How do I find the bottleneck in my hiring funnel?

Two per stage, read together. Conversion rate tells you how many candidates move from one step to the next. Time in stage tells you how long they wait there. A stage with low conversion is filtering, which may be correct or may mean the bar is misapplied or the pool is wrong. A stage with a long dwell time is blocked, almost always by scheduling or by a decision nobody is making. The pattern matters more than any single figure. A funnel that loses most candidates at first screen has a sourcing or advert problem. One that loses them after final interviews has a bar or an offer problem. Pull both from your recruitment analytics rather than reconstructing them by hand.

Compare against your own history and against similar roles, never against a published benchmark. Every funnel narrows, so a low conversion rate is only meaningful if it is worse than what the same role produced before, or worse than a comparable role today. Three questions sharpen it. Did this stage always convert at this rate, or did it change. Are candidates being rejected here, or are they withdrawing, which is a completely different problem. And are the people who pass this stage succeeding later. A stage that rejects heavily but produces strong finalists is working. One that passes people who then fail consistently is not filtering at all. Ask the interviewers at that stage what they are rejecting for, since the answer often differs from the scorecard.

Fix one stage, then measure again before changing anything else. Simultaneous changes make the result uninterpretable. If the block is time in stage, the fix is usually mechanical: self-service booking through interview scheduling software, a feedback deadline, or pre-agreed offer approvals. If the block is conversion at screen, look at the advert, the sourcing channels and the screening criteria. If it is conversion at offer, examine compensation, speed and how well the role was sold. Give each change a few weeks and a named owner. Also record the reason candidates leave at each stage, because a stage that leaks through withdrawal needs a different intervention from one that leaks through rejection. Write down what you expect the change to do before you make it, so the result stays interpretable.

Weekly at a glance for live searches, monthly in depth by role family. Weekly review catches the practical problems while you can still act: interviews not booked, feedback outstanding, a source that stopped delivering. Monthly review by role family is where structural patterns appear, since a single requisition rarely has enough volume to say anything reliable. Keep segments separate. Engineering, sales and support funnels behave differently, and averaging them hides the problem you were looking for. Watch the trend rather than the absolute value, and treat a stage that is deteriorating over two or three months as a signal even when it still looks acceptable. Bring the review into the same weekly meeting where blockers already get owners and dates.