Five sources. A dozen we didn't pick. Here's why.

The constraint that saved us

Launch day was locked. Non-negotiable. We had customers waiting, a Slack integration written, a classifier trained on thousands of labelled examples. Adding a new source meant retraining the classifier, testing the ingestion pipeline, building the UI layer, handling edge cases when the API goes down. Two weeks in, I realised we had a choice: launch a solid product covering five sources well, or launch something that tried to cover thirty sources and launched broken.

That's when we stopped asking what sources "would be nice" and started asking what sources actually matter to the people we're building for. Mobile app studios. SaaS founders. Brand managers trying to track what the world is saying about their products. Not TikTok influencers. Not private Discord servers. Not email inboxes.

The five sources we landed on, App Store reviews, Google Play reviews, Twitter slash X, Reddit, and Google News, share something crucial: they're public, searchable, and persistent. Someone writes a one-star review or posts a bug report; it stays there. We can monitor it. We can correlate it with other signals. We can route the noise to the right team.

Why we binned the obvious ones

Instagram came up. Of course it did. Our first instinct was yes. But then we asked: how many of our customers' app users mention them on Instagram? Not repost screenshots. Not tag the brand in a Story. Actually discuss the app itself, the bug, the feature request. For most mobile studios, that number is near zero. Instagram is visual and ephemeral. By the time we've ingested a post, it's been buried under fifty other posts.

YouTube comments were another. The thinking was obvious: if someone makes a video about your app, the comments are gold. True. But we'd also need to handle YouTube's API limits, deal with spam moderation at scale, and somehow make sense of thousands of comments that had nothing to do with feedback. We decided to start without it. If five of our customers ask for it in month two, we'll reconsider. So far, none have.

Telegram, Discord, Slack communities. These were the hardest to cut. They're where real users congregate. They're where product people listen. But they're also private by default. You can't monitor a private community you don't own without explicit permission. We'd rather build trust and let customers opt-in to private monitoring later than scrape private spaces from day one. That's not who we are.

Product Hunt, Hacker News, Trustpilot. These matter for some founders, sure. But they're not where most app users go to report bugs or ask for features. App Store reviews are. Google Play reviews are. Reddit is where technical communities congregate. That's what we optimised for.

What sealed it: the classifier

Here's the thing nobody tells you about ingestion: it's only useful if you can make sense of it. We built a classifier that tags every signal as bug_report, feature_request, pr_crisis, positive_feedback, or noise. That classifier was trained on thousands of examples across our five sources. Each source has its own linguistic texture, its own vernacular. Reddit sounds nothing like an App Store review. Twitter replies to your brand are structured differently from Google News articles mentioning your company.

The moment we started thinking about that classifier, the choice became clear. Five sources, deeply understood. The classifier gets better. The routing gets smarter. Our customers get signal, not noise.

Twitter slipped in late, by the way. We almost cut it. The API setup is manual; our customers have to bring their own bearer token. It's a friction point. But the moment we started fielding questions from SaaS founders, we understood why it mattered. Real-time mentions of your brand, your competitor, your feature. It's not review feedback, but it's signal. It moves faster than anything else we ingest. And it correlates beautifully with spikes on the App Store. A tweet thread goes viral. Your app gets reviewed. Our correlation engine connects the dots.

The version two conversation

Launch is not end-game. It's the beginning of a conversation with people actually using the product. Two weeks in, someone asked for TikTok. Someone else asked for Product Hunt. Both sensible asks. But the answer isn't yes yet. The answer is: show us the pattern. If five product studios ask for TikTok monitoring, we'll prioritise it. If it's one founder with a specific use case, we'll help you find a workaround.

That's not us being stubborn. It's us being honest about what we can maintain and what we can make good. Every source we add is a source we're responsible for. If Twitter's API breaks, we have a problem. If Google Play's HTML structure changes, we have a problem. We'd rather own five sources completely than own fifteen sources badly.

What this means for you

If you're a mobile studio with apps on the App Stores and a presence on social, Monitr does what you need. If your users live in private Discord communities and you want to monitor every message, we're not the product for that. Not yet. We're honest about that because the moment we pretended we did everything, we'd do nothing well.

The sources we picked cover the public places where app feedback actually happens. The classifier we built learns the tone and structure of each one. The routing engine we deployed gets smarter every week as customers tune their rules. That's the product we wanted to build, and that's why we cut the rest.