At hundreds of renewals per quarter, you cannot sort by renewal date.

At 50 accounts, the CSM carries the portfolio in their head. They know which accounts are healthy, which have sponsor risk, which had a bad QBR three weeks ago. Renewal date is a reasonable proxy for attention because the CSM's instinct fills in everything else.

At 500 accounts, that instinct breaks. Three structural reasons:

1. The signals that predict churn fire everywhere, not just at the renewal date. Usage decline, stakeholder turnover, budget-cycle shifts, support spikes, champion departures. They are happening right now on an account whose renewal is 180 days out. Sorted-by-date misses them entirely until the account is already lost.
2. Risk and ARR are not correlated. Your riskiest accounts are not necessarily your biggest, and your biggest are not necessarily your riskiest. Sorting by one dimension guarantees the other gets under-served.
3. Effort-to-save varies 10x across accounts. Some accounts are recoverable with one well-timed email. Others require a six-task coordinated play. Sorted-by-date treats them identically. The CSM spends 20 hours on an account that was unrecoverable and skips the one that would have been saved with 30 minutes.

The architectural fix is not a better sort. It is a scored queue with tier-based actions, and a prioritization formula that accounts for all three dimensions at once.

Every renewal in the portfolio sits in exactly one tier, scored daily. The tier determines the action, the owner, and the cadence slot. No account is "in review" or "pending a conversation." The tier is the state.

Two governance rules keep the queue honest. First: every account must be in exactly one tier. No account sits in "we'll figure it out next week." Second: transitions are auditable. When an account moves from T3 Monitor to T2 At-Risk, a log entry captures which signals fired and who decided the transition. Six months later, you can replay what happened.

Inside T2 At-Risk, you still have to rank. A CSM with 40 at-risk accounts cannot work all of them. The ranking is a simple three-variable formula.

Each variable needs definition:

1. ARR: annualized contract value, ideally net of expected contraction. A $200K account where the customer has already signaled they want to reduce seats by 30% is really a $140K save target, not $200K.
2. Save Probability: historical win rate of the save play applicable to this account's signal pattern, scoped to segment and risk profile. If your renewal_save play has a 62% win rate on Enterprise-tier accounts with champion-departed signals, this account inherits that base rate. No guessing.
3. Expected Effort: CSM hours required to execute the save play, from the play library's estimated effort field. A five-task deal-rescue play is ~15 hours. A one-touch re-engagement is ~2 hours. Effort is an explicit number, not a vibe.

A CSM looking at a queue sorted by Priority makes fundamentally different choices than one looking at a queue sorted by renewal_date. The highest-priority account might be six months out with $350K ARR and a 70% save probability at 8 hours of work. Sorted by date, it doesn't show up until Q4. Sorted by priority, it's the first save of the week.

Every T2 and T1 account needs a save play assigned. "I'll figure out what to do" is the tribal-knowledge failure mode that breaks at scale. A library of four to six named save plays covers the majority of at-risk patterns, with a clear rubric for when to deploy each.

Two rules that keep the play library credible. First: plays have measured win rates, tracked per segment and signal pattern. Under 30% win rate on enough samples means retire or rebuild the play. Second: plays have a walk-away criterion. If a play has been active for its defined duration and the customer has not engaged, the CSM escalates to a save-or-close decision with leadership. No eternal open saves.

Not every at-risk renewal is saveable. The hardest discipline in triage-at-scale is deciding which accounts get the CSM's time and which get a graceful close. Three criteria make this decision explicit instead of political.

Calling the walk-away early is not giving up. It is re-allocating scarce CSM capacity to accounts where a save is actually possible. Teams that never walk away end up saving 20% of the unsaveable and losing 40% of the saveable because they ran out of hours.

Every resolved save, whether the outcome is renewed, expanded, churned, or lost, produces a data point. Every outcome feeds back into the save-play library's win rates, which feed back into the prioritization formula, which feeds back into which accounts get CSM time next quarter. The system gets smarter with every save.

1. Every save play closes with a structured outcome. Renewed at full value, renewed with contraction, expanded, churned, lost to competitor, lost to budget. Not "done."
2. Outcomes roll up to per-play win rates, scoped by segment and signal pattern. Retire plays under 30%. Promote plays over 65%.
3. Win rates feed Save Probability in the prioritization formula. Next quarter's triage ranks accounts with more accurate probability estimates than last quarter's.
4. Walk-away decisions produce their own learnings. Patterns of walked-away accounts surface ICP-fit issues, product-gap flags, and segmentation mistakes that inform Phase 1 work upstream.

Without the flywheel, save-play effectiveness is a perpetual hypothesis. With the flywheel, it is a number that improves every quarter.

For EdTech orgs with two motions on shared infrastructure (K-12 plus higher ed, academic plus corporate L&D, horizontal plus vertical specialty), the triage queue stays shared but tier thresholds and save plays diverge. Three adjustments:

1. Tier thresholds are motion-specific. "Usage decline" triggers T2 At-Risk at different thresholds for an academic-cycle customer (45-60 days inactivity mid-term) versus a credentialed-outcome customer (10-14 days pre-assessment). Shared queue, scoped triggers.
2. Save play library forks. Renewal Save, Champion Rescue, Adoption Recovery, Competitive Displacement. Each motion gets its own variant with motion-specific talk tracks, stakeholder maps, and win-rate benchmarks. Shared library shape, divergent content.
3. Cross-motion handoff at save time. If a Motion A customer at risk is also a Motion B prospect, the save play explicitly flags the cross-sell implication. Losing the Motion A account closes the Motion B door for at least a year. The save calculation includes the downstream motion too.

Twelve yes/no questions to audit triage maturity. Count the no's. That's your rebuild order.

Six two-week phases. Ship the queue first. Then the prioritization math. Then the play library. Then the flywheel. Do not try to build everything in parallel.

Two anti-patterns to avoid. First: launching the tier system without the prioritization formula. T2 At-Risk becomes a dumping ground, and nothing actually gets worked first. Second: launching save plays without measured win rates. The library becomes folklore instead of infrastructure, and you cannot improve what you cannot measure.