About Longbow Research
We read the job postings employers publish, at scale, and tell job seekers what they actually say.
This page is the reference behind every report we publish: where the data comes from, how we count, and what we will never publish. Every report links here, and every change to these rules is dated at the bottom of this page.
Last updated September 2026
Research at a glance
- Postings collected since April 2026
- More than 160,000
- Employers represented
- More than 27,000
- Hiring systems covered
- More than 50
- Feed coverage
- United States and Canada, some elsewhere
- Published research covers
- United States postings only
- Data used in published research
- Employer postings only
- Smallest group we publish a number for
- 300 postings
What Longbow is
Longbow is a job search platform built by Interview Guys Media, the team behind The Interview Guys. It starts with one action. Paste a job posting, and Longbow prepares you for that specific job.
That means a plain read of what the posting is really asking for, the questions that interview is likely to include with coached answers for each, and research on the company. A career coach named Robin works alongside you through all of it.
Longbow also finds jobs. Our Find Jobs feed collects postings straight from employers’ own career pages and shows you the date we first saw each one, so you know how fresh a listing is before you spend an evening on it.
Longbow Research exists because of that feed. Helping people prepare for jobs means watching a very large number of job postings very closely. This is what we see when we step back and look at all of them at once.
Where the data comes from
Our feed collects postings published by employers on their own career sites. Those sites run on hiring systems such as Workday, Greenhouse, SAP SuccessFactors, SmartRecruiters, Paylocity, ADP and Oracle, and we currently collect from more than 50 of them.
The full feed refreshes twice a week. A smaller incremental pull runs several times a day to catch new postings between refreshes.
The feed covers the United States and Canada, with a small share of postings from elsewhere. Since April 2026 it has collected more than 160,000 postings from more than 27,000 employers.
Published research uses the United States portion only. The Canadian and other postings stay in the product but not in the numbers we publish, because pay, law and hiring norms differ by country and pooling them would blur every finding.
Before we count, exact duplicate postings are removed and employer names are matched to one company record, so an employer that posts under several names counts once.
For each posting we keep the title, the employer, the location, the work arrangement, the employment type, the skills the posting lists, the benefits it mentions, the pay range when one is stated, the hiring system it was published on, the date the employer posted it and the date we first saw it.
Some fields are read from the posting text by machine. Work arrangement, industry and listed skills come from the data vendor that supplies our feed. Years of experience required, degree requirements, seniority, pay period and role family are extracted from the posting text by our own pipeline, which records what the employer actually wrote rather than a vendor’s guess. We say so wherever a finding depends on one of these fields, and we report how many postings carry the field before we report a percentage.
Our product also shows listings collected from job boards. In research, a board copy of a posting that already exists on the employer’s own site is an exact duplicate and is removed before we count. A board listing with no career-site twin counts like any other posting, and each report says how many postings it covers.
How we count
Every number we publish comes with three things: the window, the universe and the denominator. If a figure does not say when, what and out of how many, it is not ours.
The window. Each finding names the dates it covers. A posting is dated by the day the employer published it, and the window is closed before we count.
The universe. Unless a report says otherwise, the universe is employer postings in the United States: active postings, exact duplicates removed, employer names matched to one company record, with the structured fields read from the posting text. When we say a posting is live, we mean it was active in our feed at the time of counting.
The denominator. Percentages are calculated over the postings that actually carry the field in question. If a fifth of postings have no work arrangement recorded, the remote share is a share of the four fifths that do, and we say so.
Minimum size. We do not publish a figure for any group smaller than 300 postings. Most groups we publish are in the thousands.
Measured, not estimated. We publish counts we made ourselves. We do not carry statistics from other studies into our own findings, and we do not present a model's estimate as a measurement. If a number is a sample or a projection, it is labeled as one.
What the data cannot prove. A posting tells us what an employer wrote, not what happened next. Our feed can show that a posting expired and reappeared. It cannot show whether anyone was hired in between, so we do not call a relisted posting a ghost job. A stated pay range is what the employer listed, not what anyone was paid. We report the posting and leave the intent to the reader.
Reproducibility. These definitions are fixed across reports. When a definition changes, the change is dated in the corrections log below and earlier reports are not silently restated.
What we never publish
Longbow Research is built from job postings that employers published for anyone to read. Nothing our users type, upload or report is used in these publications. Not resumes, not the jobs people are pursuing, not interview outcomes, not conversations with Robin.
If that ever changes, the rules below apply before a single figure is released. They are the same rules already written into the Longbow product.
Nothing individual. No finding will ever describe one person, one resume or one application, at any volume.
Five people before a company is named. No claim about a specific employer's hiring process is published unless at least five different users contributed to it.
Ten people before any pay figure. Pay is published only in bands of $5,000, and only when at least ten different people contributed. Exact figures never leave the raw data.
Staff and test accounts are excluded from every count, every time.
No selling of individual data. Longbow Research is publication, not a data product.
Employer names appear in our research only as they appear on public postings, and only in aggregate unless the finding is about a public posting itself.
Questions about any of this go to support@longbow.careers.
Corrections and changes
If we get a number wrong, we correct it on the page where it appeared, mark the correction with a date, and say what changed. We do not quietly replace figures.
If a definition in How we count changes, the change is logged here with the date it took effect and the first report it applies to.
Changelog
- September 2026Page published. Definitions v1.
Cite us
Longbow Research, Interview Guys Media. Report title, publication date, longbow.careers/research. For press or data questions: support@longbow.careers.