Answers
Pipeline
Applications per opening, and what a tripled pool does to a process that was built for a smaller one.
81 articles
Everything on pipeline
- Use AI to Choose Where to Apply, Not to Apply Everywhere AI job-search tools split into four categories that fail differently. Automate discovery and tracking. Keep writing and submitting for yourself.
- The AI Wage Premium Is Measured on Postings, Not on People The AI wage premium compares postings, not people. How the 28% and 62% figures were built, and the one check that actually applies to you.
- What to Put in a Cold Message So It Gets Answered A template reads as a mass send before the first line ends. What still works: three sentences, one specific line, and an ask small enough to answer in a line.
- How to Tell a Live Opening From One That Won't Hire Most stale postings trace to a rehire pool, an unconfirmed budget, or a forgotten close, not a deliberate fake. Check the tells before you spend an hour.
- Automate the Tracking, Never the Judgment A tracker is safe because it records instead of representing you. What to save for every application, and why the posting text matters more than the link.
- Build a Pipeline of Dated Evidence, Not a List of Interested People A stored candidate rating decays faster than the record holding it. Keep dated artifacts, the rules in force, and one sentence of observation, with a one-year shelf life.
- Offer Scope and Decision Rights, Not Upside You Cannot Price Candidates can price your offer against the market now. Name the decisions the candidate owns, write the equity as three numbers, and say where the offer stops.
- Cost Per Hire Counts the Recruiter and Skips the Panel Cost per hire is comparable only inside a boundary you wrote down. Track recruiter hours, interviewer hours and tooling separately, and compare against your own year.
- Measure the Barrier, Not the Group: Disability Has No Denominator The four-fifths rule has no group to divide by for disability. Audit each stage for the capability it silently demands and instrument the drop-off.
- Most Lost Callbacks Come From a Minority of Employers A fifth of discriminating firms accounted for nearly half the callbacks lost to Black applicants. Why a market average says nothing about your own company.
- Applying Early Matters More Than Applying Perfectly The 24-hour rule is repeated with no mechanism attached. Check whether the posting reviews on a rolling basis or a stated date, then decide how fast to move.
- Easy Apply and the Company Site Are Different Applications A board form and the company's own careers page are not always the same road in. Check where the questions come from, apply once, and skip the duplicate.
- Entry-Level Hiring Fell Where AI Automates, Not Everywhere A Stanford payroll study found young-worker hiring down in AI-exposed occupations. Here is the conditional finding the headlines cut, and what it means for your search.
- How Recruiters Search, and How to Be in the Results Recruiter sourcing is a database search, not a resume screen. Title, the plain name of your work, location and recency decide if you appear; one line gets you opened.
- Cut Calendar Time, Not Evidence, When the Start Date Is Fixed Most compressible time in a small company's hiring is waiting, not stages. Cut by evidence produced per hour and protect the stage where somebody makes something.
- Settle Four Things at Intake, Then Test Them on Three Real Profiles Four things a recruiter and a hiring manager settle before the req opens, plus the twenty-minute profile exercise that shows whether the stated must-haves are real.
- The Longest Gap Between Stages Is the One a Candidate Feels A candidate feels the longest gap between stages, not the 44-day total. Put a service level on each gap and publish the elapsed time you expect.
- Fewer Applications With Evidence Beat a Hundred Without The number vendors sell (twenty, thirty) is their engagement metric, not your strategy. Split real hours into a selective layer and a wide, unedited one.
- Hundreds of Applications, No Replies: What Actually Changed Silence at scale is a denominator problem before it's a rejection. What two decades of rising application volume does to your odds, and what to do instead.
- Budget the Whole Loop in Candidate Hours, Not Round by Round Nobody publishes a whole-loop total, so set your own: four to six hours end to end for most roles, with every stage buying its share.
- The CS Degree Question Is a First-Job Market Question The first-job market has tightened, partly for reasons that predate AI, and full replacement of work skills stays rare. Spend the degree on specifying and verifying.
- Pick the Title for the Level You Will Assess Against A title binds the level: what the interview loop assesses, who the candidate is compared with, and the band they arrive expecting. Findability is the easier half.
- Auto-Apply Bots Cost You the Application You Could Defend Auto-apply tools raise your application count and lower what each one carries. What to automate in a job search, what to keep yours, and what the evidence shows.
- Compute Selection Rates Per Gate: The Total Hides the Bad Gate A per-gate table, with the denominator set to who entered that gate. The end-to-end hire rate can read 1.00 while the resume screen sits at 0.60.
- A Must-Have Is Any Requirement You Are Willing to Test For One rule sorts the list: a must-have is tested by a named stage with a rule that rejects. Everything else is a preference, and the bullet count is noise.
- The Nice-to-Have List Stopped Filtering and Started Attracting A preferred-qualifications block worked by discouraging applications. It discourages far fewer now, so either delete it or replace it with one line the loop asks about.
- Offer Acceptance Rate Measures Your Speed More Than Your Appeal Offer acceptance rate depends on where your offer record starts and on who declined. Published bands are not comparable: read declines by reason, track days to decision.
- Source Outbound When You Need Evidence, Not When You Need More People The conversion gap vendors quote for sourced candidates measures pre-selection. Choose outbound or inbound by what you are short of, and judge both on one pass rate.
- Stop Paying an Agency for Access and Start Paying for Evidence A placement fee prices a submission, not verification. Sign when reach is genuinely the constraint, and make every agency candidate run the same evidence stage.
- Cap Referrals at a Third of Hires and Screen Them Like Everyone Referred candidates pass screens at a higher rate partly because the screener can see the label. Set a ceiling, keep the bonus, and assess everyone the same way.
- If the Screen Never Overturns the Resume Read, It Is a Calendar Step Pass-through rates measure throughput. Count how often the screen reverses the pre-call read, in both directions, and you learn whether the round produces anything.
- Most AI Jobs Are Ordinary Jobs With One New Requirement AI job titles split into three families, and only one is open to most career-changers without a specific technical background.
- Source of Hire Records the Route, Not Where the Person Heard The source field records which system delivered an application, not where someone heard about the role. Read it by hires, and ask each hire directly.
- Pick One Clock, Write Down Its Start, Stop Benchmarking Outward Time to fill runs from the requisition being opened to acceptance; time to hire starts when the candidate enters. Report one, publish its timestamps, and read the legs.
- Too Few Hires for a Ratio: Measure the Process Instead At single-digit selection counts one candidate flips the verdict. Run the flip test, report counts, and measure the decisions instead, which you have enough of.
- Where Your Application Actually Goes After You Hit Submit Six stages run between your click and a decision: parsing, dedup, rule checks, ranking, recruiter review, and a status code. Here is what each one can and cannot do.
- A Procedure Is Biased When Decisions Move on Job-Irrelevant Facts A hiring process is biased when its decisions move with something not related to the job. That is testable on ten past decisions, and a motive is not.
- There Is No Safe Job List. There Is a Task Exposure Test. Safe-jobs lists expire within months and score a title, not the tasks inside it. Here is the four-question test you can run on any job, including your own.
- Move the Decision to the Stage That Still Produces Evidence The resume screen sorts, it no longer selects. Put the decision at the first stage where a candidate produces work under conditions you set, and price it first.
- The Stage That Opens a Gap Is Rarely the Stage That Shows It A late-stage gap is mostly inherited. Compute entry composition and pass rate at every stage, then name the one boundary where the gap actually widens.
- Years of Experience Is the Weakest Requirement You Will Write Prehire experience correlates .06 with later job performance, so a years minimum sorts on career shape rather than capability. What to write in its place.
- Why Do Candidates Who Interviewed Brilliantly Struggle in Their First Quarter? A great interview samples a supervised, framed hour; the first quarter samples unsupervised generation. The gap scales with how much of the role starts as a draft.
- Drafting a Job Description With AI Without Inventing Requirements A model writes a usable job description draft. The failure is the qualifications section, where requirements arrive with no origin. Feed it tasks, then trace each line.
- How Do You Write an AI Requirement Without Flooding the Pipeline? Name a tool and every applicant who has opened it qualifies. Name the task the job runs, plus the sentence saying the claim gets checked, and the pool sorts itself.
- Put the AI Rule in the Job Post and Watch Who Still Applies One sentence in the process section: which stages are AI-open, and what the next call asks about. Published before anyone applies, it is the only version you can enforce.
- Required or Preferred: Where AI Skills Belong in a Posting Required means you would decline a strong candidate on day one for lacking it. That is rarely true of AI tool use, so the AI line belongs in the duties.
- What Do You Do When a Candidate Has No Idea They Applied? An agent applied for them, which is a consent gap and not fraud. Add one confirmation act between application and screen, sized to cost a real applicant thirty seconds.
- Friction Filters Effort, Not Ability. Spend It Where It Pays Back. Friction filters only when it costs candidates differently, and text no longer does. Add a step that produces evidence and hands the candidate something back.
- Why Did Applications Per Opening Triple, and How Many Are Real? Volume rose because producing an application costs a candidate almost nothing and one-click apply removed the friction. Measure intent density in your own funnel.
- Is It Cheaper to Assess Every Applicant Than to Screen Resumes? Assess-first pays only below roughly four dollars per completed assessment. The break-even arithmetic, the five numbers it needs, and where the line sits for campus.
- What Replaces an ATS Keyword Filter When Every Resume Matches? Keyword screens measured effort to learn a job's vocabulary, and that effort is now free. Replace them with checkable hard requirements and an earlier work sample.
- What a Bad Hire Costs, and Why a Salary Percentage Cannot Answer It A percentage of salary describes no company in particular. Build the number from four internal lines, then price the mistake nobody puts on the other side of the ledger.
- What to Delete From a Job Description When AI Does That Part The deletion list nobody publishes: production verbs an assistant now drafts, software-proficiency lines it absorbed, and the years-of-experience proxy under both.
- The AI Screening Notice Belongs in the Job Posting, Not the Careers Footer No federal duty. NYC wants notice ten business days before a covered tool runs, and names the job posting as one route. Illinois has required notice since January 2026.
- Hiring Funnel Conversion Rates, and Why Yours Dropped Published stage benchmarks divide by applications received, and that number moved. Index the funnel backward from hires and compare against your own history.
- Which Funnel Metrics Still Mean Anything Now That Candidates Use AI? Stage pass rates now track how cheaply a candidate can produce what the stage asks for. Keep three metrics, count evidence density, and re-baseline per job family.
- How Long to Leave a Job Posting Open Now The two-to-four-week rule assumes applications trickle in. Set the close by how many a person can read, publish that limit in the posting, and reopen on a rotation.
- Write a Job Description for a Role AI Already Changed Sort every task three ways before rewriting: an assistant drafts it, a person owns the call, or nothing changed. Only the second column belongs in requirements.
- Your Req Closed in Two Days: Did You Select for Bots? A count-based cap closes on arrival time, so it samples alerts, time zones and auto-apply tools. Replace it with a timed window, a random draw, or a short work act.
- Turn the AI Line in Your Job Posting Into One Answerable Question Write the interview question first and the posting line second. The rewrite that turns AI fluency required into something a candidate can pass or fail.
- Your Matching Engine Scores Against the Posting You Wrote A matching engine compares applications against your requisition text. Write the three tasks the role does in a normal week, then read your own top twenty.
- Literacy, Fluency, Prompt Engineering: Only One Is a Hiring Bar Literacy is an afternoon of onboarding and prompt technique a fortnight. Fluency is the judgment nobody installs after the hire, and the only bar worth writing.
- Measuring Quality of Hire Without a Satisfaction Survey Name one or two role-specific outcomes before the offer, collect them on a schedule the hiring manager doesn't control, and compare cohorts rather than people.
- Naming AI Tools in a Posting Versus Naming the Behavior A tool name in a posting is the keyword every applicant mirrors back. Name the artifact and the behavior instead, and name a product only where nothing substitutes.
- Proving a Hiring Process Change Actually Worked A before-and-after chart across two quarters proves nothing. Assign at the req, hold the role family constant, and name the outcome before the numbers arrive.
- Publishing the Evaluation Process Inside the Job Posting In the job posting, list the stages, what each examines, where AI touches the file, where a named person makes the call, and how long it takes. Then run exactly that.
- Automate Everything That Doesn't Change Who Advances Scheduling, reminders and status updates change nobody's outcome and are worth automating now. Anything that can move a candidate's stage is a selection procedure.
- The Parts of a Posting That Actually Reduce Generated Applications Requirements cannot deter a generated application. Publish the pay range, state the location rule exactly, and ask for one thing that cannot be mass-produced.
- Does 2024 Interview Evidence Still Hold When You Rehire From the Pool? What 2024 proved about judgment, domain depth and communication still holds. What it proved about how they work is stale. Re-verify that half with one work act.
- Releveling a Job After AI Absorbed Part of the Work AI takes the production half first, which was the junior half. What remains is review and consequence. Set the level from that, or cut the scope to match the range.
- A Remote Posting Adds the Candidate's State AI Law to Yours The trigger in these laws is where the candidate or the position sits, not the head office. Two levers exist, and both live in the posting's location field.
- Should Your Job Post Require AI Experience, or Does It Screen Out the People You Want? Requiring tool names filters for trivia and requiring nothing filters for nobody. Name the act the role performs with a model, in that occupation's own language.
- Naming the Assessment Stage in the Post Moves Drop-Off Earlier Three facts near the end of the post: how many stages, that one is an assignment, how long it takes. Expect fewer applications, better completion, fewer wasted hours.
- Cutting Twelve Hundred Applications to a Shortlist You Can Defend Write the cut rule before you open the pile, sample instead of reading down the queue, cap the read at what two calibrated people can do, then audit the rejections.
- Should You Skip the Resume Screen and Go Straight to a Work Sample? Send a work sample to everyone only while applicants times review minutes fits your hours. Above that, keep a documented resume check and move the sample earlier.
- How Do You Test AI Skills Without Adding an Hour to the Loop? Cut the round whose output a model now writes in seconds and spend its minutes on an AI work sample. The stage to cut differs for an engineer and an analyst.
- Where Did the Time Go After AI Entered Every Hiring Stage? AI cut the cost of applying far more than the cost of judging, so volume rose and rounds got re-added. Find the stage whose pass rate jumped, then cut below it.
- How Do You Find Out What AI Does in a Role Before Writing the Job Post? Pull the occupation's task statements, mark the tasks a model already drafts, then check the list against three people doing the job now. What survives is the ask.
- An AI Skill Is a Decision, Not a Tool You Have Open Four decisions carry almost all of it: what you hand over, how much context you give, what you check before it goes out, and what you refuse to delegate at all.
- How Do You Write an AI-Skills Requirement That Isn't Legally Vague? Name the occupational task, the judgment the person keeps, and the proof a candidate can show. Six before-and-after rewrites, plus the ADA test a vague line fails.
- Years of AI Experience: What to Require Instead A year count reports accumulated exposure, which a three-year-old tool cannot supply. Ask for recent use on work of this kind, then test the claim once.
This beat is written by Ella McChesney & Olive.