People Analytics Playbook for Indian SMBs: Metrics to Act On

Most small and mid-sized Indian businesses already have a lot of people data: attendance logs, payroll registers, leave records, hiring pipelines and exit forms. What they often lack is a way to turn that data into decisions. Leaders ask questions like "Why are we losing people in the first six months?" or "Is overtime getting out of control in the plant?" and the answer comes from gut feel or a late-night spreadsheet.

This guide is a practical people analytics playbook for Indian SMBs. You will learn what people analytics is (and is not), which metrics matter, how to build a simple analytics stack without a data science team, how to analyse attrition, attendance, payroll and hiring data, how to present insights so managers act on them, and how to respect employee privacy while doing it. We also cover how AI can help, and where it can mislead.

What is people analytics?

People analytics, sometimes called HR analytics or workforce analytics, is the practice of collecting, cleaning and analysing data about employees and the workplace to improve decisions about hiring, pay, engagement, performance and retention. It exists on a spectrum:

1. Descriptive analytics: what happened? Headcount, attrition, absenteeism, cost per hire.
2. Diagnostic analytics: why did it happen? Attrition by manager, tenure, location or salary band.
3. Predictive analytics: what is likely to happen? Which teams are at risk of attrition next quarter?
4. Prescriptive analytics: what should we do? Recommended actions, such as a targeted compensation review.

For SMBs, 90% of the value comes from the first two levels, done consistently. Predictive models are exciting, but they need clean data, enough volume and careful governance. Start with the basics and move up only when the foundation is solid.

Why SMBs should care

Large corporates have analytics teams, but SMBs have a different advantage: they are close to the data and can act quickly on it. A few practical benefits:

• Spot problems early: a rise in late arrivals or a drop in leave usage can signal burnout long before resignations arrive.
• Control people cost: salaries, overtime, benefits and attrition-related replacement costs are typically the largest expense lines. Visibility saves money.
• Make fairer decisions: data on pay, ratings and promotions can reveal unintended bias.
• Improve hiring: knowing which sources produce good, long-staying employees focuses your recruitment spend.
• Support the leadership conversation: numbers turn HR from a cost centre into a strategic partner.
• Prepare for audits and inspections: consistent, accurate records are also compliance assets.

Start with questions, not dashboards

The most common failure in people analytics is building dashboards no one uses. Avoid it by starting with business questions. Meet three or four leaders and ask what keeps them up at night. Common themes include:

• Why are new hires leaving within six months?
• Which departments have the highest overtime, and is it sustainable?
• Are we paying competitively, and where are the largest gaps?
• Is our hiring process fast and cost-effective enough to meet growth plans?
• Are absenteeism and leave patterns normal, or a warning sign?
• Do performance ratings differ unusually across teams or demographics?
• What will our payroll cost look like next year under different hiring plans?

Write each question down along with the decision it would inform. If a metric would not change any decision, do not track it.

The core people metrics every SMB should track

Here is a starter set organised by theme. Each is simple to calculate from a standard HRMS.

Workforce and headcount

• Headcount: active employees at a point in time, by department, location, grade and employment type.
• Headcount movement: joiners, leavers and transfers per month.
• Span of control: number of direct reports per manager.
• Contractor ratio: contract and temporary workers as a share of the total workforce.

Attrition and retention

• Attrition rate: leavers during a period divided by average headcount, expressed as a percentage. Our turnover formula guide shows the variations.
• Regretted attrition: loss of employees you would have preferred to retain. See our guide on regretted attrition.
• Early attrition: share of joiners leaving within the first 90 or 180 days.
• Tenure at exit: the average time between joining and leaving.
• Reasons for leaving: categorised from exit interviews.

Hiring

• Time to hire and time to fill, as defined in our recruitment metrics guide.
• Cost per hire across internal and external costs.
• Offer acceptance rate and offer dropout rate.
• Source effectiveness: hires and quality-of-hire by channel, such as referrals, job boards, agencies and campus.
• Quality of hire: performance and retention of new hires after six or twelve months.

Attendance and leave

• Absenteeism rate: days absent divided by scheduled working days.
• Late-arrival rate and regularisation volume, which can indicate process or commute issues.
• Overtime hours per employee, and its share of total hours.
• Leave utilisation: average leave taken versus entitlement, and the liability for unused leave.
• Unplanned leave patterns: such as Monday and Friday spikes.

Payroll and cost

• Total employee cost and cost per employee.
• Payroll as a percentage of revenue where available.
• Payroll variance month over month, with explained drivers. See our payroll variance analysis guide.
• Statutory cost such as PF, ESI and gratuity provisions.
• Compa-ratio: an employee's pay divided by the midpoint of the salary band for the role.
• Pay distribution: by grade, gender and location, to spot gaps.

Performance and development

• Rating distribution by team and manager.
• Goal completion rate.
• Promotion rate and time to promotion.
• Training hours and completion rate.
• Internal fill rate: share of openings filled by existing employees.

Engagement and experience

• Engagement or pulse survey scores, with response rate.
• Employee net promoter score, if you use it.
• Self-service adoption: percentage of employees using the portal and app, a proxy for digital experience.
• Grievance volume and resolution time.

You do not need all of these on day one. A sensible starting dashboard has eight to twelve metrics.

Building your analytics foundation

1. Fix data quality first

Analytics is only as good as the data behind it. Common problems in SMBs include duplicate employee records, inconsistent department names, missing joining dates, outdated manager mappings and attendance gaps. Before you build anything, run a data audit: check for completeness, consistency and accuracy, and fix at the source. Our employee master data audit guide gives a ready checklist. Assign a data owner for each key field, such as HR for designation and finance for cost centre.

2. Standardise definitions

Decide, and write down, how you define each metric. Does attrition include contract-end exits? Does headcount include interns? Are probation terminations counted as voluntary or involuntary? Is time to hire measured from requisition approval or job posting? Without agreed definitions, two people will produce two different numbers and trust will collapse.

3. Choose a single source of truth

Use your HRMS as the primary source. If the data lives in several tools, such as separate attendance software, an ATS and a payroll spreadsheet, consolidate them via integrations or scheduled exports. The fewer manual copy-paste steps, the fewer errors.

4. Build a lightweight toolset

SMBs can do a lot with modest tools:

• HRMS built-in reports and dashboards for standard metrics.
• Spreadsheets for ad hoc analysis, with pivot tables and simple charts.
• A BI tool for interactive dashboards, when volume and complexity justify it.
• Survey tools for engagement and exit feedback.

Do not buy a heavy analytics platform before you have mastered the basics in a spreadsheet.

5. Create a reporting calendar

Consistency builds habits. For example:

• Weekly: operations snapshot, covering attendance, overtime and open positions.
• Monthly: HR dashboard with headcount, attrition, hiring and payroll variance. Our monthly HR dashboard guide offers a template.
• Quarterly: deep dive on a theme, such as attrition drivers or pay equity.
• Annually: workforce plan, compensation review and compliance health check.

Analysing attrition: a worked approach

Attrition is the most requested people analysis, so here is a step-by-step method.

Step 1: Calculate the overall rate

Use a consistent formula, such as leavers in the last twelve months divided by average headcount. Present both total and voluntary attrition.

Step 2: Slice the data

Break attrition down by:

• Department and team.
• Manager.
• Location.
• Tenure band (0 to 3 months, 3 to 6, 6 to 12, 1 to 2 years, more than 2 years).
• Grade or level.
• Age band and gender, with care and respect for privacy and small-number issues.
• Source of hire.
• Compa-ratio band.
• Performance rating.

Look for concentrations. If 40% of leavers come from one team that holds 10% of headcount, you have found a place to investigate.

Step 3: Look at the timing

Plot exits by month. Do they spike after appraisal results, after bonus payout, after festive seasons or after a change in policy? Seasonality is itself a lever, as it helps you plan hiring and retention efforts.

Step 4: Combine with qualitative data

Numbers tell you where; people tell you why. Read exit interview themes, stay interviews and survey comments. Our guide on exit interviews and offboarding shows how to capture useful reasons.

Step 5: Quantify the cost

Estimate the cost of attrition: recruitment fees, interview time, onboarding, productivity loss and any knowledge loss. Even a rough figure helps leadership prioritise.

Step 6: Form hypotheses and test

Examples: "New hires in sales leave because the ramp-up period is unclear." "People in a certain location leave because of the commute and shift pattern." Pick one or two hypotheses and design a small intervention, such as a revised onboarding plan or a shift change, then measure the effect over a quarter.

Step 7: Report and follow up

Present three to five insights with a recommended action and an owner. Review progress in the following quarter.

Analysing attendance and overtime

Attendance data is often the most granular and timely signal you have. Questions to explore:

• Which teams have the highest unplanned absence, and does it follow a pattern by weekday, shift or month?
• Are late arrivals concentrated among specific shifts or commute routes?
• Is overtime concentrated among a handful of employees? That could indicate workload imbalance, understaffing or a risk of burnout.
• Are regularisation requests rising? That may point to biometric problems or a confusing policy.
• Are some employees never taking leave? That may be a sign of overwork, or a culture where people feel they cannot take time off.

Combine attendance with other signals. For example, a team with high overtime, low leave usage and rising early attrition is telling a coherent story about workload. Our guides on chronic absenteeism, overtime calculation and attendance regularisation provide policy-side responses.

Analysing payroll data

Payroll data is rich and reliable because it is audited every month. Useful analyses include:

• Cost drivers: break the month-on-month change into headcount, increments, overtime, allowances and one-offs.
• Pay equity: compare pay for similar roles, grades and experience across genders and other groups, controlling for relevant factors. Where gaps appear, investigate and correct them.
• Compa-ratio and band positioning: identify employees paid well below the band midpoint who might be flight risks, or well above, which may limit future increments. Pair it with market benchmarking, covered in our salary benchmarking guide.
• Increment modelling: simulate different increment budgets and distributions to see the impact on total cost.
• Overtime and shift allowance cost, to decide whether hiring more staff is cheaper than paying overtime.
• Statutory cost forecasting: project the effect of wage ceilings, new rates or headcount changes on PF, ESI and gratuity.
• Anomaly detection: unusual changes such as sudden bank account updates, duplicated names or atypical net pay. This supports the controls in our guide on payroll fraud prevention.

Analysing hiring and onboarding

Treat hiring as a funnel and look for the narrowest points:

• Funnel conversion: applicants to screened, interviewed, offered and joined.
• Stage duration: where do candidates wait the longest?
• Source quality: do referrals produce better retention than job boards?
• Interviewer consistency: are some interviewers systematically harsher or more lenient? Structured scorecards help, as covered in our interview scorecard guide.
• Offer dropouts: analyse by notice period, salary gap and time from interview to offer.
• Onboarding completion: share of new hires completing documentation and training on time.
• 90-day retention: by hiring manager and by source.

These insights directly support improvements in job descriptions, sourcing channels, interview process and the pre-boarding experience.

Workforce planning with data

Workforce planning combines historical data and business plans to forecast headcount needs. A simple approach:

1. Start with current headcount by team.
2. Add planned growth from the business plan.
3. Subtract expected attrition, using your historical rates.
4. Add expected internal movements.
5. Compute the gap and the hiring plan, with lead times from your time-to-hire data.
6. Translate the plan into payroll cost with expected salary levels and statutory costs.

Revisit the plan quarterly. Our workforce planning guide and FY budget planning playbook describe the full cycle.

Predictive analytics: when and how to start

Once you have at least two years of reasonably clean data and a few hundred employee records, you can explore simple predictive models, for example, estimating attrition risk. Start gently:

• Begin with rules, not algorithms. A simple "risk flag" using factors such as tenure, compa-ratio, time since last promotion and recent changes in manager can be surprisingly effective.
• Validate against history. Check how the flags would have performed in past periods.
• Keep humans in the loop. Use the output to prompt a stay conversation, never to penalise an employee.
• Be transparent. Tell employees what data is used and why, in line with privacy law.
• Monitor for bias. Check whether the model's flags differ unfairly across groups.
• Avoid sensitive inferences. Do not use health data, family status or protected attributes as inputs, and do not infer them.

If your volumes are small, remember that statistical models on 50 employees can be noisy. Qualitative insight and manager conversations will often serve you better.

How AI can help, and where it can mislead

AI features in HR tools can speed up analysis: natural-language questions on your data, automatic summaries of survey comments, anomaly alerts and draft narratives for dashboards. They are useful as assistants. But apply caution:

• Check the numbers. AI-generated summaries can misstate figures. Verify against the source.
• Understand the data used. Where does the AI access data, and is it processed in line with your privacy commitments? Our guide on generative AI usage policies in HR is a useful companion.
• Avoid automated decisions about people. Hiring, firing, rating and pay decisions need human judgement and accountability.
• Beware of spurious patterns. With many variables, some correlations will appear by chance. Treat them as hypotheses.

Privacy, ethics and legal considerations

People analytics deals with personal data, so governance matters.

1. Purpose limitation: collect and use data for stated, legitimate purposes, such as improving retention or planning workforce, not for unrelated surveillance.
2. Data minimisation: use only what you need. Aggregated and anonymised views are often enough.
3. Consent and notice: inform employees about what you collect and why, consistent with the Digital Personal Data Protection Act and your internal policies. Our DPDP guide for HR explains the key duties.
4. Access control: limit who can see individual-level data. Managers usually need only their team's data.
5. Small-group protection: do not show metrics for groups so small that individuals can be identified, such as a team of three.
6. Fairness: monitor analyses for discriminatory patterns, and do not use protected attributes to make adverse decisions.
7. Transparency: be open about the use of analytics, and invite questions.
8. Retention and deletion: keep data only as long as needed.
9. Security: protect data in transit and at rest, and log access.

Trust is the foundation. If employees believe analytics is used to watch them, they will resist; if they see it used to fix real problems, such as unfair workloads or pay gaps, they will support it.

Presenting insights so people act on them

A good analysis that nobody acts on is wasted. Use these principles:

• Lead with the answer. "Early attrition in the sales team is twice the company average, driven by unclear ramp-up expectations" is better than a chart with no message.
• Keep it simple. Three to five key insights per report, each with a clear chart or table.
• Use comparisons. Show trends over time, benchmarks and internal comparisons.
• Explain the so-what. Link each insight to a business outcome, such as cost or productivity.
• Recommend actions. Provide options, an owner and a timeline.
• Make it visual, but honest. Use appropriate chart types with clear labels, and avoid truncated axes that exaggerate changes.
• Tailor to the audience. Executives want a one-page summary. Managers want team-level detail. Employees want to see what changes for them.
• Close the loop. Report back on what was done and what changed.

A 90-day roadmap to get started

Days 1 to 30: Foundation

• Interview leaders, and list the top five questions.
• Audit data quality and fix the biggest gaps.
• Agree definitions for core metrics.
• Draft a privacy note describing what data you use and why.

Days 31 to 60: First insights

• Build the starter dashboard of eight to twelve metrics.
• Run the first attrition and attendance analysis.
• Share findings with leadership and agree on one or two actions.

Days 61 to 90: Embed

• Set up the weekly, monthly and quarterly reporting rhythm.
• Train managers to read their team dashboards.
• Pilot one intervention, such as a revised onboarding programme.
• Review what worked, and plan the next quarter's questions.

Common mistakes to avoid

1. Starting with tools rather than questions.
2. Ignoring data quality, leading to reports nobody trusts.
3. Inconsistent definitions, so numbers cannot be compared over time.
4. Too many metrics, which dilute focus.
5. Reporting without recommendations.
6. Over-interpreting small samples.
7. Using analytics to monitor individuals in ways that harm trust.
8. Neglecting privacy, exposing the company to legal and reputational risk.
9. Treating correlation as causation.
10. Not following up, so insights never become action.

How an HRMS supports people analytics

An integrated HRMS is the natural foundation for analytics because it already holds the data and keeps it consistent. Look for capabilities such as:

• A single employee record that connects attendance, leave, payroll, performance and onboarding data.
• Ready-made dashboards for headcount, attrition, attendance and payroll cost.
• Filters and drill-downs by department, location, manager and tenure.
• Scheduled report delivery by email to managers and leadership.
• Role-based access, so each viewer sees only what they should.
• Easy exports for deeper analysis in spreadsheets or BI tools.
• Audit logs that show who accessed or changed data.

CozyHR brings these data sets together so that small teams can answer common people questions without building their own data warehouse.

A worked example: investigating high early attrition

Imagine a 150-person services business that notices five of its last twelve leavers had joined less than six months earlier. The HR lead opens the HRMS and slices the data. Early leavers cluster in two teams and share one hiring source, a recruitment agency used for urgent roles. Compa-ratio analysis shows that these hires were offered pay close to the top of the band, which suggests that the problem is not pay. Attendance data shows high late-arrival rates and frequent regularisation requests in the first month, mostly from one office location with a long commute and an early shift start.

Exit interview themes confirm the pattern: unclear expectations in the first weeks, a long commute and a feeling of being unsupported. The HR lead proposes three small interventions: a structured thirty-sixty-ninety day plan with a buddy for each new joiner, a shift-start change at the affected location for a trial quarter, and a review of the agency's briefing process. After a quarter, early attrition in the pilot teams falls and the lateness rate improves. The numbers are modest, but they were produced with ordinary HRMS reports and a spreadsheet, and the decision trail is clear. That is people analytics at its most practical: a question, a slice of data, a hypothesis, an action and a measurement.

Choosing and designing charts that people understand

Good presentation multiplies the value of an analysis. For trends over time, such as monthly attrition, use a simple line chart. For comparisons across teams, use a horizontal bar chart sorted from highest to lowest. For composition, such as the split of attrition reasons, use a stacked bar or a simple table, and avoid pie charts with many slices. For distributions, such as pay within a grade, a box plot or a dot plot shows spread better than an average. Always label the axes, state the time period and the definition used, and show the number of employees behind each percentage, because a rate of 50% means little if it comes from two people. Keep colours consistent across reports, so that the same department or metric always looks the same. Finally, add a single sentence above each chart stating the takeaway, so readers do not have to guess.

Frequently asked questions

1. Do we need a data analyst to do people analytics?

Not to begin with. A capable HR or finance professional using HRMS reports and spreadsheets can deliver substantial value. Consider specialist support when you have complex questions, large data volumes or a need for predictive models.

2. How many employees do we need before analytics is useful?

Descriptive analytics is useful at any size, since it is simply organised record keeping. Statistical modelling needs larger samples. Below a few hundred employees, favour simple metrics, trends and conversations.

3. Which three metrics should we start with?

Most SMBs get quick value from attrition rate (with early attrition), absenteeism and overtime, and payroll cost variance. Add hiring metrics as you scale recruitment.

4. Is it legal to analyse employee data?

Generally yes, when done for legitimate purposes with appropriate notice, security and respect for privacy law. Avoid sensitive data and covert monitoring, and seek legal advice for complex cases or cross-border data.

5. How do we avoid bias in people analytics?

Use clear definitions, check outcomes across groups, avoid protected attributes and proxies, review models regularly, and keep humans accountable for decisions. Treat analytics as decision support, not decision maker.

6. How often should we review dashboards?

Operational metrics such as attendance weekly, strategic metrics such as attrition and cost monthly, and deeper analyses quarterly. Align reviews with your management meeting rhythm.

7. Can employees see the analytics?

Employees should see their own data and, where appropriate, aggregated company-level information. Sharing high-level insights builds trust, particularly when you explain actions taken in response.

8. How do we handle small teams where individuals could be identified?

Set a minimum group size for reporting, such as five or more, and combine small groups into larger categories. Avoid sharing individual-level data with people who do not need it.

Conclusion

People analytics does not require an army of data scientists. It requires good questions, clean data, agreed definitions, a handful of well-chosen metrics, and a habit of turning insight into action. Start small with attrition, attendance, payroll variance and hiring funnel data. Respect privacy at every step. Present findings so leaders can act. As your data and confidence grow, add deeper diagnostics and, carefully, predictive tools.

If you would like to see how an integrated HRMS can bring your attendance, payroll, leave and hiring data into ready-made dashboards, you can explore CozyHR and try the reports with your own data. And when your analysis touches statutory costs or compliance, verify current rates and rules with official sources or a qualified professional.