Customer Success Playbooks skill for Codex

Customer success works when every account has a clear next action: new customers reach first value on a known timeline, risk is spotted from data weeks before the renewal, and renewals and expansions follow a calendar rather than a scramble. A health score is only useful if it would have predicted last year’s churn.

Workflow

1. Gather data: account list with ARR, plan, renewal dates, CSM owner; product usage events; support tickets; NPS/CSAT; billing status; churned and expanded accounts from the last 12-24 months with reasons.
2. Segment accounts into coverage tiers.
3. Define the onboarding journey and the first-value milestone.
4. Build the health score and back-test it.
5. Write trigger-based plays (risk, expansion, advocacy).
6. Set the renewal timeline and business review cadence.
7. Instrument: where each signal lives (CRM, CS platform, warehouse) and who acts on it.

Customer-facing emails, in-app messages and CRM updates are actions on the user’s systems: draft them, and execute or send only with explicit approval.

Segmentation

Tier	Typical criteria	Coverage
High touch	Top ARR accounts or strategic logos	Named CSM, monthly cadence, executive business reviews
Mid touch	Mid ARR	Pooled or named CSM with larger book, quarterly check-ins, triggered outreach
Tech touch	Long tail, self-serve	Automated onboarding, in-app guidance, email sequences, community; humans on triggers only

Set ARR thresholds from the actual distribution and CSM capacity (book size per CSM varies widely by product complexity; derive it from hours per account per month).

Onboarding to first value

• Define first value per product as an observable event: “first report shared with a teammate”, “first 100 invoices synced”. Measure time to first value (TTFV).
• Kickoff (high/mid touch): confirm the customer’s goals, success criteria, stakeholders (champion, executive sponsor, admins), timeline, and technical requirements.
• Success plan:

# Success plan - <Account>
Customer goals: <1-3 measurable business outcomes, in their words>
First value milestone: <event> by <date>
Stakeholders: champion <name>, sponsor <name>, admin <name>
| Milestone | Owner (us / them) | Due | Status |
|---|---|---|---|
| SSO configured | Customer IT | 11-10 | |
| Data import complete | CSM + customer admin | 11-14 | |
| First workflow live for 1 team | Champion | 11-21 | |
Risks: <e.g. IT capacity, competing project>

• Escalate when a milestone is more than a week late; onboarding stalls are the strongest early churn signal.

Health score

Build from a few signals that are available for every account and plausibly cause retention, then test them.

Component	Example signal	Example weight
Usage depth	Weekly active users / licensed seats, last 30 days	30%
Adoption breadth	Key features used out of the 4-5 that predict retention	20%
Support	Open high-severity tickets, ticket trend, CSAT	15%
Sentiment	NPS or CSM pulse, champion engagement	15%
Relationship	Executive sponsor identified and engaged in last 90 days	10%
Commercial	Payment status, contraction requests, discount dependency	10%

Score each component 0-100 with explicit bands, then weight. Example usage band: 0 if under 20% seat utilisation, 50 at 20-50%, 100 above 50%. Map totals to Green (70+), Yellow (40-69), Red (under 40), and allow a CSM override only with a written reason and expiry date.

Back-test before rollout:

1. Compute the score for every account as it stood 90-120 days before its renewal over the past year (or as close as the data allows).
2. Compare to outcomes (renewed, contracted, churned).
3. Check: what share of churned accounts were Red or Yellow beforehand (recall), and what share of Red accounts actually churned (precision).
4. Drop or reweight components that do not separate churned from retained accounts. A simple score that predicts beats a complex one that does not.

Typical SQL shape for the usage component (warehouse with an events table):

select a.account_id,
       count(distinct e.user_id) filter (where e.occurred_at >= now() - interval '30 days')::numeric
         / nullif(a.licensed_seats, 0) as seat_utilisation
from accounts a
left join events e on e.account_id = a.account_id
group by a.account_id, a.licensed_seats;

Trigger-based plays

Trigger	Play	Owner	Within
Health drops to Red	Diagnose (usage, tickets, stakeholders), call champion, agree a recovery plan	CSM	5 business days
Champion leaves (email bounce, LinkedIn change, CRM update)	Identify new champion, re-onboard, inform sponsor	CSM	10 business days
Seat utilisation under 30% for 60 days	Adoption workshop, admin training	CSM / tech touch email	2 weeks
Usage above 90% of plan limits	Expansion conversation	CSM + AE	2 weeks
Repeated severity-1 tickets	Support escalation review, exec apology if warranted	Support lead + CSM	1 week
NPS promoter (9-10)	Ask for review, reference or case study	CSM / marketing	2 weeks

Each play: entry criteria, steps, message templates, exit criteria, and how the outcome is logged.

Renewal timeline

Days before renewal	Action
120	Review health, usage and value delivered; flag risks; agree renewal strategy with AE
90	Business review with customer showing outcomes vs success plan; discuss next year’s goals
60	Send renewal proposal (pricing, terms, expansion options); identify signer and procurement steps
30	Resolve open terms; escalate stuck renewals to leadership
0	Signed; update CRM and success plan for the new term

Auto-renewal notice windows in contracts can be 60-90 days; check them and start earlier if needed.

Customer business reviews (QBR / EBR)

Structure: their goals restated, outcomes achieved with their data, adoption and usage trends, support summary, roadmap items relevant to them, recommendations, goals and success plan for next period, asks (from both sides). Lead with their business results, not your feature list. Internal quarter reviews belong to okr-reviews.

Churn analysis

For each churned or contracted account record: primary reason (from a fixed list such as product gap, low adoption, budget, acquired/shut down, competitor, poor experience), whether health predicted it, and what play ran. Review quarterly; feed product gaps to product with ARR attached.

Gotchas

• Health scores built on opinion (CSM gut feel) with no back-test; they look scientific and predict nothing.
• Counting logins as engagement; measure the actions that deliver value.
• Treating all accounts the same; tech-touch accounts drown CSMs and high-value accounts get neglected.
• Starting renewal conversations at 30 days, after budgets were set.
• Expansion pitches to accounts with open severity-1 issues.

Output

Segmentation rules, onboarding journey and success plan template, health score definition with back-test results, play library, renewal calendar, business review template, and churn reason taxonomy.