September Update: €139 MRR (+€31) - Highest new subscribers but also highest churn
What happened this month
September was my best month for new subscribers (7) but also my worst for churn (3). Deployed programmatic SEO targeting 5 professional segments. Starting to see patterns in user behavior that challenge my subscription model.
Metrics
- MRR: €108 → €139 (+€31 / +29%)
- New paying customers: +7 (highest ever)
- Churn: -3 (highest ever, 2 cancelled immediately after subscribing)
- Net growth: +4 customers
- Signups: 62 total, 52 linked (84% link rate)
- Traffic sources:
- 27 Google (52%)
- 10 AI search engines (Perplexity/ChatGPT - 19%)
- 5 referrals (10%)
- 3 DuckDuckGo (6%)
- 1 Heise article (2%)
- Language split: 32 German, 16 English, 2 Dutch, 1 French, 1 Spanish
- Conversion by language: English 12.5% vs German 6.25%
What worked
Automated feedback emails are paying off
Started getting meaningful user feedback through automated emails. Discovered two concrete use cases:
- Personal relationships (couples with different communication preferences)
- Shopping list memory tool from long voice messages
Also got a feature suggestion for Signal group transcription that multiple users seem interested in.
AI search engines bringing premium users
Got a cybersecurity consultant (founder at a German agency) via Perplexity. AI search engines (Perplexity + ChatGPT) now represent 19% of traffic and seem to attract higher-quality profiles.
Email routing validated (sort of)
One German power user actively using email routing feature. However, no one else has adopted it yet despite promoting it. Unclear if it's a niche feature or if discovery/onboarding needs work.
Programmatic SEO deployed
Launched 5 profession-specific landing pages (IT managers, office managers, remote workers, consultants, content creators) in 4 languages. Too early to see impact (only 4 days live), but hope it will bring more qualified traffic.
What didn't work
High link rate ≠ more subscriptions
Despite maintaining 84% link rate, it hasn't translated into proportionally more paid conversions. 7 new subscribers is my best month, but it's still modest growth for 52 signups. I suspect the users who weren't linking before were occasional users anyway, not power users - so improving link rate just brought in more of the wrong profile.
Subscription model doesn't fit occasional users
2 out of 3 churns were immediate cancellations - users subscribed, used the service, then cancelled to avoid renewal. They have real needs but sporadic usage patterns. Current pricing doesn't capture this segment.
Not attracting power users
Most users have light, occasional usage. Need to figure out how to attract profiles that will use the bot intensively and get consistent value from monthly subscriptions.
Email routing adoption is unclear
Only 1 user actively using this premium feature. Either it's too niche, poorly communicated, or the UX needs rethinking. This was supposed to be a premium differentiator.
Reality check
My goal of €500 MRR by end of December 2025 is looking difficult. At +€31/month growth, I'd need 11+ months to reach €500. The programmatic SEO needs to significantly accelerate this, or I need to reconsider the timeline.
The subscription model might be holding me back. I'm losing users who have legitimate needs but can't justify €5.99-19.99/month for sporadic usage.
What's next
October priorities:
- Monitor programmatic SEO impact (first full month of data)
- Consider Signal group feature (bot as member to auto-transcribe) based on user feedback
- Explore pay-per-use or credit system for occasional users
- Start exploring paid advertising (Facebook/Google Ads) if SEO shows promising signals
- Focus on attracting power user profiles who can justify premium plans
Open question: Should I pivot pricing strategy before scaling marketing, or scale first and let the market tell me what works?
Ask
For those who've dealt with the occasional-user problem: Did you add pay-per-use alongside subscriptions, or did you find ways to convert occasional users to regular ones? How did you balance the two?
Hey Guillaume,
Thanks for sharing the details so openly. What stood out to me was your point about churn from “occasional users” who clearly have a real need but can’t justify a monthly subscription. That pattern comes up a lot in AI tooling people dip in for one big task, get value, then cancel.
I’m an AI researcher working on ways to make custom models more accessible to solo founders and small teams without a data science background. Reading your update, I’m curious: for the folks who cancel after a quick burst of usage, do you think a pay-per-use or credits system would actually retain them, or are they just not your target power-user segment?
Also, if you’re open to it, I’d love to learn a bit more about how you’re seeing user behaviour diverge from the subscription model. Happy to share some ideas on lightweight ways to test alternative pricing or usage models.
– Ulk
Hey Ulk,
Thanks for your comment !
I'm honestly still figuring this out.
Here's what I'm seeing:
The occasional user pattern:
My hypothesis on pay-per-use/credits:
Arguments FOR:
Arguments AGAINST:
I don't have enough data yet to know which hypothesis is right.
Current priority:
I just deployed programmatic SEO targeting 5 professional segments. Once I see what profile converts best from that traffic, I'll have better data on whether pay-per-use is worth building vs just attracting better-fit customers.
Would love to hear your thoughts on lightweight ways to test alternative pricing or usage models !
Thanks!
Guillaume
Hey Guillaume,
That makes sense. For lightweight tests, a simple "fake door" button for credits or a quick exit survey for churned users could give you some immediate signals.
Thinking about the data you already have, though, you could probably predict which new sign-ups will be "burst users" vs. potential "power users" based on their source and first few actions.
If you're open to it, I'd be happy to build a quick predictive model for you. All I'd need is an anonymised CSV of your user activity. It'd be a great real-world test for my research, and in return, you'd get a concrete way to segment new users for different offers.
Let me know if that sounds interesting.
Best,
Ulk