Once you've connected Stay to Claude using the Stay MCP, you can plan campaigns, build your own dashboards, and do even more with your subscription analytics.
This playbook gives you ready-to-use prompts for the most high-impact analyses — organized by topic so you can jump to the area most relevant to your business right now.
Each play includes a prompt you can copy directly into Claude. For deeper analyses, we recommend using Opus 4.7 or Opus 4.8 for the most accurate results.
📌 Note: Not connected yet? Start with Using the Stay MCP: Connecting Stay to Claude before running any of these plays.
Retention Plays
Play 1: Break Down Cancel Reasons by Subscriber Lifecycle Stage
What it tells you: You might know your top cancel reason is "too much product" — but do you know whether those cancellations happen mostly in Month 1 or Month 6? This play breaks cancel reasons down by how many orders a subscriber completed before canceling, so you can see exactly where in the lifecycle each reason spikes.
Why it matters: A "too much product" cancel in Month 3 points to a cadence or quantity problem at onboarding. The same cancel reason in Month 6 might mean a loyalty offer or a swap to a smaller size could have saved them. Knowing the lifecycle stage changes the intervention.
Prompt to use:
Use the Stay AI integration to break down our cancellation reasons by the number of orders completed by the customer for the last month. For example: we know X% of cancels come from "too much product" — but of that group, what percentage canceled in orders 1-2 vs. orders 3-4 vs. orders 5+? Build an interactive table I can sort by cancel reason and order range.
What to look for:
Cancel reasons that cluster heavily in Orders 1-2 could be onboarding problems — cadence, quantity, or expectation mismatch
Cancel reasons that spike in Orders 5+ are loyalty problems — these subscribers had a good run; a targeted offer or product swap may retain them
Any reason with 30%+ of volume in the first two orders is a strong signal to revisit your welcome flow or product pacing
💡 Tip: Once you have the table, you can follow up in the same Claude conversation: Based on this data, suggest what retention offer I should route to subscribers who cancel for [reason] at Orders 1-2.
Play 2: Surface What Subscribers Are Actually Saying in Cancel Survey Free-Text
What it tells you: The sentiment and themes hiding in your cancellation survey's free-text field — without manually reading through every response one by one.
Why it matters: Structured cancel reasons only capture what's already an option in your dropdown. Free-text responses often surface issues — a specific side effect, a pricing complaint, a competitor mention — that aren't represented anywhere in your reporting until someone reads them manually. A sentiment summary tells you whether your survey options still reflect why people are actually leaving.
Prompt to use:
Use the Stay AI integration to pull the free-text responses from our cancellation survey for the last [30/60/90] days and summarize the sentiment themes. Group similar responses together and give me a rough percentage breakdown for each theme (for example, "X% mentioned digestive issues," "Y% mentioned price"). Flag any themes that aren't currently covered by our structured cancel reason options.
What to look for:
Themes that show up repeatedly but aren't already a structured reason option — a sign to add or revise your survey's dropdown reasons
Sentiment clusters that skew toward a specific product or plan tier
Language you can reuse directly in win-back or save-offer messaging
💡 Tip: Once you have the theme breakdown, ask Claude to draft a save offer or messaging angle for the most common theme.
Play 3: Find Save Offers Sitting Unused
What it tells you: Whether subscribers who accepted a save offer in the cancellation flow are actually using it, or just remaining subscribed while skipping or delaying their next shipment.
Why it matters: A subscriber staying "Active" after accepting a save offer looks like a win on paper. But if they're sitting on an unused discount and repeatedly skipping their next order, they haven't actually re-engaged — they're just delaying the eventual cancellation. Catching this gap lets you nudge them before it becomes churn instead of after.
Prompt to use:
For subscribers who accepted [save offer name] in the cancellation flow, show me their order activity since accepting the offer using the Stay AI integration. Did they place their next order on schedule, skip it, or delay it? Break down how many are actively using the offer vs. still holding an unused discount code.
What to look for:
A large share holding an unused code without placing their next order — that's an engagement gap, not a retention win
Whether "still Active but not ordering" correlates with eventual cancellation elsewhere in your data
✅ Best Practice: Build a segment for subscribers holding an unused save offer, then add a Customer Portal banner with a direct "Get It Now" quick-action CTA linking to their next order.
Play 4: Catch Cadence Mismatches Before They Cause Churn
What it tells you: Subscribers whose shipping frequency doesn't match the product variant they're actually on. For example, a 90-day supply shipping every 30 days.
Why it matters: A cadence mismatch means product is piling up faster than the subscriber can use it. That's a quiet, common cause of churn. They're not unhappy with the product, they just have too much of it and eventually cancel rather than deal with the backlog.
Prompt to use:
Using the Stay AI integration, find subscribers whose current product variant duration doesn't match their shipping frequency (for example, a 90-day supply variant shipping every 30 days). List the mismatched combinations and how many subscribers fall into each.
What to look for:
Any variant/frequency combination with meaningful volume — that's worth building a segment around.
✅ Best Practice: Build a segment for the mismatched combination, then launch a dynamic Customer Portal banner prompting those customers to update their frequency. Link it directly to a quick action so they can fix it in one click.
Dunning & Billing Recovery Plays
Play 5: Identify Your Biggest Billing Recovery Opportunities by Error Type
What it tells you: Not all failed billing attempts fail for the same reason — and the right recovery strategy depends on the error type. This play builds a table showing how many subscriptions failed for each error message, how many were recovered, and your recovery rate per error over the last 6 months.
Why it matters: "Insufficient funds" tends to recover well with retries because the problem is temporary. "Card expired" or "payment method no longer valid" won't recover from retries at all — those subscribers need a payment update prompt. Knowing which errors drag down your overall recovery rate tells you where to act and how.
Prompt to use:
Help me use the Stay AI integration to build an interactive HTML table with my dunning/billing attempts data showing recovery rates by error message for the last 6 months. Filter it down to only include subscriptions that failed their first recovery attempts so we avoid double-counting subscriptions under multiple retry error messages
Output: Interactive HTML table with sortable columns (by volume, recovery rate ascending, recovery rate descending, unrecovered count). Include summary cards at the top for total failed, recovered, unrecovered, and overall recovery rate. Tag rows as "opportunity" where unrecovered count is high and recovery rate is below 45%.
What to look for:
Sort by "Unrecovered" descending — the top rows are your biggest revenue leak
Low recovery rate on card/payment errors (expired, revoked, no longer valid) means retries aren't the answer — these subscribers need a direct prompt to update their payment method
High recovery rate on "insufficient funds" is expected and healthy — your dunning sequence is working for transient failures
Rows tagged "opportunity" are where improving recovery would have the most impact on revenue
⚠️ Important: The prompt above filters to first failures only to avoid double-counting. Without that filter, a single subscription can appear under multiple error messages across retries, which inflates failure counts and distorts recovery rates.
Daily Monitoring Play
Play 6: Set Up a Daily Metrics Review That Flags Anomalies
What it tells you: Your top-line metrics can look fine while something breaks quietly underneath. This play runs a daily snapshot of your key subscription metrics and compares each one to its trailing 30-day average — so you see changes in behavior or technical issues before they compound.
Why it matters: If your upsell tool is only running on 30% of orders instead of 100%, your dashboard won't flash a warning — your AOV will just look low. A daily comparison to your own trailing average is the early detection layer that catches those issues.
Prompt to use:
Use the Stay AI integration to pull today's subscription metrics and compare each one to its trailing 30-day daily average. Flag any metric where today's value deviates by more than 15% from the average. Metrics to include: new subscribers, cancellations, cancellation rate, dunning entries, orders processed, AOV, and save rate. Format the output as a table with columns for: Metric, Today's Value, 30-Day Avg, % Change, and Status (Normal / Watch / Alert). Write a 2-3 sentence summary at the top highlighting anything flagged as Watch or Alert.
What to look for:
Dunning entries spike — could signal a payment processor issue, a bulk card expiration, or a recent price change that surprised subscribers
AOV drop — could indicate a product swap issue, a discount applied more broadly than intended, or an upsell or add-on tool running at reduced coverage
Cancellation rate spike — pair with the cancel reason breakdown (Play 1) to identify whether it's a specific reason driving the increase
New subscriber drop — often a traffic or conversion issue upstream, not a Stay issue, but worth flagging early
💡 Tip: You can run this prompt on a recurring cadence each day by keeping a Claude conversation open for your daily check-in or by creating a scheduled task in Claude Cowork.
Revenue Modeling Play
Play 7: Model LTV Impact Before Making a Subscriber Decision
What it tells you: Before you contact, offer a discount to, or make a change affecting a specific subscriber segment, this play estimates the revenue impact. It calculates LTV for the segment and projects the potential revenue at risk if a portion of those subscribers churns as a result.
Why it matters: Decisions like "let's reach out to all 6-month subscribers" or "let's offer 20% off to anyone who's been active over a year" carry real revenue implications. Modeling the downside before acting — especially for segments with high LTV — helps you weigh the risk and right-size the offer.
Prompt to use:
Help me calculate LTV for [describe the subscriber segment — e.g., subscribers who have been active for 6+ months] and model the potential revenue impact if I [describe the action — e.g., reach out to all of them with a check-in offer].
Assume a [X%] churn rate from outreach. Show me: current active count in the segment, average order value, average order frequency, estimated LTV per subscriber, total segment revenue value, and projected revenue loss at the assumed churn rate. Then suggest what churn rate threshold would make this action not worth taking.
What to look for:
Segment revenue value vs. projected loss — if the revenue at risk from churn exceeds the revenue you'd gain from the action, reconsider or narrow the segment
The churn rate threshold Claude surfaces — this tells you how much attrition the action can absorb before it's net negative
High-LTV segments warrant conservative offers — a smaller, more targeted intervention is usually safer than a broad outreach
✅ Best Practice: Pair this play with a cohort comparison if you're considering a discount. Ask Claude to compare LTV for subscribers acquired through a promotional event (like a Black Friday sale) vs. your standard acquisition cohort before extending similar offers to your full base.
Reporting Plays
Play 8: Build a Recurring Subscriber Health & Revenue Report
What it tells you: A reusable weekly or monthly snapshot combining subscriber health (active, new, reactivated, churned, churn rate) and revenue breakdown (first-time vs. recurring vs. add-on) — without rebuilding it from a platform export every time.
Why it matters: Recreating the same report from scratch each week eats time that's better spent acting on what the numbers show. A recurring prompt gives you the same view every time, pulled fresh from Stay, so you can spot anomalies — a dip tied to an ad spend change, a spike around a sale event — as soon as they show up.
Prompt to use:
Build me a recurring [weekly/monthly] subscription health report that includes: active subscriber count and churn count, churn rate, MRR change, save rate, top cancellation survey save offers, top cancel reasons, revenue breakdown (first-time vs. recurring vs. add-on), and period-over-period % change for each. Format as a table with a short summary of any notable changes at the top.
What to look for:
Week-over-week or month-over-month swings in new vs. churned subscribers — flag anything outside your normal range
Whether revenue growth is coming from new subscribers or your existing base
Save rate trending down even if churn count looks flat — an early signal your save offers are losing effectiveness
💡 Tip: If you're using Claude with Cowork, you can schedule this report to run automatically at the start of each week or month instead of re-running the prompt manually.