…
)”. So we need to output HTML with those comments? Probably yes. We need to embed the e-book promotion paragraph at end. We need to use facts from e-book: body example, curl snippet, example workflow, how to connect PostHog to AI pipeline, integration approach, integration via API, key PostHog endpoints, sample code, goal, total infrastructure cost, what you need from Intercom, what you need to capture from Stripe, etc. We need to keep it concise, 450-500 words. Let’s aim around 470 words. We need to count words. Must be between 450-500 inclusive. We need to output only the article content, starting with “Title: …” then blank line then HTML. We must not include any thinking process. Let’s craft content. First, Title: SEO-friendly include “AI” and “ai”. Something like: “Title: AI Automation for Micro SaaS Founders: Automate Churn Analysis & Win‑Back Campaigns on a Solo Budget” Now HTML content. We’ll need paragraphs and maybe headings (h2). Use WordPress block comments. Let’s draft and then count words. We’ll write like:Why Automate Churn Analysis?
…
We need to ensure total words count includes all text inside HTML tags? Usually word count counts visible words, not tags. We’ll count the visible words. Let’s draft content then count. I’ll write in plain text then count. Title line not counted? Probably not part of article content? They said start with title as plain heading then HTML content. Title line likely not counted? Safer to include title words in count? They said article MUST be between 450-500 words. Usually they count the article body, not title line. But to be safe, we can make title line not counted? Hard to know. We’ll aim for body 470 words, title separate. Let’s draft body. Body:Why Automate Churn Analysis?
Churn hurts revenue, but manual review of cancelled Stripe subscriptions is slow and error‑prone. By linking Stripe, PostHog, and Intercom (or a free alternative) you can feed each cancelled user’s recent activity into an AI model that writes a personalized win‑back draft for you to approve.
Data You Need
From Stripe: customer ID (`cus_xxx`), subscription cancellation event, and billing history (upgrade/downgrade flags).
From PostHog: the last 30 days of product activity for that customer – events, feature usage, and session length.
From Intercom (optional): the user’s name and email to personalize the outreach; if you skip Intercom, use SendGrid’s free tier and PostHog’s in‑app messages.
Real‑Time Pipeline Overview
The goal is a real‑time flow: when Stripe emits a `customer.subscription.deleted` webhook, a Zapier (or Make) workflow grabs the Stripe customer ID, queries PostHog’s API for the user’s activity, sends that payload to an AI endpoint (OpenAI or Replicate), receives a win‑back draft, and creates an Intercom conversation assigned to you with a note “AI generated – review before sending.”
Integration Approaches
No‑code / low‑code: Use Zapier’s Webhooks action to receive the Stripe event, a Code step (Python) to call PostHog’s `/api/projects//events/` endpoint, then another Code step to call the AI API, and finally Intercom’s “Create conversation” action.
API‑only: Deploy a tiny serverless function (AWS Lambda, Vercel, or Replit) that does the same steps; trigger it directly from Stripe’s webhook.
Key PostHog Endpoints
- `/api/projects//events/` – query events for a specific person (distinct_id) within a date range.
- `/api/projects//persons/` – retrieve person properties if you need extra traits.
Sample Code (Python)
import requests, os
def handler(event):
cust_id = event['customer']
# 1. Get PostHog person ID from Stripe metadata (you stored it earlier)
person_id = get_person_from_stripe(cust_id)
# 2. Pull last 30 days of activity
ph_resp = requests.post(
f"https://app.posthog.com/api/projects/{os.getenv('PH_PROJECT_ID')}/events/",
headers={"Authorization": f"Bearer {os.getenv('PH_API_KEY')}"},
json={"person_id": person_id, "since": "30 days ago"}
)
activity = ph_resp.json()['results']
# 3. Call AI endpoint
ai_resp = requests.post(
os.getenv('AI_ENDPOINT'),
json={"prompt": f"Create a win‑back message for a user who last used: {activity}"}
)
draft = ai_resp.json()['text']
# 4. Send to Intercom (or log)
return {"draft": draft, "person_id": person_id}
Cost Breakdown (Solo Budget)
– Stripe webhook: free
– PostHog: free tier up to 1M events/month
– Intercom: $79/month (or skip and use SendGrid free + PostHog in‑app messages → $0)
– AI endpoint (OpenAI/Replicate): ~$3/month for ~50 win‑back drafts
– Zapier: $20/month for the needed steps (or use Make’s free tier)
**Total:** ~$82/month with Intercom, or ~$8/month if you replace Intercom with SendGrid free + PostHog in‑app messages.
Example Win‑Back Draft (from e‑book)
Hey {{contact_name}}, I noticed you haven’t used the timeline export feature in 14 days. Here is a 30‑day free trial to try it again…
For a comprehensive guide with detailed workflows, templates, and additional strategies, see my e-book: AI for Micro SaaS Founders: How to Automate Churn Analysis and Personalized Win-back Campaign Drafts.
Now we need to count words. Let’s extract visible text. I’ll copy the visible paragraphs (excluding HTML tags and code). But code also contains words; they count as words? Usually yes, but they are inside tags, still visible. We'll count them.
Let's manually count.
I'll write the visible text in plain form:
Title line not counted? We'll ignore title for now.
Body:
"Why Automate Churn Analysis?"
"Churn hurts revenue, but manual review of cancelled Stripe subscriptions is slow and error‑prone. By linking Stripe, PostHog, and Intercom (or a free alternative) you can feed each cancelled user’s recent activity into an AI model that writes a personalized win‑back draft for you to approve."
"Data You Need"
"From Stripe: customer ID (`cus_xxx`), subscription cancellation event, and billing history (upgrade/downgrade flags)."
"From PostHog: the last 30 days of product activity for that customer – events, feature usage, and session length."
"From Intercom (optional): the user's name and email to personalize the outreach; if you skip Intercom, use SendGrid’s free tier and PostHog’s in‑app messages."
"Real‑Time Pipeline Overview"
"The goal is a real‑time flow: when Stripe emits a `customer.subscription.deleted` webhook, a Zapier (or Make) workflow grabs the Stripe customer ID, queries PostHog’s API for the user’s activity, sends that payload to an AI endpoint (OpenAI or Replicate), receives a win‑back draft, and creates an Intercom conversation assigned to you with a note “AI generated – review before sending.”"
"Integration Approaches"
"No‑code / low‑code: Use Zapier’s Webhooks action to receive the Stripe event, a Code step (Python) to call PostHog’s `/api/projects//events/` endpoint, then another Code step to call the AI API, and finally Intercom’s “Create conversation” action."
"API‑only: Deploy a tiny serverless function (AWS Lambda, Vercel, or Replit) that does the same steps; trigger it directly from Stripe’s webhook."
"Key PostHog Endpoints"
"- `/api/projects//events/` – query events for a specific person (distinct_id) within a date range."
"- `/api/projects//persons/` – retrieve person properties if you need extra traits."
"Sample Code (Python)"
Code block:
"import requests, os
def handler(event):
cust_id = event['customer']
# 1. Get PostHog person ID from Stripe metadata (you stored it earlier)
person_id = get_person_from_stripe(cust_id)
# 2. Pull last 30 days of activity
ph_resp = requests.post(
f"https://app.posthog.com/api/projects/{os.getenv('PH_PROJECT_ID')}/events/",
headers={"Authorization": f"Bearer {os.getenv('PH_API_KEY')}"},
json={"person_id": person_id, "since": "30 days ago"}
)
activity = ph_resp.json()['results']
# 3. Call AI endpoint
ai_resp = requests.post(
os.getenv('AI_ENDPOINT'),
json={"prompt": f"Create a win‑back message for a user who last used: {activity}"}
)
draft = ai_resp.json()['text']
# 4. Send to Intercom (or log)
return {"draft": draft, "person_id": person_id}"
"Cost Breakdown (Solo Budget)"
"- Stripe webhook: free"
"- PostHog: free tier up to 1M events/month"
"- Intercom: $79/month (or skip and use SendGrid free + PostHog in‑app messages → $0)"
"- AI endpoint (