Why I'm Sunsetting Papertrail
In June I wrote that building the product is the easy part. I still believe that. What I didn’t expect was the next lesson: you can build the product, start on the marketing, and still have the reason for the product disappear.
I’m sunsetting Papertrail. I hope to open source the code so anyone who wants to keep using it can run it themselves. Here’s why I’m stopping, and what I’m taking with me.
What it was
I built Papertrail after a document expiry caught me off guard. There were reminder apps and scanner apps, but nothing that owned the whole problem. I wanted something that would track the document, know when it expires, walk you through renewing it, and then get out of your way.
You’d scan your passport, insurance, licence or visa once. Papertrail pulled out the expiry date, watched it, and reminded you by push, email or calendar before it lapsed. When renewal time came, it gave you a step-by-step checklist for your document type and your country.
It launched in March 2026 at $3.99 a month. It reached ten active users with zero marketing spend. That’s small, but they were ten real people.
Why I’m stopping
There are three reasons.
Agents made it unnecessary. This year, the problem Papertrail solved got much easier to solve without Papertrail. People now have AI agents that read their inbox, spot a renewal notice, put the date on their calendar and nudge them when it gets close. They’re already paying for those tools. Once a product becomes something your existing assistant does in passing, $3.99 a month stops being an easy yes. I couldn’t make a convincing case for why someone should pay for a dedicated app. If I can’t make that case, I shouldn’t be asking people for money.
People didn’t trust it with their documents. Papertrail asked people to upload their passport, their visa and their driver’s licence to a product run by one person. Those are the most sensitive documents they own. I built it carefully, with time-limited share links, watermarking, revoke controls and an SSRF guard on the email pipeline. But conversations kept coming back to the same question: why would I trust you with this? It’s a fair question. The security you build and the trust people feel are two different things. A solo founder with ten users doesn’t have the track record to close that gap.
It cost more to run than it needed to. Nearly every infrastructure decision I made started at zero cost. Supabase, Resend, OneSignal and Sentry all have free tiers that would have carried Papertrail a long way. Hosting was the exception. I put it on Vercel’s $20-a-month plan from the start, before the product had earned it. At $3.99 a subscription, I needed five paying users just to cover hosting, before counting anything else. That was poor judgement on my part. It’s a small number, but small numbers matter when revenue is small too. A fixed monthly cost turns “let’s see if this grows” into “this has to grow.”
The awkward part is that I used it a lot. It’s genuinely useful to me. But a product that’s useful to the person who built it isn’t a business.
What I’m proud of
Some parts of Papertrail I’d happily build again.
The Drop. Premium users got a personal forwarding address. You forwarded any document-related email to it, and Papertrail read it, pulled out what mattered and asked you to confirm. There was no typing and no manual entry. Getting it right meant parsing attachments, handling .eml threads, and retrying with backoff because Resend’s webhook sometimes arrived before its API was ready. When it worked, it felt like magic, which was the goal.
Handoff. You’d be logged in on your laptop, scan a QR code with your phone, photograph the physical document, and it showed up on your desktop. It was a small flow with a lot of session-linking work behind it.
Country-aware renewals. Renewing a Nigerian driver’s licence is nothing like renewing a UK passport. Papertrail knew that, and its checklists reflected it. Most tools treat “renew document” as one generic task. It isn’t.
Retroly, again
With Retroly, the product worked. We converted almost 60% of signups with no marketing budget. Then lockdown ended and the growth we’d borrowed from that moment went with it. The lesson I took was to figure out marketing first.
With Papertrail I tried to do exactly that. I offered a free trial, posted on LinkedIn and wrote in public. This time it wasn’t really a distribution problem. Retroly lost its moment. Papertrail lost its reason to exist.
So I’m adding two questions to the one I learnt last time. Will people still need this in a year, and will they trust me with it? For a product built on sensitive data, trust isn’t a feature you ship. It is the product.
There’s a smaller, practical lesson too. Stay on free tiers until revenue tells you to leave them. Every fixed cost you take on early shortens how long you can wait for the answer.
In June I wrote that I couldn’t wait for my $40 moment, when someone would trust Papertrail enough to pay for a year upfront. Looking back, that line was about trust all along.
Open-sourcing it
I hope to open source the full codebase. I’m not ready to do that yet, so I can’t promise a date, but it’s the outcome I’m aiming for. It’s built on Next.js, Supabase, Resend, OneSignal and Tailwind. Self-hosting would also answer the trust question: your documents would stay on your own infrastructure, with nobody else in the middle. And if you host it yourself, you can do what I should have done and start on a free tier.
What’s next
Something fun. I’m working on Sidequest, an app for missions in real life. You pick a quest, do it, and keep the story, either as a private memory or as something you share with friends.
It’s about as far from document renewals as I could get. I’m taking the questions from Papertrail with me, though: will people still need this in a year, and is it something people will use without me having to nag them?
Sunsetting something isn’t failure. It’s what happens when you take the evidence seriously. Papertrail taught me more about trust, timing and what people will actually pay for than any product that “worked.” And I’ll probably keep running my own copy.
Thanks to everyone who used it, gave feedback, or asked me the hard questions.