How I research

Every article here is written by the person who did the research, and the wording tells you what kind of work sits behind each claim:

• “I compared”, “I went through”, “I read” mean I read the sources, and the claim links to them.
• “I ran” means a logged check, and the article links to its raw data. You won’t see it until those checks exist.

Where the facts come from #

I work down this list, and I only move to the next source when the one above can’t answer the question:

1. Official pricing pages, docs, changelogs and status pages
2. GitHub: code, issues and releases
3. Independent benchmarks and published studies
4. Review platforms such as G2, Capterra and TrustRadius
5. Developer communities: Reddit, Hacker News, Stack Overflow and public Discord logs
6. Other blogs, as a last resort and never as the only source

Comparisons and roundups cite at least 20 sources, and explainers at least 10. Every non-obvious claim links to its source.

When sources disagree, I say so and explain which one I trust and why. When I can’t verify something, I say that too, instead of repeating the vendor’s claim as fact.

Prices, limits and dates #

Every price, limit and spec is checked against the vendor’s own page and carries a “verified” date. I re-check pricing at least every 90 days, and straight away when a vendor announces a change.

The “updated” date on an article only moves for a real change, such as new pricing, a correction or a new section. It never moves for a cosmetic edit.

What the “How I researched this” box means #

Every comparison has a short box near the top called “How I researched this”. It lists:

• How many sources the article rests on, counted by type (official docs, GitHub, community threads and so on)
• The date range of the research
• What I couldn’t verify
• A link back to this page

If the box says I couldn’t verify something, treat that part of the article with extra care.

Small-sample checks #

Reading sources only gets you so far. So the plan is to add light checks that anyone could repeat, under strict rules:

• Free tiers or trial credits only. Checks are scripted, with fixed settings written down before the run.
• Only our own targets. They point at test pages I host with known correct answers, or at sandboxes built for scraping practice, such as toscrape.com. Never at a third-party site that forbids it.
• Terms first. Before a vendor is checked, I read its terms for any clause about publishing performance results, and I log what I found.
• Everything logged. Each run keeps the script version, the raw results, screenshots and a short summary. Articles link to a public page with that data, failed runs included.
• Same setup for everyone. Every vendor in a round gets the same pages, settings and time window. If a bug on my side breaks one vendor’s run, I re-run all of them.
• Called what they are. These are “small-sample” checks, never benchmarks. A few dozen requests against test pages can’t predict what you’ll see at your volume or on the sites you care about.

If a vendor disputes a result, I re-run the same setup and publish both runs with a dated note.

Vendors whose terms restrict publishing results #

Some vendors’ terms don’t allow customers to publish performance results without their written consent. I respect that. Those vendors are covered from their docs, pricing pages and other public sources only, and the article says so plainly.

That isn’t a mark against them. It only means you won’t see my numbers for that vendor.

What I never claim #

• That I used a product, contacted support or saw a dashboard, unless it’s a logged check
• Quotes, statistics, customers or credentials that I can’t link to
• A vendor’s marketing claim as fact without saying where it came from

AI assistance and money #

I use AI tools to help with research and drafting, and every claim is checked against its live source. A named person reviews each article; for the first articles of this soft launch, that review is still in progress and each one says so. The editorial policy explains the rules and how corrections work.

Money never changes the research. The disclosure explains how affiliate links and sponsored placements work.