The numbers that actually run a recruitment desk

A while back I wrote a long post on how the weighted forecast on the Jobs dashboard is calculated. The question I get now is the one that arrives the moment a manager starts trusting that number. The forecast tells you what is coming. It does not tell you what to do about it, and it does not answer the five things a desk manager actually loses sleep over.

Start with what a recruitment desk really is. It is a throughput business with a cash lag. You get paid only when a job fills, and most open jobs never will. So the entire game reduces to three things. Work the right jobs, move fast, and do not leak. You lose that game to three silent killers: dead jobs quietly eating recruiter capacity, conversion leaks nobody can see, and clients who waste your time and never decide. A single forecast number sees none of them.

This post goes deeper than a feature tour, because each read only makes sense once you can see the decision it drives. Here are the five questions, and for each one the insight and the statistically honest way to show it. A desk view that answers these is worth building. One that shows you activity totals is not.

Whether you will hit number is a ratio, not a feeling

The first question every desk manager asks near the end of a billing period is whether the desk will hit number. The forecast already computes the weighted pipeline. Divide it by the target you have left to bill and you get a coverage ratio, and that single figure is the answer. At around 1.4 times cover you are comfortable. At 0.7 you will miss, and you need to generate now rather than find out on the last working day of the month.

The value is in the reframe. A weighted pipeline figure is something you read. A coverage ratio against target is something you act on. It turns the forecast from a number into a verdict. Show it one way per desk and the same way per consultant, colour-coded, so a manager sees in a second which desks are covered and which are about to come up short while there is still time to fix it. Anything below 1.0 is not a worry to note. It is an instruction to go generate.

The highest-value screen ranks jobs by fee at risk, not by status

This is the single most valuable screen on the dashboard, and it is the one almost no agency has. The biggest capacity leak in any desk is recruiters spending hours on jobs that were never going to fill. To find them you have to stop classifying jobs by status and start classifying them by behaviour.

Starved. The job is open and almost no pipeline is being added to it. That is a sourcing failure, not a pipeline problem. Clogged. Plenty of pipeline, no forward movement. Either the recruiter is not pushing it or the client has gone quiet. Stalled. It was progressing, and then nothing has changed stage for more than your chosen number of days. Dead. Open past two to three times your median time-to-fill with nothing happening. Close it or escalate it, but stop pretending it is alive.

Then sort the screen by money, not by count. The headline a manager needs is a sentence like "£148k of fees sitting in nine stalled or dead jobs", with the worst offenders at the top. A stalled £25k role matters more than five stalled £2k roles, and a count-based list buries that completely. Sorting by fee at risk is what tells a recruiter to stop working ghosts, and working ghosts is the largest single capacity leak there is.

None of this needs new data. The stage dates already sitting on each job's pipeline give you the date of last movement. The fee and the created date give you the age and the value. The whole classification is arithmetic on fields you already hold. How you set that fee in the first place, fixed or a percentage of salary, changes what "at risk" actually means, and the guide on agency fee structures covers those trade-offs.

Conversion only means anything against your own baseline

The instinct when someone asks where a desk is leaking is to show a funnel. A funnel on its own is close to useless. It tells you where candidates drop out. It does not tell you whether that drop is normal for your desk or a genuine problem, and without that you cannot act on it.

Show each conversion against the desk's own trailing twelve-month baseline instead, flag the one stage that is abnormal, and split the stages by who owns the fix. Applied to CV sent is your screening and your speed. CV sent to interview is the client buying the quality and the match you put in front of them. Interview to offer is client decision and fit. Offer to placement is closing, counter-offers and candidate drop-out. Four conversions, two of them mostly yours and two of them mostly the client's.

Then attribute the rejection reasons to the stage that is failing, using the reason recorded against each rejection and whether it came from the client or from your own team. Now the read becomes a diagnosis. "CV to interview is running at half your baseline, and the top reason is insufficient experience" tells a recruiter, in one line, that the desk is sourcing the wrong people. That is something to act on by Friday, not a chart to nod at in the Monday meeting. The reporting layer this sits inside is covered in the guide on recruitment KPIs and reporting.

Speed is the leading indicator, and the average hides it

Three numbers answer whether a desk is fast enough to win, and the way most dashboards present them actively misleads.

Time to first quality CV is the real leading indicator, and almost nobody tracks it. On a contingent, multi-agency brief the desk that submits a good candidate first usually takes the fee, and everything after that is reaction. If you measure one speed number, measure this one, because it predicts winning better than anything downstream of it.

Time to fill belongs as a distribution, not an average. Time-to-fill is heavily right-skewed. A handful of 200-day zombies drag the mean somewhere meaningless, so a blended "141 days" describes none of your jobs. A contract role that fills in five days and a permanent role that fills in sixty do not average into anything you can use. Read the median, split it by job type, and the figure starts telling the truth.

The cohort survival curve is the honest way to talk about fill rate given the cash lag. Take the jobs opened in a month and plot the share still unfilled at day 30, 60 and 90. The curve shows the inflection where "not filled yet" quietly turns into "never will". That inflection is not a hunch. It is what sets the dead-job threshold in the triage screen, so the line between a stalled job and a dead one comes from your own data instead of someone's guess.

Who and which clients are worth the capacity is a two-axis question

The last question is where the capacity should go, and it splits in two. Both views plot results against load, and both deliberately ignore activity.

Consultants, fill rate against job load. This finds the recruiter sitting on twenty-five open jobs and filling one, who is either overloaded or unfocused, and it finds the quiet one filling four of six. An activity dashboard rewards the first recruiter for looking busy. This one shows you which of the two is actually billing.

Clients, fill rate against fee value. This is the portfolio call. High fee and low fill is where you are bleeding effort into work that looks valuable and never lands. Low fee and low fill is a client you fire. Add the client's decision speed, measured as the lag from CV sent to interview, and the ghosters who waste a desk's time and never decide expose themselves on the chart. Most agencies fly completely blind on this, and it is pure margin sitting on the table. The same results-over-activity thinking runs through how we approach performance and revenue reporting across the product.

What the screen deliberately refuses to show

It matters as much what a desk view leaves out. Three things stay off it on purpose. Blended average time-to-fill, because it hides the skew that makes the median honest. Raw activity counts, because calls made and emails sent measure effort, and effort is not progress. Vanity totals like "533 applied this month", because a big number with no conversion attached feels like insight and delivers none.

The reason to refuse them is not taste. A dashboard that flatters you is worse than no dashboard, because you pay for it twice. Once in the screen space, and again in the decisions you get wrong while trusting it. Effort is not progress, and averages hide skew. Every read on this view is built to survive both of those traps.

Put the five together and the desk reads as a system rather than a pile of figures. Coverage tells you whether to generate. Triage tells you which jobs to drop. Attributed conversion tells you whether the leak is you or the client. Speed tells you whether you will win the brief at all. The two quadrants tell you who and what deserves the capacity you have left. The forecast started this by telling you what is coming. These five tell you what to do about it.

We hold every number on the dashboard to the same standard. It has to be one you can take apart when it looks wrong, the same way I walked through our candidate dedup pipeline rather than ask you to trust a black box. If you would rather watch these reads move against your own desk than read about them, book a walkthrough and bring a month of open and closed jobs with you. The numbers are far more convincing when they are yours.