Why Your Pipeline Is Lying to You

A pipeline board full of cards is a picture of activity, not of health. It shows you that work exists; it says nothing about whether work is moving. Four numbers do: how long the oldest candidate in each stage has been sitting there, the conversion rate between each pair of adjacent stages, the ratio of candidates entering a stage to candidates leaving it, and the number of roles with no forward movement in 21 days. None of those are visible on a standard board, which is why a stalled role usually announces itself as a client email rather than as something you spotted first.

I ran a recruitment agency for four years before I built Pickr. My board was always full. It was full in the month we billed well and full in the month we billed nothing, and I could not tell those two months apart by looking at it. That is not really a complaint about tooling. A board is built to answer "where is this candidate". It was never designed to answer "where is my time going".

Why a full pipeline board looks healthy when it is not

Three numbers get quoted in every Monday meeting, and all three are close to useless on their own.

Total candidates in process. 180 cards across 12 roles sounds like a strong week. It is equally consistent with 140 of those cards having been untouched for 30 days, which is not a pipeline. It is a graveyard with good lighting.

Applications received. Volume at the top measures your job ad and your brand. It measures nothing about whether the work of assessing those people is happening. Doubling applications while your screening stage already runs 9 days behind makes the number worse, not better.

Active roles. "Active" is nearly always a status somebody set at kickoff and nobody has touched since. Ask a straight question of any agency board — how many of these roles have had a candidate move stage in the last 21 days — and on the desks I have seen, the answer has never been all of them. The rest are open in the system and closed in reality.

The pattern is the same in all three. They count things that exist rather than things that changed. A metric that cannot go down when you stop working is not a health metric.

The 4 pipeline metrics that show where the work is stuck

Stage ageing, not stage counts

Not "how many are in screening" but "how long has the oldest one been in screening". Age is the number that gets worse on its own while you do nothing. A candidate 19 days into a first-interview stage is not a data point. They are a person who has probably accepted something else and has not told you yet.

Read the oldest item per stage and the median age per stage together. A high median means the stage is slow for everyone, which is a process problem. A normal median with two very old outliers means two candidates were dropped, which is an attention problem. The fixes are opposite, and the card count is identical in both cases.

Conversion between adjacent stages

Overall funnel conversion — applications to hire — averages over so many separate steps that it can hide a collapse in any one of them. Adjacent-stage conversion is where the information lives: CV sent to client feedback, feedback to interview booked, first interview to second, second to offer.

On my own desk, the ranges that looked healthy sat somewhere near 40-60% from client-submitted CV to interview and 25-40% from first interview to final stage. Those were my numbers on my roles, not a benchmark to hold yourself to. What matters is spotting the step that runs at 15% while the steps either side of it run at 50%.

One trade-off worth knowing before you build the report: below roughly 20 candidates through a step, this ratio is mostly noise. Read it weekly on a small desk and you will spend the week chasing sampling error. Read it quarterly, or pooled across similar roles, and it holds up.

The entry-to-exit ratio

For each stage, over the last 30 days: how many candidates entered, and how many left in either direction. If entries outnumber exits, the stage is filling faster than it empties, and your time to fill is already growing whether or not anyone has noticed.

This is the earliest warning available. Stage ageing tells you a stage is blocked. The entry-to-exit ratio tells you it is becoming blocked, often 2 to 3 weeks earlier, while the fix is still a scheduling conversation rather than a lost mandate.

Roles that have silently stalled

A stalled role is not one that was rejected. It is one where nothing happened and nobody decided anything. Count the roles with zero stage movement in 21 days and zero candidate-facing activity in 14. Almost nobody tracks it, because a stalled role generates no notification, no red flag and no email. Silence is the failure mode, and silence looks exactly like a quiet week.

Where the hiring bottleneck actually is

Ask a recruitment team where the process breaks and you will hear "sourcing" or "the candidates aren't out there". Measure it and you will rarely find it there.

In the boards I have looked at, the stage that swallows the most calendar days is one of three: waiting for client or hiring manager feedback on submitted CVs, scheduling a first interview across three calendars, and the gap between a final interview and a decision. On a 45-day time to fill, it was entirely ordinary for a third of those days to sit in a stage where the next action belonged to somebody who is not a recruiter and never opens the board.

There are two reasons for the misattribution. Sourcing is the stage recruiters personally feel effort in, so it feels expensive; waiting costs nothing to do and is therefore invisible. And waiting stages usually have no owner. Every stage where the next action belongs to "the client" or "the panel" is a stage with nobody's name on it, and unowned stages are where time goes to die. Once you know which stage it is, reducing time to hire covers the interventions themselves.

What a pipeline board should surface without being asked

These numbers are rare not because they are hard to compute but because they need stage-change timestamps that most systems record and never show, and somebody has to go looking. Nobody goes looking on a busy Tuesday.

Pickr is an AI-native recruiting platform for recruitment agencies and in-house teams, built around movement rather than card count. Pickr's pipeline runs one board per client mandate — with client portals and placement tracking native rather than assembled out of permission workarounds — and flags candidates that have sat in a stage too long and roles that have stopped moving. Recruiting analytics covers time in stage and stage-by-stage conversion broken down by job and by client, and what happened to the people you actually placed feeds back into how the next candidates are ranked.

The honest limitation: none of this makes a slow client faster. Visibility converts a silent failure into a visible one, which is worth a lot, because you can chase a visible problem. It is not the same as solving it, and any vendor telling you a dashboard fixes a bottleneck is selling you a mirror.

How to find your pipeline bottleneck this week

You do not need to buy anything to do this once. Export the last 90 days of stage-change history from whatever you use today and compute three things: median days in each stage, entries against exits per stage over 30 days, and every role with no movement in 21 days. That is about half a day of spreadsheet work. Do it before you evaluate software, because if the answer is an unowned feedback stage, no purchase fixes it and you have saved yourself a migration you did not need.

If you would rather not do it by hand, the free recruiting audit is a 2-minute questionnaire, with the option of a read-only connection to your current system that reports where candidates drop off stage by stage. Auditing the process around those numbers — who owns which stage, and what happens when nobody does — is covered in the recruiting process audit.

Count changes, not cards

A pipeline is healthy when candidates are moving and roles are progressing, and every honest measure of that is a measure of movement over time: age per stage, conversion between neighbouring stages, entries against exits, and the roles that have gone quiet. Your board is not lying to you on purpose. It is faithfully answering a question you stopped needing answered — where is everyone — while staying silent on the one that costs money, which is where is everything stuck.