Hiring Funnel

What inputs does a credible top-of-funnel plan need?

Four things, and most plans are missing at least two of them. The hires required and the dates they are required by. Observed transition rates for the relevant segment rather than the company average. The elapsed time each stage consumes, because volume that arrives too late converts into nothing useful. And recruiter capacity, since a plan that orders more interviews than the team can physically run is a work of fiction that will be discovered in week three. Capacity is the constraint that gets omitted most often and bites hardest.

How much history should the ratios be built on?

Long enough to contain complete cycles and to average out seasonal distortion, short enough to still describe the current process and market. Exclude candidates who are still in flight from the denominators, or conversion is understated in a way that grows worse the faster you are hiring. Re-derive the ratios after any material process change: adding a stage, removing one, or changing who screens invalidates the previous numbers immediately, and continuing to plan on them is a slower version of using someone else's benchmark.

How should a funnel plan be reported honestly?

As a range with its assumptions visible: which ratios were used, over which window, for which segment, and what happens to the answer if the channel mix moves. Put the sample size next to any conversion figure so a reader can judge how much weight it carries. A plan presented as a single confident number stops being a forecast the moment it is circulated and becomes a commitment, at which point the only available response to reality is to argue with it rather than to revise.

How do you work backwards from a hiring target?

Start at the number of hires required and multiply back through the inverse of each transition rate the team has actually observed. If your own history shows that roughly half of screened people reach an interview and a minority of those reach an offer, then the hires you have promised fix the interview volume, and the interview volume fixes the sourcing volume. The arithmetic is trivial. Sourcing the honest inputs is the work.

Two disciplines make the output usable. Build the ratios on a trailing window long enough to contain complete cycles, so people still in process are not counted as failures. Then present a range rather than a single figure. A precise top-of-funnel number invites a confidence the underlying sample cannot support, and it becomes a commitment the moment somebody puts it on a slide. Ranges survive contact with reality; point estimates get quoted back at you.

Why does one blended company-wide funnel mislead?

A blended rate is dominated by whichever segment supplies the most events, which is normally high-volume junior hiring. That average then gets applied to a specialist or senior search where it is wrong in both directions at once: it over-predicts conversion at the top and under-predicts the effort required at the bottom. The number is arithmetically correct and operationally useless, and it fails silently, because nothing in the report signals that the segments underneath it behave differently.

Cut by role family first and by channel second, because a referral and a cold advertising response do not behave alike at any stage. A plan built on last year's blended rate under-orders sourcing the moment the channel mix shifts toward paid volume. The right level of segmentation is the finest one where each cell still contains enough events for the ratio to hold still.

Why must stage definitions be governed before ratios mean anything?

A ratio compares two counts, and both counts depend on a definition that different people apply differently. Where one recruiter records an interview when it is scheduled and another records it when it is completed, the resulting interview-to-offer ratio measures administrative habit rather than hiring. No amount of dashboard work repairs that, because the damage happened at data entry. The only real repair is to fix the definition and restate the history it affected.

Governance here is unglamorous and foundational: a written definition for each stage, an agreed moment the clock starts and stops, one place the transition is recorded, and an occasional audit checking that the recorded stage matches what actually happened. Teams that skip this step spend the following year arguing about whose report is correct instead of about which transition needs fixing. The audit is the part everyone skips and the part that keeps the rest honest.

What is the difference between a leaky funnel and a narrow one?

A leaky funnel admits enough people and loses too many at a particular transition. A narrow funnel converts perfectly well and simply never had enough people in it. They look identical in the hire count and demand opposite responses, which is why the distinction is worth making explicitly before anybody approves spending. One calls for process work at a named transition; the other calls for more or different sourcing.

Separating them is straightforward once the ratios are segmented. Compare each transition against that same segment's own history: if the transitions held and the hires came up short, the shortage is volume; if the volume held and the hires fell, one transition moved and it can be named. Buying more advertising to fix a leak is the most common expensive error in this whole area.

Where does funnel thinking genuinely stop working?

At the top of the market and at the bottom of the volume range. Ratios need events, and a search producing a handful of serious conversations in a quarter cannot support a stable conversion rate. Any percentage computed on that base swings violently when one person changes their mind, and treating the swing as a signal leads directly to a wasted intervention. Sample size deserves to be printed next to the number every single time.

For senior, executive and genuinely low-volume searches, manage the named list instead of the ratio: who has been approached, who is engaged, who is realistically convertible, and what would have to change for the rest. Report it as coverage against a shortlist with names and dates. Keep the funnel arithmetic for the roles where the sample is large enough to carry it, and resist blending the two reporting styles into one review.