Sumwise
AI Summarizer for Web, PDF & YouTube
Numbers since last update:
- 138 cumulative Chrome installs (~130 net), ~20 WAU — flat. Classic leaky bucket.
- 27 countries. Singapore is #1 and I genuinely don't know why.
- Biggest funnel leak: ~130 installs → 18 signups. My auth wall eats 86% of installs.
- First fully organic funnel completion: a user in Chile went install → signup → follow → first digest, zero founder touch.
- First returning user: Bolivia. Text summarizer, Spanish output, hit the free daily cap in 12 minutes. I emailed them exactly one question: the day you hit the limit, what did you do next?
The part I'm actually proud of: before building anything this sprint, I wrote a tripwire. If by Sep 10 there are (a) zero organic follows from the two new surfaces AND (b) zero new week-2 returns, I formally open Plan B. Criteria locked before the work — so I can't negotiate with myself later.
Building the machine is the easy part. Finding out on a schedule whether anyone wants it — that's the discipline.
Day 10 update: 119 visitors, 4 signups. Since then: ~250 visitors, 12 signups, ~5% conversion. But the real problem was underneath — half of signups never ran a summary, and nobody was around by day 7.
So last week went into one thing: you can now follow YouTube channels and get new videos summarized weekly, in your language, in your inbox. New-account onboarding is now "pick 2 channels" instead of an empty box.
Best moment: posted the update at 9pm. By morning, a user in Turkey had followed 2 channels and woken up to their first digest — in Turkish. The loop closed while I slept.
Honest state: $0 revenue, retention experiment is one day old. Three weeks will tell — I'll post the numbers either way.
-
1
This is a good example of how early usage data can reveal a different problem than the one you initially expected.
I’m curious — after seeing that many signups didn’t reach the first summary, what surprised you most about where users were dropping off?
-
1
Honestly? That the biggest drop wasn't motivation — it was a bug. We served both www and apex, and localStorage is per-origin, so people who signed in on one variant looked signed out on the other. They didn't churn; they assumed the product was broken.
Runner-up surprise: even activated users never came back — D7 was literally zero. That's why we shipped weekly channel digests. Too early for numbers; the next two weeks will tell.
Ever seen churn that turned out to be a bug?
-
1
That’s a useful distinction. I’ve definitely seen cases where what looks like churn is actually a product or onboarding issue. The D7 result is especially interesting, though.
-
1
Thanks! What made D7 click for us: the people who showed up actually used it — ~3 summaries each — then never came back. A summarizer is a pull tool; the need doesn't remind you of the product. So the digest is us bolting a push loop onto it: follow your channels once, get a weekly reason to return. That's the bet on the table — if it doesn't move week-2 return, more funnel polish won't save it.
-
-
-
Launched Sumwise 10 days ago — one-click AI summaries of articles, YouTube videos and PDFs, in 7 languages. I read in English but think in my native language; built the tool I wanted.
First 3 days converted at ~1%. The data said people left without seeing value, so I rebuilt: 3x faster summaries, a no-signup demo on the landing, and a hero about the real differentiator — summaries in your language.
Conversion's trending toward 3–5% since. Tiny numbers, not calling it yet.
Lesson so far: analytics before marketing. Everything I shipped came out of the data, not my roadmap.
-
2
Summarizers are crowded, so differentiation usually comes from context or workflow.
What’s the main reason users choose this over alternatives?
-
1
Good question. Our differentiation is the reader, not the summary: most tools assume you think in English — we assume you don’t. One click, structured summary, in the language you actually think in. Our first signups so far: India, Switzerland, Burkina Faso. Non-English-first isn’t a niche — it’s most of the internet.
-
1
Your Swiss signup is a researcher who reads English fine but prefers German for dense papers — and will pay $20/mo for precision. Your Indian signup is a professional who processes English fine but is 40% faster in Hindi — and will share your tool in 12 WhatsApp groups if it's free. Your Burkinabé signup is a student who struggles with English entirely — and needs French academic vocabulary, not just translation.
The risk: you build for the Swiss researcher (high willingness to pay, narrow use case) while your landing page attracts the Indian professional (viral loop, low willingness to pay) and your product frustrates the Burkinabé student (needs completely different language support).
Quick test: Track activation by country + language selected. My guess: one combo drives 80% of your 4 signups, and the other two are false positives who'll churn in 7 days.
I run PreLaunch AI — we simulate buyer segments for language/region products to find which market justifies premium pricing vs. which needs freemium virality. Happy to model your India/Switzerland/Burkina Faso split if you want to know which country to double down on before you localize. No strings.
-
-
-
1
What stood out to me is that you treated early traffic as a way to learn where value becomes visible, not as a verdict on the product.
It's easy to chase more visitors, but sometimes the bigger leverage comes from making the first moment of understanding happen sooner.
1 Comment
The first organic completion and returning user seem more informative than the 138 installs.
Curious which signal you’ll trust most when deciding whether Plan B is actually necessary.