"That's for big companies." I hear it every time I talk about HR Analytics with SMB managers. And I get it: it sounds like something that requires a team of data scientists and millions in software.
But the reality is simpler. If your company has more than 30 people and you keep decent records of attendance, leave, and salaries, you already have the basics to start. I'm not talking about artificial intelligence or sophisticated predictive models. I'm talking about looking at your own data and finding patterns you don't see today.
What is HR Analytics, simply put?
It's using human resources data to make better decisions about people. Instead of deciding based on gut feeling or "we've always done it this way," you use numbers.
For example: instead of believing turnover is "more or less the same," you calculate that 18% of people left last year, that most left before reaching 6 months, and that the sales department has three times the turnover of everywhere else.
With that information, you can do something. Without it, you just have feelings.
The data you probably already have
Before thinking about buying software or hiring someone, check what data already exists in your company. You'll be surprised.
In your payroll system or ERP:
- Start date for each employee
- End date (and reason, if recorded)
- Position, department, location
- Gross and net salary
- Leave days
In your attendance tracking system:
- Hours worked
- Tardiness
- Overtime
In Excel or various systems:
- Performance evaluation results
- Training completed
- Climate surveys (if you do them)
With that, you can already do quite a bit. You don't need more—at least to start.
The 4 metrics I'd start with
If I had to choose the first HR Analytics metrics for an SMB, these would be them:
1. Turnover rate by department
The basic formula: (People who left in the period / Average headcount) x 100.
The interesting part isn't the overall number, but the differences. If you have 8% turnover in admin and 35% in operations, something is happening in operations worth investigating.
2. Average tenure before resignation
Do people leave after a month? A year? Three years? Each answer points to a different problem.
If they leave before 6 months, there's probably an onboarding problem or mismatched expectations from the hiring process. If they leave at 2-3 years, maybe there are no growth opportunities.
3. Absenteeism by department
Not just total days lost, but where they're concentrated. A department with much higher absenteeism than others might have climate problems, workload issues, or leadership problems.
4. Estimated cost of turnover
Losing someone costs more than it seems: severance, recruitment time, training time for the replacement, lost productivity while they learn. For an operational role, estimate between 3 and 6 months of salary. For a professional role, it can be 12 months or more.
When you put a number to this, conversations with management change. It's no longer "turnover is a problem," it's "turnover cost us $150,000 last year."
How to build your first analysis
Here's a process you can do this week:
Step 1: Export employee data from your ERP or payroll system. You need: ID, name, position, department, start date, end date if applicable, gross salary.
Step 2: Calculate the basic indicators in Excel. Turnover by department, average tenure, distribution of resignations by seniority. You don't need anything more sophisticated to start.
Step 3: Look for anomalies. Any department with very different turnover from the rest? Any time of year when more people leave? Any supervisor with much higher turnover than others?
Step 4: Present the findings to management. No embellishments, with concrete numbers. "Department X has 3 times more turnover than average. If we cut that turnover in half, we'd save $X per year."
Typical mistakes when starting out
After working on this with several companies, these are the mistakes I see most:
Wanting to predict before describing. I've been asked for "predictive turnover models" at companies that don't even know their current turnover rate. First describe well what's happening. Then, if you want, you can try to predict.
Dirty data that nobody reviews. Employees showing as active who resigned 2 years ago. Incorrectly entered dates. Positions nobody updated. Before analyzing, clean. It doesn't have to be perfect, but obvious errors need to be fixed.
Measuring without doing anything. HR Analytics is useless if the numbers don't lead to actions. If you discover that turnover in a department is a problem, what are you going to do? If the answer is "nothing," don't waste time measuring.
When do you need something more sophisticated?
If you already have the basic analyses working and want to go further, that's when it might make sense to invest in tools or more complex analysis:
- A dashboard that updates automatically (instead of exporting data by hand every month)
- Cross-referencing HR data with business data (do departments with better climate sell more?)
- Predictive turnover models (who is most likely to leave?)
- Organizational network analysis (how does information flow?)
But honestly, most SMBs don't need that to start. They need the basics done well first.
A real example
I worked with a service company of about 200 employees. They had never looked at their HR data systematically. When we pulled the numbers, we discovered that:
- 40% of resignations happened in the first 90 days
- Those who made it past 90 days stayed an average of 2.5 years
- The cost of those early resignations was about $80,000 per year
The problem wasn't the company in general—it was that the onboarding process was a disaster. They improved onboarding, put in a buddy system, and the following year early resignations dropped by half.
They didn't use artificial intelligence or expensive software. They used data they already had, Excel, and the will to do something with what they found.
Want help building your first HR analysis? We can review what data you have and build something useful in a week. Let's talk.