Visit Website →
← Back to blog

Overtime Forecasting: How to Predict and Prevent It

October 29, 2026
Overtime Forecasting: How to Predict and Prevent It

Overtime forecasting means projecting which employees will cross the overtime line, and what that will cost, before the hours are worked. You do it by combining scheduled hours, hours already clocked this workweek, expected demand and time-off patterns, then changing the schedule while there is still time to change it. Done weekly, it turns overtime from a surprise on the payroll report into a decision you make on Monday.

This guide shows what data to use, how to run the math by hand, how to read the forecast, and how to check that it is working.

Heyhive
Keep Overtime in View When Scheduling
Heyhive generates weekly shifts while accounting for employee availability, certifications, and overtime limits, with managers approving schedules before publication.
Explore Heyhive scheduling

TL;DR

  • Forecast overtime per employee, per workweek, not as a single monthly budget number. Overtime is triggered by individual hours.
  • Start with three inputs: hours already worked, hours still scheduled, and demand you expect but have not yet scheduled.
  • Under federal law, covered non-exempt employees earn at least 1.5 times their regular rate after 40 hours in a workweek. Your state may add daily rules.
  • Act on the forecast early. Planned overtime costs less and disrupts fewer lives than last-minute overtime.
  • Judge the forecast by what it changes: fewer surprise overtime hours, and no chronic understaffing hiding behind the "approved" ones.

Table of Contents

What overtime forecasting is and why it matters

Most managers see overtime in one place: the payroll export, after the pay period closes. At that point the money is spent. A forecast moves the same information earlier. Operationally, that means projecting each employee's hours through the pay period and flagging anyone trending toward overtime before those hours are worked, as Axiom's workforce guidance describes it.

There are two levels worth separating:

  • Employee-level forecast. Who will hit 40 hours this week, and by which day?
  • Budget-level forecast. How many premium hours, and how many dollars, should we expect across a location or a season?

The first protects you week to week. The second helps with planning. Research on government overtime, for example, breaks overtime into seasonal, trend and leave-driven components so that budget alerts fire before money runs out, as Mannava (2026) describes. Small businesses rarely need machine learning for that. A spreadsheet with the same three components gets you most of the way.

The rule your forecast is built on

The federal baseline is simple. Under the Fair Labor Standards Act, covered non-exempt employees must be paid at least one and one-half times their regular rate for all hours worked beyond 40 in a workweek, according to the U.S. Department of Labor. Two details matter for forecasting:

  1. The workweek is fixed. The Department of Labor defines it as a fixed, regularly recurring period of 168 hours, seven consecutive 24-hour periods, and it does not have to match the calendar week (same fact sheet). Your forecast must use the same start day and time as your payroll workweek, or every projection will be off.
  2. Most hours count. Hours worked include time on duty or at the workplace. Short rest periods of 20 minutes or less are compensable hours, per the Department of Labor. A forecast that quietly subtracts short breaks understates the real total.

State rules can be stricter. Some states count overtime by the day, not only by the week, and for some occupations, such as home care, where both federal and state law apply the employer must follow whichever pays the employee more. Check your state labor department's current rules, then build them into the forecast as a second trigger.

Classification matters too. The Department of Labor recovered $358,200 in back wages for 31 Dairy Queen managers who were entitled to overtime because their pay fell below the required minimum salary for managers. Before you forecast, confirm who is actually non-exempt. For a deeper walk through the rules, see our FLSA overtime guide for employers.

What data to use

A good overtime forecast needs inputs from three groups.

Hours you already know

  • Clocked hours so far this workweek. This is the most valuable number you have, and it is only reliable if clock-ins are accurate. Our guide to employee attendance tracking for hourly teams covers how to get there.
  • Scheduled hours still to come. Published shifts, including shifts picked up through swaps.
  • Approved time off. It lowers an employee's projected hours and often shifts them onto coworkers.

Demand you expect

Overtime is usually a coverage problem, not a people problem. Labor scheduling research from Cornell's Gary Thompson lays out eight steps for forecasting demand, including identifying labor drivers (the thing that creates work, such as covers served, deliveries or service calls), testing whether they vary with time, and tracking forecast errors. For a restaurant the driver is hourly sales or covers. For a field crew it is booked jobs and their estimated durations. For a retailer it is traffic and transactions.

Three labor demand drivers by business type

The federal government's own workforce planning guidance points the same way: demand analysis looks at future tasks, roles and staffing requirements and includes benchmarking, according to the Office of Personnel Management.

History and benchmarks

  • Last year's overtime by week, to find seasonal peaks.
  • Overtime by employee, to find the few people who absorb most of it.
  • Outside context. The Bureau of Labor Statistics publishes average weekly hours and overtime by industry, which gives you a sense of where your own numbers sit in your sector.

How to forecast overtime by hand: a worked example

Here is the method with a small, made-up kitchen team. The numbers are illustrative, not data from any real business.

It is Wednesday close. Your workweek runs Monday to Sunday. A line cook earns $20 an hour.

ItemHours
Clocked Monday and Tuesday17
Clocked Wednesday so far8
Scheduled Thursday to Sunday4 shifts x 8 = 32
Projected total57

The cook is headed for 17 hours of overtime. Overtime pay is $20 x 1.5 = $30 an hour, so the premium portion alone is 17 x $10 = $170 above straight time. Multiply that by the three or four people in the same position, and you see why the weekly view matters.

Now the forecast does its job. You look at Thursday to Sunday and ask:

  1. Is the demand real? If Sunday's sales forecast is soft, drop a shift.
  2. Who has room? Another cook at 24 projected hours can take the Saturday shift.
  3. Can the shift be shorter? Trimming two hours of overlap on Friday may remove the risk.

Repeat this for every non-exempt employee each day, and you have an overtime forecast. Software just runs the loop faster.

Reading the forecast: planned versus reactive overtime

Not all overtime is a failure. The goal is to know which kind you are buying.

One workforce management reference describes overtime as authorized when actual or forecast staffing falls below the net requirement and other options cannot close the gap. It separates reactive overtime, found through real-time monitoring, from proactive overtime, planned one to seven days ahead when short-range forecasts show demand above scheduled coverage (WFM Labs). The same source argues proactive planning is generally preferred because staff can plan their personal time and you avoid the premiums often needed to get last-minute agreement. Treat that as a practitioner view rather than a law of nature, but it matches what most managers experience.

A practical triage works well:

  • Green: projected under 36 hours. No action.
  • Yellow: projected 36 to 40 hours. Watch it. One late closer or a sick call tips it over.
  • Red: projected over 40, or over your state's daily threshold. Decide now: reassign, shorten or approve on purpose.

Then look at patterns, not just weeks. The same WFM Labs reference notes that persistent overtime signals systematic understaffing, while persistent voluntary time off suggests overscheduling relative to demand and possibly a forecasting accuracy or shift design issue. If the same location is red every Friday, you do not have an overtime problem. You have a Friday staffing problem.

Turning the forecast into schedule changes

A forecast that nobody acts on is a report. These moves convert it into savings.

  1. Rebalance before publishing. Check projected hours as you build the schedule, not after. Our guide to avoiding scheduling overtime explains why systems beat simply adding headcount.
  2. Spread open shifts across people with room. Sort by projected hours and offer the shift to the lowest first.
  3. Cross-train for the peaks. More people who can cover a station means more candidates under 40 hours.
  4. Cap approval at the manager level. Overtime should need a yes from someone who can see the forecast.
  5. Re-forecast midweek. Call-outs and demand shifts change the answer. Run the numbers again Wednesday.

For a longer playbook, see how to prevent overtime scheduling and cut labor costs.

Tying the forecast to payroll

The forecast only holds up if the hours behind it match what payroll pays. Mismatched workweek definitions, missing punches and manual timesheet corrections are the usual causes of a forecast that looks fine and a paycheck that does not. If you are tightening the payroll side, an overview of online payroll for small businesses can help you think through how overtime flows from timekeeping into pay. Also compare timesheets and a time clock, because the way you capture hours decides how early you can see trouble.

Perspective: why we build the forecast into the schedule

We see overtime forecasting fail for one reason more than any other: the forecast lives somewhere other than the schedule. A manager exports hours, opens a spreadsheet, spots a problem on Thursday, and by then half the week is worked.

Perspective: why we build the forecast into the schedule — overview diagram

So we build the other way around. In Heyhive, scheduling respects employee availability, certifications and overtime limits as the schedule is generated, and you approve it before anyone sees it. The forecast is not a separate report. It is a constraint on the draft. GPS-verified clock-ins feed real hours back in, so the numbers you review are the numbers that happened.

We also believe a person should make the final call. AI can build a full week of shifts in seconds, but you know the new hire is shaky on the grill and the regular who needs hours. That is why nothing publishes without your approval.

— Heyhive

How to tell whether your forecast is working

Do not judge a forecast by whether it was exactly right. Judge it by whether it changed decisions. Track these each month:

  • Surprise overtime hours. Overtime that was not flagged at least two days ahead. This should shrink.
  • Planned versus actual hours by employee. Large, repeated gaps point to bad demand inputs or missed clock-ins.
  • Overtime concentration. If three people carry most premium hours, the problem is distribution, which you can fix.
  • Overtime by day and location. Persistent spikes mark understaffed shifts, not unlucky weeks.
  • Forecast error on your labor driver. Thompson's research includes tracking forecast error as a formal step (Cornell). If sales forecasts are off by a wide margin, fix that first.

If you want these figures in one place, our piece on attendance data reporting shows what to pull and how to read it.

Forecast overtime inside your schedule

Heyhive's AI Scheduling, GPS Time Clock and payroll-ready hours work together: generate the week, keep every shift inside your overtime limits, verify who actually showed up, and export clean hours to payroll. You review and approve before anything is published.

Heyhive

If you manage hourly teams across restaurants, retail floors or field crews, start free and see your next week's schedule built around your overtime limits.

FAQ

What is overtime forecasting?

It is projecting each employee's hours for the workweek, using clocked time, scheduled shifts and expected demand, so you can see who will cross the overtime threshold before it happens. You then adjust the schedule or approve the overtime deliberately.

How many hours of overtime is too much?

There is no federal cap for most adults, so the answer is about cost and people. Persistent overtime usually signals systematic understaffing, and sustained long weeks tend to hurt reliability. If the same people are over 40 week after week, treat it as a staffing signal and review coverage.

How do I calculate overtime pay?

Multiply the regular hourly rate by 1.5 and apply it to hours beyond 40 in the workweek. At $20 an hour, overtime is $30 an hour. Your regular rate can include more than base pay, so check the Department of Labor's overtime fact sheet and your state's rules.

Do I need software to forecast overtime?

Not for a handful of employees. A weekly spreadsheet works if clocked hours are accurate. Software becomes worth it when you have several locations, frequent shift swaps or field teams, because projections need to update as hours change.

Sources