Article
HR KPIs: the five people numbers worth a weekly look
People problems announce themselves late. Someone resigns in November over something that went wrong in July, and the turnover number records it a quarter after that. The rows in this guide are chosen for the opposite property: five weekly measures that move while a fix is still cheap, seven monthly ones that confirm whether the weekly habits are doing anything.
By the Key Performance Indicators editorial teamPublished Updated 10 min read
01
What is HR accountable for week to week?
Filling seats and keeping the people already in them, and both fail quietly. HR owns the process that produces hires and the habits that keep good people from drifting: candidates in play, interviews held, one-on-ones that actually happen, new starters with a written plan. Turnover and engagement are the receipts, and by the time they move the decision was made months ago.
Nobody argues with that split until it reaches the weekly board, where the people rows are usually a headcount and a turnover percentage. Both describe a year, and neither tells a manager what to do before Friday.
The awkward part is that the weekly rows measure managers rather than the HR function. A missed one-on-one belongs to whoever missed it. HR owns whether the habit exists and whether anybody notices when it stops.
02
Which five people numbers belong on the weekly board?
The ones a manager can fix inside a week. Open seats with a live candidate, interviews held, scheduled one-on-ones that happened, new starters with a written plan, and applications per open seat. None of the five needs a survey tool or a closed month, and all five say something about work that is either happening or quietly not happening.
Two of these rows make people uncomfortable, which is a good sign. One-on-ones held and new starters with a plan name a manager rather than a market condition, and the discomfort is the information. A row nobody feels responsible for has not been assigned properly.
Open seats with a live candidate
Approved open roles that have at least one candidate past a first conversation, divided by all approved open roles. Leading, weekly, owned by the recruiter. Take the band from your own trailing 13 weeks. Gamed by counting an unread application as a candidate, so define the stage that qualifies.
02Interviews held
Interviews that happened with a candidate for an approved seat, counted after the fact. No formula, just a count. Leading, weekly, owned by the recruiter. Band off your own trailing 13 weeks. Gamed by counting a screening call as an interview, which lifts the row and fills no seat.
03Scheduled one-on-ones held
One-on-ones that took place divided by those in the calendar for that week. Leading, weekly, owned by each department manager. The target is structural rather than historical: everything scheduled happens or gets rebooked inside the week. Gamed by deleting the meeting instead of moving it, so count deletions beside it.
New starters with a written 90-day plan
People in their first week who have a dated plan with named milestones, over all people in their first week. Leading, weekly, owned by the hiring manager. Structural target of every new starter. Gamed by a template with nothing filled in, so a plan counts only once someone has signed it.
Applications per open seat
Applications received this week divided by approved open seats. Leading, weekly, owned by the recruiter. Band from your own trailing 13 weeks, per role family rather than across all of them. Gamed by leaving old job postings running to collect volume, which buys a good number and no shortlist.
03
What belongs in the monthly people review?
The outcomes, plus the shape of them. Voluntary turnover over a trailing year, the share of it you regret, first-year departures, median days to fill a seat, offer acceptance, an engagement score read beside its response rate, and how many seats were filled from inside. All seven need more than a week of people to mean anything.
Run the review with names in front of you rather than percentages alone. In a company of sixty, one resignation moves the turnover figure by a full point, so the list matters more than the number. Two departures from one team is a manager conversation rather than a market one.
Voluntary turnover, trailing 12 months
People who chose to leave in the last twelve months divided by average headcount over the same period. Lagging, monthly, owned by the people lead. Band from your own trailing four quarters, since a single quarter of it is mostly noise. Gamed by reclassifying a managed exit as voluntary.
Regretted share of leavers
Departures the hiring manager would have preferred to keep, over all departures. Lagging, monthly, owned by the people lead. Band from your own trailing four quarters. Gamed by deciding after the fact, so ask the manager to mark regretted or not within a week of the resignation and leave it alone.
First-year turnover
People who left inside twelve months of starting, divided by everyone who started in the matching period. Lagging, monthly, owned by the people lead. Band from your own trailing four quarters. Gamed by a long probation that ends in a mutual parting, which keeps the row clean and teaches nothing.
Median days to fill a seat
Median days from an approved requisition to a signed offer, on filled seats only. Use the median, because one hard search will drag a mean past anything useful. Lagging, monthly, owned by the recruiter. Band from your trailing four quarters. Gamed by approving the requisition late.
Offer acceptance rate
Offers accepted divided by offers made. Lagging, monthly, owned by the recruiter. Band from your own trailing four quarters. Gamed by making an offer only once the candidate has confirmed they will say yes, which produces a perfect row and hides every pay conversation that went badly.
06Engagement score with response rate
The recommend question asked of your own staff, scored as the share of promoters minus the share of detractors and reported next to the response rate. Lagging, reviewed monthly and asked quarterly, owned by the people lead. Band from your own history. Gamed by chasing responses only from teams expected to answer well.
Seats filled from inside
Open roles filled by an existing employee divided by all roles filled. Lagging, monthly, owned by the people lead. Band from your own trailing four quarters. Gamed by relabeling a sideways move as a promotion, so count only moves that came with a changed job description.
04
Where do HR targets come from?
Your own history for most of them, a decision for a few, and almost never an outside figure. Turnover benchmarks in particular mix industries, pay bands and company sizes into one number that describes nobody. Read your last thirteen weeks for the weekly rows, your last four quarters for turnover, and set the band a little better than where you sit now.
The rows worth setting by decision are the habit ones. One-on-ones held and new starters with a plan are not questions of what the market allows but of what this leadership team will insist on. Pick the standard and let the row report whether it survived a busy month.
For the hiring rows, split the history by role family before you set anything. A warehouse seat and a controller seat fill on different clocks, and one band across both produces a number that is wrong in two directions at once and defensible in neither.
05
Which people numbers should you drop?
Headcount, training hours logged, total applicants and any survey score shown without its response rate. Each is true and none of them would change a decision a manager makes on Monday. Headcount is a cumulative total wearing a people label, and an engagement score from a third of the company is a mood reading from whoever felt like answering.
Training hours is the row people fight hardest for, usually because somebody is measured on delivering them. Hours booked say nothing about whether anyone can now do something they could not before. Measure what the training was meant to change instead.
Total applicants has a subtler problem: it rewards a wide job posting, and a shortlist a small team never works through. Applications per open seat, read beside interviews held, catches that early.
Vanity metric
Headcount, hours logged, applications received: numbers that grow on their own and settle no argument.
02Examples by department
Ten people measurables with an owner role apiece, alongside the lists for the other departments.
03Leading vs lagging
Sorting your own rows by when they move, which matters more here than in any other department.
06
What does the one-on-one row look like written out?
Take scheduled one-on-ones held. The row needs five fields written down: name, owner, definition, band and source. The definition is the part that decides whether the number means anything, because a fifteen-minute corridor conversation and a booked thirty minutes with an agenda are not the same event, and only one of them survives a busy quarter.
Name: scheduled one-on-ones held. Owner: each department manager, reported as one company figure by the people lead. Definition: a booked meeting of at least twenty-five minutes between a manager and a direct report, held in the week it was scheduled, with a rebooking inside that week counting and a cancellation not. Band: 95 to 100 percent, a standard this leadership team set. Source: the calendar export, pulled Friday by the people lead.
In EOS® this sits on the weekly Scorecard next to the sales and cash rows rather than in a separate people report, which is most of the value. A missed week gets read out to the whole leadership team, and the manager answers for it the way the sales lead answers for a thin pipeline.
Scorecard
Where one or two people rows sit once the weekly meeting has a fixed shape, and why only one or two.
02Building a Scorecard
How a leadership team assembles its weekly list from nothing, with a picker that drafts a starting set.
03Metric owner
Why a habit row belongs to the manager who performs it rather than to the person who reports it.
Start with a real question
Your people rows only move once somebody has already resigned?
Send us the people numbers your leadership team sees and how often it sees them. We will mark which ones are habits you can watch weekly, which are outcomes that need a year behind them, and where a manager is being measured on something they cannot change. One reply from a person, inside a business day.
A person reads every one.