What We Learned Building an Honest Software Directory

We Did Not Set Out to Build a Directory

We set out to solve a problem we kept hitting ourselves. Every time we launched a product, we faced the same wall: where do you list it so people actually find it?

Most directories felt like graveyards. Some felt like scams. Almost all of them hid how they worked, slapped fake 9.2/10 scores on everything, and stuffed their pages with barely-readable comparison tables built to harvest ad clicks.

So we built directree differently. The pitch was simple: paste a URL, get a structured listing in about 30 seconds, and every field is labelled as either Observed, AI-inferred, or Founder-edited. No fake scores. No pay-to-rank. Every claimed listing gets a real do-follow backlink.

That was the idea. Here is what actually happened.

Lesson 1: Honesty Requires Infrastructure

Saying "we're honest" is cheap. Proving it takes work.

We had to build a three-tier provenance system before a single listing went live. Observed facts come from crawling - pricing pages, free trial buttons, integration lists, social links. AI-inferred fields - summaries, best-for tags, strengths and weaknesses - are always labelled, never disguised as fact. Founder-edited fields are the only ones that carry a gold star, because they were verified by someone who actually owns the product.

This sounds like a design choice. It is actually a database schema, an enrichment pipeline, a review queue, and a UI pattern that repeats on every single card and page. The honesty model explained took its own blog post because the details matter.

The lesson: if your differentiator is "we don't lie," you need systems that make lying structurally impossible. Trust is not a tagline. It is a constraint you code against.

Lesson 2: Fake Precision Is Everywhere

Within the first week of researching competitors, we found a directory that gave a project-management tool 8.7/10 for "docs" and 7.3/10 for "collaboration." The decimal looked scientific. It was completely made up.

We wrote an entire post on why we refuse to show fake precision scores. The short version: composite ratings with decimal precision signal expertise they do not possess. They are either random, bought, or reverse-engineered to keep advertisers happy. All three are dishonest.

Instead, we use qualitative bands. A tool is either "Great for," "Good for," or "Not ideal for" a given use case. That is less impressive at a glance. It is also accurate, which is the entire point.

Lesson 3: Automation Breaks in Embarrassing Ways

We built an automated pipeline that pulls new launches from Product Hunt, enriches them, and turns them into directree listings. Smart, right?

For three days straight, the daily showcase feature kept picking the same tool. Every. Single. Day. The bug was tiny: exactly one user had qualified as a "top contributor," and their only verified tool was the only item in the pool. The randomness had nothing to randomize over. We fixed it with a minimum pool guard and wrote the rule down: if the verified pool is smaller than three, fall back to a broader random draw.

Then there was the Product Hunt vote problem. Our ingest script was accidentally copying Product Hunt's vote counts into directree's own community signal. For a few hours, our homepage showed PH popularity as if it were our own. We caught it, reset the data, and added a hard separation: PH popularity is ingest-only. Our community signal comes from our community.

These are not heroic war stories. They are the kind of dumb mistakes you make when you ship fast. The build-in-public part is admitting them.

Lesson 4: Content at Scale Without Quality Is Just Spam

We launched the directree blog with ambitious plans: daily posts, SEO automation, the whole machine. Within a week, we realized we were writing filler. The posts were correct. They were also interchangeable with anything else on the internet.

So we cut the cadence from seven posts a week to three. We added a spam-update quality gate: every post must contain either a named source with a real figure, an original observation from our own experience, or a specific concrete how-to. No generic listicles. No rephrased search results.

The result? Twenty-seven posts, all indexed on Google, climbing from zero to over a thousand impressions in under three weeks. One click so far. That one click is real, and it came from a post that said something specific.

Lesson 5: The Backlink Promise Actually Matters

Every claimed listing on directree gets a do-follow backlink. We say this everywhere. We mean it.

Most directories give you no-follow or nothing at all. Some charge for the do-follow upgrade. We decided to make it free and automatic because the founders who need backlinks most are the ones with the smallest budgets.

The catch? You have to claim your listing. We verify domain ownership. Once verified, the badge goes live and the link passes authority. It is a real backlink from a real directory, and we explain the mechanics in our guide to do-follow vs no-follow backlinks.

Lesson 6: Build in Public Means Showing the Cracks

We shipped an iOS app and submitted it to the App Store with a known flaw: account deletion happens on the web, not natively in the app. Apple might reject it. If they do, we will fix it, resubmit, and explain what changed.

We pivoted an entire product feature - an SEO "build pack" that was going to generate styled landing pages - into a content strategy product after realizing that customers wanted recommendations, not another design tool. The old spec is archived. The new one is live.

None of this is polished. It is honest. That is the whole brand.

FAQ

Q: What does "build in public" actually mean? A: It means sharing what you are building, what is working, and what is breaking - while you are still building it. Not the polished case study after the fact. The messy middle.

Q: Is directree really free? A: Yes. Anyone with an account can submit a tool and get a listing. Claiming your listing - which gives you the do-follow backlink badge and editing access - is also free. We monetize through optional featured placements and an SEO content build pack, not by paywalling basic directory access.

Q: How do you prevent spam listings? A: Every submitted URL goes through enrichment and a basic quality check. Low-traffic or broken sites get flagged. The community can report listings. And we do not let unverified founders edit fields that are marked as Observed.

Q: What is the difference between Observed and AI-inferred? A: Observed means we crawled it and it is verifiable - pricing, free trial, integrations. AI-inferred means our model made a reasoned guess - best-for tags, strengths, a summary. The label is always visible. You can read the full breakdown in our honesty model guide.

Q: Does a do-follow backlink from directree help my SEO? A: A single backlink will not catapult you to page one. But it is a real, contextual link from a relevant directory, and it counts. The real value compounds when you combine it with other honest listings, good content, and structured data. We explain the full picture in our backlinks guide.