Customer Retention & Churn Signals

When to use

Triggers:

• "Why are customers leaving?" / "We lost [specific customer] — why?"
• "Churn is up" / "Retention dropping" / "Monthly repeat is down"
• "NPS dropped" / "CSAT is trending down"
• "How do I keep customers longer?"
• "Repeat business from [segment] is soft"
• User runs a subscription product, retainer service, repeat-retail business, or hospitality operation

Don't fire for:

• Pure acquisition questions (use a different growth skill)
• One-time purchase businesses with no repeat relationship (different model entirely)
• Enterprise SaaS with dedicated CSMs and customer success software — use a different framework

The retention problem for AU SMEs

Most AU SMEs under 50 staff don't have a retention strategy. They have an acquisition strategy, an occasional thank-you email, and a vague sense that "our customers are happy". Then churn accelerates and they can't explain it.

The reality: customers don't churn when they stop paying. They churn weeks or months earlier — you just see the invoice stop last.

This skill gets you to the early signals so you can intervene before the invoice.

Step 1 — Define "churn" for your model

Not all businesses churn the same way. Pick the definition that matches yours:

• Subscription / membership: Paid customer who cancels or doesn't renew
• Service retainer: Retainer not renewed, or paused for >90 days
• Repeat retail: No purchase in [industry-typical] window (most retail: 90 days; hospitality: 60 days; local services: 180 days)
• B2B project: Client who was supposed to come back for a new project and didn't
• Professional services: Client whose last invoice was >2x the typical cycle ago

Write down your definition. Share it with your team. Without a shared definition, every conversation about "churn" is incoherent.

Step 2 — Churn signal scoring model

Customers don't arrive at churn uniformly. There's a behaviour/usage/sentiment pattern that precedes it. Build a simple score per customer:

Behaviour signals (weight: 40%)

• Frequency drop. Last visit / login / purchase is older than their typical cycle × 1.5
• Engagement drop. Emails opened, app usage, meeting attendance — declining trend over 2+ periods
• Channel change. Previously engaged via phone, now only email. Or previously answered same-day, now 3-day delays.
• Decision-maker change. Your champion left the organisation. Flag this immediately — it's the single biggest B2B churn predictor.

Usage signals (weight: 35%)

• Seat/scope shrinkage. Fewer users, smaller job sizes, lighter-tier usage
• Feature pullback. Not using the features they originally bought the product for
• Support tickets up. Especially "why doesn't this work" tickets
• Data exports. Any customer who exports their data has at least considered leaving

Sentiment signals (weight: 25%)

• NPS drop. A customer who rated 9 last survey and 6 this survey is amber, not green.
• Complaint frequency. Even one escalation changes the risk profile
• Review activity. If they left a public review, good or bad, they're thinking about you — lean in
• Direct feedback. "We're exploring other options" said even once is a red signal

Score each customer 0–100. Above 70 = healthy, 40–70 = at-risk, below 40 = critical.

For an AU SME without a BI tool, this lives in a spreadsheet. One row per customer, one column per signal, a formula for the total. Refresh monthly minimum, weekly ideally.

Step 3 — Tiered re-engagement cadence

Healthy (score >70) — maintain

Don't disrupt working relationships with too much marketing. Keep to:

• Monthly value newsletter (the same as all customers get)
• Quarterly check-in from their main contact — genuine, not scripted
• Public shout-out on social when they hit a milestone worth noting
• Renewal reminders well ahead of any expiry

At-risk (score 40–70) — re-engage

Personal touches, not automated sequences:

• A direct call or email from a senior person (not the account manager, not a mass send). "Been a bit since we connected — everything good?"
• Specific reference to their last engagement: "We noticed your team hasn't used [feature] in a while — worth a refresher session?"
• Offer a specific value add that costs you little: a 1-pager on something relevant to them, an intro to another customer, a preview of an upcoming feature
• No discounts yet. Discounts signal "we know you're leaving" and often accelerate the decision

Critical (score <40) — save or learn

• Executive-to-executive call where possible. If your CEO/founder isn't calling when a customer is critical, nobody should be
• Diagnose before proposing. "What's driving this?" before "here's what we'll do"
• If they're leaving, learn as much as you can before they do. Exit survey, exit call, exit email — some signal is better than none
• Don't discount your way out unless discounting was the cause (pricing objection). Discounting a product-fit or service-quality problem masks it.

Step 4 — The 5 questions for every churned customer

Even the ones who left quietly. Send a short email (not a survey tool — plain email from a real person):

1. What were you trying to accomplish when you started with us?
2. What prevented you from accomplishing it?
3. What did we do well? (capture signal even when they're leaving)
4. Who or what will you use instead?
5. Under what conditions would you consider coming back?

Reply rate on these plain-email versions is typically 15–30% vs <5% for SurveyMonkey-style exit surveys. Read every reply personally. Log each answer in a shared sheet — patterns emerge over 10 responses.

Step 5 — Retention ops cadence

Build the rhythm into your operations:

Weekly (30 min)

• Score update for any customers who had activity this week
• Review the at-risk list — any deterioration?
• Assign one "re-engage" action per at-risk customer

Monthly (90 min)

• Full score refresh across the customer base
• Review churn from last month — root cause per customer
• 5-questions responses: what's the pattern?
• One pricing / packaging / product tweak committed based on the month's signals

Quarterly (half day)

• Cohort analysis — are customers acquired in Q1 2026 retaining as well as Q1 2025?
• Segment analysis — which industries / sizes / acquisition channels retain better?
• Share findings with the whole team
• Update the score model based on what's actually predictive

Special cases

When the champion leaves

B2B: your main contact at the customer moves on. Default behaviour — relationship withers. Better:

1. Within 24 hours of finding out, send a congratulations to the leaver
2. Within 48 hours, ask "who's picking up [project] on your side?"
3. Introduce yourself to the new contact, summarise the relationship, ask what they'd like to change
4. Send a written handover doc (what you've delivered, what's in flight, what's upcoming)

This converts 40–60% of "champion-leaves" churns to retained accounts. Most SMEs skip it entirely.

Seasonal businesses

Hospitality, tourism, some retail — customers naturally come back annually, not monthly. Your "cycle × 1.5" definition needs to match the season. A holiday park customer who hasn't booked since last Easter is healthy in July, at-risk in October, critical in January.

Service retainer pauses

Some customers pause for 60–90 days due to their own project delays. Don't treat a pause as churn until 90 days elapse AND you've had at least one conversation confirming intent. Track pauses separately.

What to measure

Month-over-month:

• Churn rate — [customers who churned this month] ÷ [customers at start of month]
• Gross revenue retention (GRR) — [recurring revenue at start] + [revenue lost to churn + downgrade] ÷ [recurring revenue at start]
• Net revenue retention (NRR) — same, but add expansion revenue; good AU SME target is >95%
• Time to first churn signal — how long after signup do at-risk signals first appear?
• Save rate — of customers who hit at-risk this month, how many returned to healthy in 30 days?

Don't measure:

• Vanity retention numbers that exclude specific cohorts
• "Logo retention" without revenue context — losing 10 small customers differs from losing 1 big one
• Industry-benchmark comparisons without context — your benchmark is your own last quarter

What this skill does NOT do

• Build the scoring model for you. You supply data; skill produces the scoring logic and a spreadsheet template
• Write re-engagement emails in your voice. Pair with au-cold-outreach-spam-compliant style guidance for the tone
• Handle active complaints. Those are a different workflow (complaint handling, not retention)
• Guarantee outcomes. Retention is a function of product + experience + price + competition. The skill helps you see signals earlier; it doesn't fix fundamental fit issues.

Tier access

Pro. Retention work compounds — the operators who invest in it consistently outperform those who don't. Base-tier members get the 5-questions template and the score-model framework; Pro-tier gets the weekly/monthly/quarterly operating cadence plus review calls.

Related skills

• au-cold-outreach-spam-compliant — for re-activation campaigns
• review-response-au-reputation — reviews are sentiment signals
• ai-roi-measurement-for-sme — same measurement discipline applies
• spreadsheet-analyst — the scoring spreadsheet analysis pairs here

References

• Reichheld — The Ultimate Question / Net Promoter Score
• ACCC — Unfair contract terms (relevant when auto-renewal is in the mix)