715K of ARR and will not renew without intervention. That is 17% of the portfolio. If the team treats them as "we'll work them in the renewal quarter," two of the three are gone. Second, Trident Energy at $310K is scoring an 11 only because nobody has had the renewal conversation yet (commercial signal of 1). The other three dimensions are strong. A single 30-minute call moves Trident from B to A. That call is the single highest-leverage action in the entire portfolio.

That is what the model is for. Not the number at the bottom — the map of where to spend the next 30 days.

How to run it

Run the scoring once per quarter, refreshed once per month inside the renewal quarter. The right time to score is the first Monday of the month — make it ritual. The two failure modes are: (a) scoring once and never refreshing, so the model goes stale, and (b) scoring every week, which turns it into busywork and dulls the discipline of actually changing scores when something changes.

Score the accounts yourself, then have a peer or your manager pressure-test the scoring. The most common error is grading your own accounts too generously — particularly on champion strength and adoption. A peer review catches this. Build it into your renewal rhythm.

When you present the forecast, do not present the spreadsheet. Present three numbers — commit, forecast, stretch — and the names of the four or five accounts that swing the spread. Executives do not want to see twelve rows; they want to know which accounts you are working and what you need from them.

What this does not do

Worth being honest about the model's limits:

• It does not forecast expansion. A separate model handles upsell and cross-sell, because the dynamics are different. This model handles base renewal only.
• It does not replace product telemetry. You still need usage data feeding the adoption score. The model assembles the inputs; it does not generate them.
• It will be wrong sometimes. An A-tier account churns because an acquisition you could not have known about wipes out your champion's role. That is fine. The model's job is to be defensible and improvable, not perfect.

The spreadsheet

A working template of this model — with the four rubrics, the probability mapping, the worked example, and formulas for portfolio-level forecast and confidence interval — is available on the Tools page. Copy it, plug in your accounts, and run your portfolio through it before your next forecast review.

This is the flagship piece of CSM Redefined. If it was useful, The NRR Playbook sends one tactical idea like this every week.