practice · operations
SLA review
Quarterly. Three questions. Not a sales conversation. The SLA reviewed quarterly stays a contract; the SLA never reviewed becomes a souvenir.
TL;DR
A quarterly SLA review is 45 minutes, driven by the PO and CS Lead with the named client, and answers three questions: did we meet · where did we approach without breaching · are the categories still right? It is not a sales conversation. Numbers, not impressions. The output is either contract holds, thresholds adjusted, or categories rewritten — signed.
What it is
The SLA review is named in Did We Serve · The Ongoing Relationship — The SLA Review. It is the recurring conversation that keeps the operational contract current. The SLA itself is the threshold past which someone is paged; the review is the conversation that keeps the threshold honest.
Distinguish from
Sales conversation — the failure shape; the SLA review is a contract conversation. Postmortem — bound to one incident; SLA review reads the quarter. Status page — operational, public uptime; SLA review is the client-facing contract review. See Confusable with at the foot.
Why it matters
Without the SLA review:
- The SLA becomes a souvenir. The contract was signed once; nobody reads it; the team measures against drift.
- Breach protocols decay. The team patches breach symptoms without confronting the threshold itself.
- Surprises arrive in escalations. A breach that was approaching for two quarters surfaces as a P0 incident.
- The client and team operate on different models. The categories that mattered at signing don't match the work that's being done now.
The corpus rule from Did We Serve · Ongoing Relationship: an SLA reviewed quarterly stays a contract. An SLA never reviewed becomes a souvenir.
How to do it
Step 1 — Read the numbers before the meeting
The PO and CS Lead read the quarter's numbers before the review, not in it. The review is 45 minutes; reading the dashboard at the table burns 20 of them.
2026-Q2 SLA performance (read before the review):
Availability (target 99.9%): 99.94% met
Response time P1 (target <15m): 12m median met
Resolution time P1 (target 4h): 3h 40m median met
Response time P2 (target 1h): 58m median met (close)
Resolution time P2 (target 1d): 21h median met
Data integrity: 0 incidents met
Approached without breaching:
- Response time P2 trended toward 1h ceiling in May
(week-over-week +8%). Cause: one CS team member
departed; cover not yet fully in place.
Breaches:
- None.Step 2 — Walk the three questions, in order
Q1: Did we meet the SLA? — numbers, not impressions. Three minutes per dimension. If met cleanly, move on; if approached or breached, name the specific story.
Q2: Where did we approach without breaching? — the leading indicators. This is where the review earns its keep. We met every threshold but we were close on P2 response in May — here's why, here's what's changing.
Q3: Are the categories still right? — the deepest question. The SLA was written against a model of the world that may have moved. Does the P1/P2/P3 split still match how the team thinks about severity? Does the maintenance window still match the client's working hours?
Step 3 — Refuse the sales agenda
The SLA review is the moment the client may try to renegotiate scope, pricing, or roadmap. It is not the venue. The PO holds the line:
Client: "Since we're talking about SLA, we'd really like to
discuss adding a third environment for the next
release cohort."
PO: "That's a roadmap conversation; let's hold it for the
bi-weekly sync. For the SLA review, we have ten more
minutes — I want to walk through whether the P2
categories still match how you experience severity."The discipline is short and contract-shaped. Sales conversations are run separately, by named owners, at named times.
Step 4 — Sign the outcome
One of three:
- Contract holds — no changes; sign for the next quarter.
- Thresholds adjusted — the numbers shift; the underlying categories don't. Sign the new thresholds.
- Categories rewritten — the structure changes (e.g., a new severity tier, a maintenance-window redefinition). This is rare and significant. Schedule a follow-up to draft the new contract; do not improvise it in the meeting.
Outcome (2026-Q2 SLA review):
Status: Contract holds.
Adjustment: P2 response target tightened to <50m (from 1h)
starting 2026-Q3, conditional on the CS team
being back to full headcount by 2026-07-15.
Signed: Alex (PO), Dina (CS Lead), client representative.Step 5 — Postmortem any breaches in this review, not as separate meetings
If there were breaches in the quarter, they were postmortemed individually at the time. The SLA review does not re-litigate them — it reads them as data:
Breach summary (2026-Q2):
INC-058 — 2026-04-10 — SEV-1 — duration 22m.
Postmortem signed; chain-level fix (cold-cache warming)
shipped 2026-05-30. Already in client's prior weekly update.Two sentences. The work was done at the time; the review confirms it.
Step 6 — File the review
The signed outcome lives next to the SLA itself, dated. Three months from now, the next SLA review opens by reading this outcome.
Evidence
Across our cycles, SLA reviews that kept the contract honest shared three properties.
- The numbers were read before the meeting. Reviews where dashboards were opened in the meeting ran 20–30 minutes longer and produced thinner outcomes.
- The PO refused at least one off-agenda topic. Reviews where the agenda held produced category-level changes 2× more often than reviews that drifted into roadmap.
- The outcome was signed. Reviews ending with "we'll figure out next steps" produced no documented change in 60% of cases. Reviews ending with a signed outcome — even contract holds — produced documented continuity.
Anti-patterns
| Pattern | What it looks like | Where to fix |
|---|---|---|
| Review becomes sales | Roadmap, pricing, new scope on the table | Clinic — An SLA review that became a sales conversation |
| Numbers read in the meeting | Dashboards opened at the table | Read before; come in with the numbers |
| Breaches re-litigated | A breach that already had a postmortem gets discussed again | Two-sentence summary; the work was done at the time |
| No outcome signed | "We'll follow up" | Sign one of: holds / adjusted / rewritten |
| Review skipped a quarter | "Things have been quiet" | Quarterly is the contract. Quiet is data; review it. |
| Client missing | Internal-only "SLA review" | This is something else (internal capacity review). The SLA review is the conversation with the client. |
Confusable with
| This | Not this | Difference |
|---|---|---|
| SLA review | Sales conversation | Contract-shaped, not deal-shaped. |
| SLA review | Postmortem | Postmortem = one incident; SLA review = the quarter. |
| SLA review | Status report | Status is one-way; SLA review is a signed mutual review of the contract. |
| Contract holds | "All green" | Even holds is a deliberate choice and a signed outcome. |
Further reading
- Canon — Did We Serve · The Ongoing Relationship — The SLA Review · The breach protocol
- Canon — Why We Build · Client Relationship Strategy
- Practice — Weekly client update — the cadence that protects the SLA review
- Checklist — SLA review · agenda
- Clinic — An SLA review that became a sales conversation
- Skill path — PO foundations · Step 5 (client cadence) · CS Lead foundations
- Reference — Area · SLA Definition · Area · SLA Breach