Monday, 9:12 a.m. Your recruiting analytics software shows a wall of green.

Messages sent: up 4x. Profiles reviewed: up 3x. Candidates touched: a number worth a screenshot.

Meanwhile your Senior Data Engineer search is on day 38. Nine people replied. The hiring manager just rejected the shortlist.

The numbers aren't lying. They're measuring the wrong thing. AI made activity nearly free, and a metric that costs nothing to inflate stops telling you anything. So what should you count instead?

Your Dashboard Is Revving in Neutral

Activity numbers used to be a rough proxy for effort. A recruiter who sent 40 messages a day was working. Then outreach tools learned to send 400, and the proxy snapped.

Activity became free

AI drafts the message, personalizes the opener, schedules the follow-up. Your send count climbs. Your effort doesn't.

• Messages sent: up.
• Profiles reviewed: up.
• Offers accepted: flat.

Volume rewards the wrong habit

When the scorecard counts sends, nobody asks if a message was worth sending. Replies dip, so the fix is more sends. You get a louder outbox and a shallower pipeline.

Take that data engineer search. 600 messages in two weeks. Nine replies. Two screens. Zero finalists. The funnel looks busy. It's just wide at the top and empty underneath.

Leaders ask for a number, so you hand over a number

Your VP wants weekly activity. You send whatever the tool can export. Nobody asks what it's for, and everyone is happy because the chart goes up and to the right.

Averages hide the stall

Time-to-fill is a trailing number. It confirms a role was slow after it's already slow.

The candidate who waited six days for feedback never shows up in an average. They show up in someone else's offer letter.

You're not tracking a hiring process. You're watching a gauge spin while the car sits in neutral.

What Recruiting Analytics Software Should Count: Replies, Waits, Wins

You can run this in a spreadsheet tonight. The test for any metric is one question: can AI make this number bigger without making a hire more likely? If yes, demote it.

Replies, not sends

Count the humans who answered. Outreach volume is cheap. A person writing back "tell me more" is not. If replies per role are falling while sends climb, you've found the leak: better targeting beats a bigger blast.

• Replies per role.
• Replies that became screens.
• Screens per shortlist.

Waits, not averages

For every role, find the longest wait: the candidate who has sat in one stage the longest with nobody moving them. Check it today, not at month end.

Set your own tolerance. If nobody has touched a candidate in three working days, that's a wait worth a call. AI can't shorten a wait. Only a decision can.

Wins, not touches

Shortlists the hiring manager accepted. Interviews that earned a second round. Offers that closed. These cost you something real to produce, which is exactly why they're worth trusting.

Now re-score that data engineer search. Replies: nine, but only two became screens. Longest wait: a strong candidate, six days without feedback. Wins: zero shortlists accepted. Same search, same two weeks, and suddenly the red is obvious. So is the fix: send less, target better, and lock the criteria with the hiring manager before sourcing another 600.

Everything else is a diagnostic. Keep sends and profile views in the small print, never on the scoreboard. We hit the capacity side of this when

every recruiter got an AI sourcing tool

and the ceiling stayed put. For a daily rhythm to read these numbers in, try the

ten-minute morning board

.

Run it for two weeks. The roles with green activity and empty wins will announce themselves.

The Knoot Angle

A scoreboard shapes behavior. Count the wrong thing and AI will help you win the wrong game.

That's the thinking behind Knoot's Activities Dashboard: one screen for the day's recruiting activity. New candidates, pipeline status and the key numbers update as things change, so what's moving sits next to what isn't. It's a daily-ops view, not a quarterly analytics suite. The question it answers is the one you actually have at 9 a.m.: what happened, and what's stuck?

It surfaces. You decide. Which role needs a call, which candidate needs a human: that part stays yours.

Let AI handle the volume. Keep your eyes on the outcome.