Measuring HR impact
·
9 min read
Key findings
In HR.com's 2025 survey of 201 HR professionals, 23 per cent rated their organisation very or extremely effective at people analytics.
Respondents rated generating data good or very good at 54 per cent, and making changes from the insights at 34 per cent.
A request becomes analysis when a decision depends on the answer, and every other request is reporting.
A one-page definitions dictionary agreed with finance prevents disputes over whose attrition figure is right.
The threshold that would change the decision is agreed with the business leader before any data is pulled.
The gap sits between the analysis and the decision, and framing the question and writing the recommendation both fall inside it.
Getting started with people analytics without a data team takes one business decision, three exports from the HR system and a spreadsheet of counts and comparisons. The deciding skills are framing the question, agreeing definitions and presenting a recommendation, which an HR business partner already has.
In HR.com's 2025 survey of 201 HR professionals, 12 per cent reported a dedicated head of people analytics.
The most common arrangement, at 28 per cent, was the HR team as a whole.
What The 2025 Survey Shows
HR.com fielded the survey from August to October 2025.
Respondents were mostly in the United States, and about 60 per cent worked in large or midsized organisations.
Only 23 per cent rated their organisation very or extremely effective at getting value from people analytics.
That share has not moved in two years.
Among organisations with 99 or fewer employees it was 14 per cent, against 27 per cent at 1,000 or more.
Respondents rated their ability to generate data (54 per cent) and understand it (50 per cent) as good or very good.
Far fewer gave that rating to making constructive changes from insights (34 per cent) or measuring their impact (31 per cent).
Agreement that analytics improves talent and business outcomes fell from 57 per cent two years earlier to 45 per cent.
The gap sits between the analysis and the decision, and framing the question and writing the recommendation both fall inside it.
About half of organisations used spreadsheet software, and 13 per cent used specialised people analytics software.
The figures are self-reported.
The report also notes that its comparison of effective and ineffective organisations shows correlation, which does not prove causation.
Reporting And Analysis
A request for last quarter's attrition by function is reporting.
A request to explain why attrition in one function rose is analysis.
Lydia Wu, head of talent analytics for Panasonic USA, gave SHRM a test in 2019.
If the sender does not expect a request to need about 30 minutes of exploration, it is probably reporting.
Her team worked for about 18 months with Excel, PowerPoint and free tools from the IT group.
It made the case for funding after that.
Reporting can be scheduled, automated or dropped.
Analysis takes the sponsor's time, so it is kept for requests where a decision depends on the answer.
The Decision Behind The Request
Requests arrive shaped as data requests, such as attrition by function or cost per hire.
Each hides a question nobody has stated, which one sentence recovers.
The sentence asks what the sponsor would do differently depending on what the number says.
Where there is no answer, the work is reporting.
Where there is an answer, it sets the measure, the grain and the deadline.
The deadline is the sponsor's decision date.
Five common requests each have a decision underneath.
Why is attrition so high. The decision is whether to change pay, managers or the job, and pay costs the most.
Are we paying market. The decision is where to spend a fixed adjustment budget, which a read by role family and regrettable loss informs.
Which managers are the problem. The decision is where to intervene, and a single indicator can indict a manager unfairly.
Is engagement improving. The decision is whether to keep funding an initiative, which needs a comparison group as well as the headline score.
How many people will we need. This is workforce planning, covered in workforce planning after the headcount freeze.
Root cause analysis in HR sets out how to test the explanations behind a request.
Five Measures In The HR System
An ordinary HR system holds five measures that carry most of the diagnostic load.
Attrition by tenure band, with regrettable leavers marked separately
Internal movement rate, by level
Time to fill, set against the dates the business needed the role
Manager span and depth, from the reporting structure
Rating distribution across comparable units
Departures inside the first year tend to point at selection, onboarding or job design.
Departures at two to four years tend to point at progression.
Sparkman separates regrettable attrition, which harms performance, from non-regrettable attrition.
The split decides whether a given rate is a problem at all.
The aggregate rate also converts into hiring effort.
With an average headcount of 100 and 10 per cent leaving, ten hires are needed to hold headcount.
Internal movement rate shows whether a career path exists in practice.
Time to fill matters where the operating plan has changed before the person arrives.
Rating distribution shows whether a rating means the same thing in two places.
In a unit of 40 people, one additional leaver moves the annual rate by 2.5 percentage points.
A Definitions Dictionary
Two people report attrition and produce different numbers.
One includes fixed-term endings and the other does not.
Or one counts internal transfers out of a unit as leavers.
A single page fixes most of it.
For each measure it records the formula, the inclusions, the exclusions, the source system and the owner.
Agreed with the finance business partner, it ends the recurring dispute about whose figure is right.
In the HR.com survey, 62 per cent of effective organisations named data cleaning and preparation among their most useful practices.
Among the others the share was 33 per cent.
The Evidence Standard
Evidence is proportionate to the cost of the decision and how easily it can be reversed.
Three levels cover most people questions.
Description. Counts and rates for a defined population over a defined period
Comparison. The same measure against the unit's own history, another unit or the plan
Explanation. A test that separates two or more rival explanations
A decision to review a process needs description.
A decision to spend an adjustment budget needs comparison.
A decision to move a manager needs explanation, and no single indicator supplies it.
Small populations need particular care.
In a unit of 40 people, one additional leaver moves the annual rate by 2.5 percentage points.
A rise from three leavers to four is a change of one person, so the count belongs beside the rate.
Two measures that move together do not identify a cause.
A finding states the pattern and lists the explanations still open.
The threshold is the figure at which the sponsor would act.
It is written down before any data is pulled.
That turns a debate about whether the finding is convincing into a check against a standard both parties set.
Spreadsheet Before Software
Most people questions are counts, rates and comparisons across defined groups.
A spreadsheet performs all three.
Pivot tables, Power Query and nested formulas are the first skills to build, ahead of Python, R or machine learning.
Excel remains among the most used tools in people analytics, for its speed, flexibility and familiarity to HR teams.
The sponsor and the finance partner can check every formula, which builds the credibility a first project needs.
AI tools are entering the work quickly.
Only 26 per cent of HR.com's respondents used none, down from 60 per cent a year earlier.
The most common uses were drafting narratives (35 per cent) and analysing unstructured text such as comments and interview notes (30 per cent).
Coding comments into themes is a reasonable use.
The practitioner reads a sample of the output against the original comments and keeps identifiable text out of public tools.
Survey results already in hand are a natural first source, and what to do after an engagement survey covers choosing the findings a business head can act on.
A Worked Sequence
Four roles take part in a sequence that runs about three weeks.
HR business partner for operations, who does the analysis
Vice president of operations, who sponsors it
Finance business partner, who supplies cost and definitions
HRIS administrator, who provides the exports
The question is one the vice president has already asked aloud, such as why plant supervisors leave within a year.
Week one, the contract. The HR business partner spends 30 minutes with the vice president and records the decision, the date, the population and the threshold.
Week one, the definitions. The two business partners agree what counts as a leaver, a supervisor and a year.
Week two, three exports. The administrator provides leavers for 24 months, monthly headcount for the same units, and open and close dates for roles filled.
Week two, the cost. The finance business partner supplies the replacement cost of one supervisor, so the finding can be stated in currency.
Week three, the page. The HR business partner builds one page and sends it to the vice president two days before the meeting.
The leaver export carries start date, leave date, reason code, role, manager and a regrettable flag.
The page holds six items.
The question and the decision it informs
The finding, in cost and in people
A comparison with the unit's own history and with the plan
Two candidate explanations and the test that separates them
A recommendation with an owner outside HR
A review date
Contracting as an internal consultant covers the first conversation in more depth.
Where the recommendation needs money, the business case for a people programme covers what finance accepts.
The Business Leader Conversation
Two conversations decide whether the work is used.
The first happens before the analysis and the second after the page is sent.
Before the analysis, ask the vice president what the answer is believed to be.
Record it as the first hypothesis.
Then ask which number would change the plan, because that number becomes the threshold.
In the meeting, open with the recommendation, then the finding, then the comparison.
A figure sent without a recommendation gets interpreted by whoever receives it, usually against their own hypothesis.
An attrition figure sent to a unit head becomes evidence that pay is uncompetitive, because that explanation is nearest to hand.
Executives work in cost, risk and time.
The finding is stated as replacement cost, months of delay or roles without cover.
When the finding is unwelcome, the threshold agreed in advance carries the argument.
The leader set the standard and the data met it.
The practice the survey points towards is to engage the business before the analysis, to recommend a concrete next step and to build the output into an existing process.
In HR.com's survey, 41 per cent of effective organisations named working through analytics issues with business partners among their most useful practices.
Among the others the share was 17 per cent.
Proportionality And Disclosure
People data describes individuals who did not choose to be studied.
The granularity of the analysis matches the decision it informs.
A question about a policy does not need individual-level reporting.
Before starting, ask whether the people in the data would be comfortable knowing the analysis exists and how it will be used.
Where the answer is no, the purpose is usually the problem.
When people experience analytics as surveillance, they answer less honestly, and the evidence available next time is weaker.
Three Costly Habits
Dashboards nobody opens. A dashboard answers questions already framed, so ask whether any decision in the past six months changed because of it
Survey sprawl. Instruments multiply because they are easy to run, and each one uses up some of the organisation's tolerance for being asked
Benchmark hunting. External benchmarks describe other populations, definitions and accounting, whereas the unit's own history cannot be disputed on definitions
After A Year
A function that has done this well holds four things.
A definitions page everyone uses
Three or four questions answered with recommendations
A standing relationship with the finance and operations analysts
A shorter reporting pack than it started with
The first sign of progress is the next brief.
It arrives as a question where it used to arrive as a request for data.
Where The Tools Sit
In Trusted Talent Advisor, the evidence standard and the translation into a business case sit in component 06, Evidence and Impact.
The sequence for establishing what the business needs is in component 01, The Core Practitioner Guide.
The Talent Template Pack includes the canvases for the measure, the threshold and the recommendation.
Related reading: internal mobility metrics without a platform.
Sources
1. HR.com's HR Research Institute. HR.com's State of People Analytics 2025-26. HR.com, November 2025.
2. Kogan Page Editorial. People Analytics 101. Kogan Page.
3. Sparkman, R. Strategic Workforce Planning, 2nd edition. Kogan Page.
4. Zielinski, D. How to Start and Scale HR Analytics for Almost Free. SHRM, 25 October 2019.
From the system
Trusted Talent Advisor sets out the full method: seven components, 51 tools, The Talent Template Pack and The Prompt Library.
See the system
Questions this answers
How do I start people analytics with no data team?
Pick one decision a business leader has already raised, agree the definitions with finance, export the data from the HR system and answer in a spreadsheet. Write the threshold that would change the decision before pulling any data, and finish with one page and a recommendation.
What are the first people analytics projects to run?
Early attrition in one function is the usual first choice, because the population is defined, the data exists and the cost converts to currency. Time to fill for a business-critical role and internal movement at one level are the next candidates.
Do HR business partners need analytics skills?
They need spreadsheet fluency, the habit of converting requests into decisions and the ability to state a finding in cost, risk and time. In HR.com's 2025 survey, only 34 per cent rated their organisation good at making changes from insights.
What tools do you need for people analytics?
A spreadsheet and an export from the HR system cover a first project. In HR.com's 2025 survey about half of organisations used spreadsheet software, and 13 per cent used specialised people analytics software.
What HR metrics should a small team track?
Five from the HR system: attrition by tenure band, internal movement rate by level, time to fill against decision dates, manager span and rating distribution across comparable units. Each needs a written definition agreed with finance.
How do you present people data to business leaders?
State the recommendation first, then the finding in cost, risk or time, then the comparison with last year or the plan. Send the page two days ahead and agree the threshold with the leader before the analysis starts.
Keep reading
Measuring HR impact
Internal Mobility Metrics Without A Platform
Seven measures an HR adviser can build from existing records, each with a stated definition, a base and a way to read it.
7 min read
Workforce and future skills
Job Rotation As A Talent Gap Strategy
A job rotation programme closes a talent gap when the role, the person, the cover and the evidence are agreed with both managers before it starts.
8 min read
Talent reviews and succession
Seeing Talent Below The Top Team
Visibility of leadership potential drops below the top team. A cross-unit forum, run on one page per unit and a written evidence standard, is the method for seeing it.
8 min read