The Competitor Job-Posting Signal Reference: What Each Role Proves

A competitor's open job postings are a window into where they are spending money and attention, but most of what you see through that window is distorted. This reference is for strategy, research, and talent-intelligence teams who read a rival's reqs to infer direction. It gives you a row per signal: what each posting element proves, how it lies, and how long the read holds before you must re-check.

The public answers are listicles that treat every posting as a strategic tell and bolt a tool pitch onto the end. They are wrong in a specific way: they skip the false-positive modes that make most single postings useless. The value here is the opposite. For each element, I tell you when not to trust it, and what second signal turns one posting into a read you can defend in a meeting.

Why a single posting proves almost nothing

A single posting is cheap, reversible, and frequently fake, so it carries far less signal than its specificity suggests. The strategic weight lives in the pattern: budget behind a cluster of reqs cannot be faked the way one aspirational listing can.

Start from the structure of how a hire happens. The sequence runs strategic decision, budget, headcount, posting, hire, execution, and finally announcement. By the time a competitor announces a product, practitioner sources say they started hiring for it 6 to 18 months earlier, though a tighter estimate puts that window at 4 to 8 months. Upstream of the posting itself, intent signals typically precede a public job ad by 20 to 30 days, while funding and leadership signals give 30 to 90 days of lead. None of these ranges is established by a regulator or peer-reviewed study. They are practitioner consensus, and you should treat them as order-of-magnitude, not precise.

The reason a single posting is weak is that so many postings are not real. Depending on what you count, somewhere between roughly one in seven and one in four live listings may not be genuinely fillable on their stated timeline. That is the noise floor you are reading against, and it is why cluster size, not role title, is the first thing to check.

The signal reference: what each posting element proves

Each job-posting element proves something specific about direction, and each has a documented way of lying. Jump to the row you need.

Read the cluster first. Engineering clusters point to product bets: three iOS engineers in a month means a mobile push, two security engineers means they are going upmarket into enterprise, and a "Founding Engineer" listing means a new product line. The location of a posting tells you about geographic intent: if a rival does not currently operate somewhere and starts posting there, suspect a new operation, expansion, or relocation, though contraction can produce stray postings too. Tech-stack requirements shed light on technologies a company plans to adopt, since specific mentions of a platform point at internal priorities, as long as the mention is not recruiter boilerplate lifted from a template.

Salary band deserves a caution. A title tells you what a competitor wants to build; the band tells you how serious they are. But the band is also the element most likely to mislead, because a high number can be a legal disclosure floor rather than a strategic bet. Convert it only by comparing against the same firm's prior postings for that title.

Reading cluster size: backfill, growth, or reorg

Cluster size is the fastest conversion from noise to signal, because budget behind multiple reqs is the thing a competitor cannot fake. One posting is a backfill; three in the same department is a growth signal; five is a reorg or expansion.

This is the single most useful heuristic in the reference. A single hire means almost nothing, because it is cheap and reversible. Patterns tell a story that is hard to fake, because each additional req represents committed headcount budget that a finance team signed off on. When you see a lone "Head of AI" posting and the listicle calls it an AI pivot, the sober read is backfill until a cluster or a second signal appears.

Exec hires follow the same logic one level up. A new VP of Engineering or CRO rarely arrives alone; new leaders rebuild teams within 90 days, so a single senior hire frequently signals three to five follow-on roles in the same function. If you see the exec and then the cluster, you have a build. If you see the exec and nothing follows inside 90 days, it was a replacement.

Funding is the other upstream amplifier. A Series B round produces 8 to 15 hires within 90 days, weighted roughly 40 percent engineering, 30 percent go-to-market, and 30 percent operations and finance. Knowing that distribution lets you predict the cluster shape before the reqs post, and flag any cluster that deviates from it as something other than a standard post-raise build.

Posting age: the shelf life of a signal

Age is the most load-bearing filter you have, because signal value decays on the same curve as hiring intent. Benchmark fill time sets the slope: genuine U.S. roles fill in roughly 42 to 44 days on average, with engineering and senior roles trending materially longer.

The practical rule is tighter than any single threshold: treat a posting as a live signal for roughly 30 days, then re-check. Because the average fill lands at 42 to 44 days, the probability that a listing is already dead climbs fast once it passes a month. A posting that has been live for 60-plus days is a substantial yellow flag; past 90 days you are in red territory; and a listing sitting at 120-plus days is almost certainly pipeline-building rather than a live req.

One structural trend may improve age reads over time. Several states are introducing legislation requiring employers to remove inactive postings or disclose future openings. Where those rules bite, stale-listing noise should fall, which makes age-based reads more reliable in regulated jurisdictions than elsewhere. This is worth re-checking locally, because the legal landscape moves.

How this read goes wrong

This is the section the listicles skip, and it is where the value concentrates. Below are the documented false-positive modes, each with the detection step that defuses it.

• Single posting read as strategy. One "Head of AI" taken as an AI pivot that is actually a backfill. Check: require a cluster of three or more, or a second signal, before calling direction.
• Stale listing read as live intent. A role filled months ago still displays. Check posting age against the 42 to 44 day fill benchmark, and confirm it on the company's own careers page. A role on LinkedIn but not on the company site has often already been filled or abandoned.
• Syndication duplicate inflates volume. One requisition appears as ten board listings and looks like a hiring surge. Check: de-duplicate to the careers-page requisition before counting.
• Ghost or pipeline posting mistaken for expansion. Placeholder language without concrete requirements is a pipeline-building tell. Check for repeated reposts under one internal job ID.
• Reposted date masks true age. Every repost resets the visible date. Check feed history or the original site date, not the aggregator timestamp.
• Salary band over-read. A high band can reflect a pay-transparency legal minimum or a hard-to-fill role, not strategic priority. Check against the same title's prior postings at that firm.
• Exec hire read without follow-on. A VP hire with no subsequent req cluster inside 90 days is a replacement, not a build. Check for the three to five follow-on roles.

The deepest trap is the stale-versus-ghost distinction, because the two look identical in a scrape but mean different things. A stale listing was genuine when posted but has since been filled, paused, or canceled and not removed. A ghost job was never genuinely intended to be filled on that timeline. The tell that separates them is the requisition number: a real role that fills and reopens gets a new req number, while a listing quietly reposted under the same internal job ID multiple times is pipeline-building.

The ghost-job noise floor: why the numbers disagree

The same phenomenon supports five different "proportion of ghost jobs" figures because the unit counted changes. Posting-level data lands near one in five; company-level and recruiter-level self-report runs far higher because one bad listing flags a whole company.

Read this table defensively. The posting-level figures cluster near one in five, and that is the rate you apply when you discount a scrape. The company-level number of 40 percent is the share of companies that posted at least one fake listing in the past year, not the share of postings that are fake. A company that posts one questionable listing among fifty still counts toward that 40 percent. A writer who cites 40 percent as a per-listing rate overstates the problem by roughly double, so when you quote a prevalence figure, quote the unit with it.

Two cross-checks anchor the floor. The Congressional Research Service has defined ghost jobs as online postings for positions that do not exist or that employers are not planning to fill immediately. And the share of employers keeping postings active for more than 30 days fell from 68 percent in 2022 to about 33 percent in 2025, which suggests the stale-listing problem is shrinking even as awareness of it grows.

When you need to go from a hunch to a named list of competitors already hiring in a cluster, plain-English search collapses the de-dupe and count work. Refolk lets you ask for the pattern directly rather than scraping boards and reconciling duplicates by hand, which is exactly the friction the first three procedure steps describe.

The procedure: from postings to a dated read

Run every competitor through the same eight steps so your reads are comparable and each one carries an expiry. The order matters: baseline before you measure spikes, de-dupe before you count, and corroborate before you ship.

Sources disagree on two points of order, and both are defensible. Some put velocity-baselining after capture rather than before, which works if you backfill the baseline from history. Recruiter-focused sources put funding and exec signals ahead of postings entirely, treating the posting as confirmation of a move you already detected. Pick one and apply it consistently so your reads stay comparable across competitors.

Corroboration: the second signal that makes a read

No read ships on one posting. Upgrade a single posting to a conclusion only when a second, independent signal points the same way, because the second signal is what defeats the ghost-job and backfill noise floors.

The documented converters fall into three families. The first is product-surface changes: pricing-page updates, product-page edits, new docs, or review-sentiment shifts that align with the role cluster. The second is funding, which both predicts the cluster shape and dates it, since a Series B round produces its 8 to 15 hires within 90 days on a known 40/30/30 split. The third is exec hires, where a new VP or CRO precedes three to five follow-on roles in the same function within 90 days.

The bottom-left quadrant, one posting and no corroboration, is where most listicle reads live and where most mistakes are made. Keep those entries in your log, but never put them in a brief.

Who actually does this work, and keeping the read current

This read is a thin, defensible capability because very few people hold the title that does it well. In Refolk's index, 365 U.S. professionals currently hold a Competitive Intelligence or Market Intelligence Analyst title, against 69 in the UK - a 5.3x gap that tells you where this capability concentrates.

The skill is more widely held than the title: 4,473 U.S. professionals list a competitive-intelligence skill at the Senior band and 9,453 at the Director band. Treat those as supply of the skill, not of the role, because they include people whose current title is not CI-specific. The practical point stands either way: doing this read rigorously, with age filters and corroboration rather than listicle shortcuts, is a capability that sits with few people, which is why it holds up as an advantage.

Keeping a read current is mechanical once the expiry dates are set. Every read carries a 30-day re-verify, and the re-verify is cheap: confirm the cluster is still live on the careers page, check whether the second signal still holds, and age any posting that crossed a threshold. The one moving part worth re-checking by hand is the legal landscape, since laws requiring removal of inactive postings will change the stale-listing noise floor in the states that pass them.