Job listings are a genre of fiction with strict conventions. Once you learn the conventions, you can read a posting in ninety seconds and know whether it deserves twenty minutes of your life. We process thousands of listings a day through our normalization pipeline, which means we've seen every trick, template and tell in the format. This guide is the decoder ring.
Start with what's missing, not what's there
The single most informative thing about a listing is usually an absence. In our dataset, only about one in five tech listings discloses any salary figure. That's not because the other four companies forgot — pay transparency is a choice, and choosing against it tells you the company either pays below the market band for the title, wants maximum room to anchor low in negotiation, or has internal pay inconsistencies it would rather not advertise. None of those are dealbreakers, but all of them are information.
The same logic applies to the tech stack. A backend role that names its stack ("Python, Django, Postgres, deployed on AWS with Terraform") is written by someone close to the actual work. A listing that says "modern cloud technologies" was written by someone at least two layers away from the codebase, and the interview process will usually have the same flavor.
Seniority labels are marketing, not measurement
Titles inflate for the same reason currencies do: issuing more of them is free. "Senior" in a fifty-person startup can mean three years of experience and ownership of half the product; "Senior" at an enterprise can mean a decade and a promotion committee. When our classifier assigns seniority levels, it reads the title and the description, because the two frequently disagree — a "Senior Engineer" posting that describes shipping tickets under close supervision is a mid-level role wearing a costume.
The reliable signal isn't the label; it's the verbs. Own, design, decide, mentor, define describe senior work. Implement, support, assist, contribute describe earlier-career work. Read the responsibilities paragraph and strike out every adjective — what's left is the actual job.
The requirements list is a wish list with an exchange rate
The old advice — treat requirements as negotiable — is true but incomplete. The useful skill is knowing the exchange rate. In practice, a listing's requirements section divides into three tiers, almost never labeled:
- Load-bearing requirements — the two or three technologies that appear in the title, the first paragraph, and the requirements. If the posting says "Rust engineer" three times, no amount of adjacent experience substitutes for Rust.
- Ecosystem filler — the surrounding tools (Git, CI/CD, "agile environments") that every working engineer already has. These exist to make the list look rigorous. Ignore them.
- The wish sub-list — "nice to have", "bonus points", and anything with a version number more specific than the team actually runs. Meeting 30% of these is normal for the person they eventually hire.
A useful rule from watching listings age in our index: postings that demand everything tend to stay open longest. A listing on its second or third month (check the posted date — we show it on every listing) is a listing whose requirements have quietly become negotiable.
Salary ranges: read the width, not just the numbers
When a range is present, its width is a message. A tight range ($130k–$145k) means the company has leveled the role precisely and negotiation will move small amounts. A cavernous range ($90k–$180k) means one of two things: the listing is really two or three roles wearing one title ("we'll level you after interviews"), or the range was published to satisfy a transparency law with minimum information content. Either way, expect the real conversation to start near the bottom third — which is exactly why you should have market data of your own before that conversation. (We wrote a separate guide on that.)
Also check the period and currency before you get excited. Our salary parser exists because "$8,000" can mean a month in Dubai, "€65.000" uses a dot as a thousands separator, and "£400" is a contractor day rate. Human eyes make the same mistakes the naive parsers do.
Quiet red flags
- "Fast-paced environment" + "wear many hats" + no engineering headcount mentioned — understaffed, and planning to stay that way.
- "Competitive salary" with no number — competitive with whom is left as an exercise for the reader.
- A reposted listing with a fresh date — sometimes genuine growth; often churn. If you saw the same role twelve weeks ago, ask in the interview what happened.
- Five interview stages for a mid-level role — process weight correlates with decision-making weight everywhere else in the company.
- "We're like a family" — families don't have performance improvement plans.
A ninety-second reading protocol
- Posted date first — is this listing fresh, or fossilized?
- Salary present? Note the width of the range.
- Find the load-bearing requirements (title + first paragraph + requirements overlap).
- Read the responsibility verbs; ignore the adjectives.
- Check the work model — and whether "remote" comes with an asterisk like "3 days in office". Our remote-work guide covers how often it does.
If a listing survives all five checks, it has earned the twenty minutes for a tailored application. Most won't — which is the point. Your time is the scarcest resource in a job search; spend it like the salary it eventually becomes.