Overtime Keeps Rising: Is It the Work or the Habit? 4 Signs to Check
Overtime keeps rising: is it the work or the habit? As of 17 Oct 2026, four signs tell you: overtime starts before the backlog does, the normal shift's first and last hours are slow, the same lines stay back every week, and output per overtime hour is lower. Overtime is paid at 2× (OSH Code s.27), so every habitual hour costs double.
By the Mama Editorial Team. We write about where hours and money leak on Indian factory floors. Last checked: 17 Oct 2026. Every ₹ figure below is either an explicitly labelled assumption or taken from the cited primary source.
Key takeaways
- Overtime is paid at twice the rate of wages. OSH Code, 2020, s.27 sets the rate and requires the worker's consent. An hour that only replaces a lost normal hour is paid twice.
- Four signs point to habit, not work: overtime that starts before the orders do, slow first and last hours, the same lines on overtime every week, and lower output per overtime hour.
- On an assumed 150-worker floor at an assumed ₹580 a day (₹15,080 a month; illustrative), 1.5 overtime hours a day cost about ₹8.5 lakh a month.
- If a third of those hours are habit, about ₹34 lakh a year buys output the normal shift should have made; ₹17 lakh of that is the pure 2× premium (assumptions below).
- The fix is usually at the shift edges and in planning, not in a ban on overtime. Cameras can show whether lines run during paid overtime; they can't tell you who consented or what anyone was paid.
Why overtime creeps up without anyone deciding it
Few owners decide to run more overtime. A rush order needs two extra hours; it ships, but the hours stay. Supervisors learn that a target missed by 5 pm can be met by 7 pm. Workers come to count overtime as take-home pay. Nobody acts in bad faith, yet overtime becomes part of the shift.
This page is about telling those two cases apart. The legal side, including consent, the wage base and the quarterly ceiling, is covered in our guide to Labour Codes for a 200–1000 worker plant, and we don't repeat it here.
What one overtime hour costs, from the law text
Three primary-source facts set the price of an overtime hour.
The rate is 2×. OSH Code s.27 says overtime is paid "at the rate of twice the rate of wages", calculated daily or weekly, "whichever is more favourable to such worker" (OSH Code, 2020). The Code on Wages, 2019, s.14 sets the floor at "not less than twice the normal rate of wages".
Short stays are rounded up. The OSH (Central) Rules, r.69(2), count 15 to 30 minutes of overtime as half an hour and more than 30 minutes as a full hour (G.S.R. 345(E), 2026). Strictly, these rules bind central-sphere establishments; a private factory follows its state's rules, which may compute differently.
The daily rate of a monthly-paid worker is 1/26 of the monthly wage under the same r.69(3). Divide by 8 hours for the hourly rate.
The Factories Act, 1948, which earlier governed factory overtime, was repealed by OSH Code s.143 when the Codes took effect on 21 November 2025. Your state's old factory rules continue only where they don't conflict with the Code.
Sign 1: overtime starts before the backlog does
Real overtime follows the order book, rising after orders jump and falling when they drop.
Habitual overtime doesn't wait. It is booked at the start of the month, on the first day of a new order, or every Saturday regardless of what is in the dispatch plan.
The test. Put two lines on one chart for the last six months: overtime hours per week from the wage register, and the open order backlog in days of output. If overtime rises before the backlog rises, or stays flat when the backlog drops, the work isn't driving it.
The Central Rules ask establishments to keep a "register for wages, over time and deduction" in Form XV (r.72, G.S.R. 345(E)). That register already holds half of the chart.
Sign 2: the first and last hour of the normal shift are slow
If a line loses 20 minutes at the start of the shift and 20 at the end, it loses 40 minutes of output every day. Somebody has to make that output up. The usual answer is overtime, at twice the rate.
The test. For one week, note two times on your bottleneck line: when the first good piece comes off after the shift bell, and when the last one does before the closing bell. Compare them with the overtime hours booked on the same line.
If the line routinely loses 30 to 45 minutes at the edges and then stays back 1 to 2 hours, part of that overtime is buying back the edges. The shift-edge leakage calculator puts a ₹ figure on them.
Sign 3: the same lines stay back every week
Demand moves between products, so real overtime moves between lines.
When the same two or three lines run overtime every week for months, the cause is usually local: an unfixed bottleneck machine, a slow changeover, material that arrives late, or a team that relies on overtime pay.
The test. List overtime hours by line for the last 12 weeks. If the top three lines account for most of the hours in nearly every week, look at those lines' normal shift before you approve more overtime there. The output-below-plan cause tree is a useful checklist for what to look for.
Sign 4: output per overtime hour is lower than per normal hour
Long hours lose productivity. John Pencavel's study of British munition workers found that below about 49 hours a week, output rose in proportion to hours; above that, it rose at a decreasing rate, peaked at about 63 hours, and output at 70 hours was little different from output at 56 hours (Pencavel, "The Productivity of Working Hours", IZA DP No. 8129, April 2014). That is wartime data, not an Indian benchmark, but the shape holds: the last hours of a long day produce less.
The test. For one bottleneck line, divide good output in the normal shift by normal hours, then good output in the overtime window by overtime hours. Use pieces counted at the end of the line, not machine counts.
| Line output (assumption) | Normal shift | Overtime window |
|---|---|---|
| Good pieces per line-hour | 100 | 70 |
| Wage per worker-hour (₹, 2× in overtime) | 72.50 | 145.00 |
| Labour cost per 100 pieces, 20-worker line (₹) | 1,450 | 4,143 |
The wage per hour comes from the worked example below. At 70% of normal output per hour, the labour cost of a piece made in overtime is about 2.9 times the cost of the same piece made in the normal shift.
The ₹ worked example: 150 workers on overtime
All inputs below are labelled assumptions.
The wage. Assume a semi-skilled worker in Noida earns ₹580 a day in basic plus DA, ₹15,080 a month over 26 days (illustrative). Take the current notification for your trade and zone from your state labour department (Uttar Pradesh Labour Department); each state revises the DA part, usually twice a year. The overtime base is "wages" as defined in Code on Wages s.2(y): basic plus DA/VDA, with excluded allowances above 50% of pay added back.
The assumptions. 150 workers stay back for an average of 1.5 hours a day, on 26 working days a month. Daily rate is 1/26 of the monthly wage, hourly rate is 1/8 of that, and overtime is paid at 2×.
| Item | Working | Result |
|---|---|---|
| Monthly wage, basic + DA (₹) | Semi-skilled, assumed ₹580 × 26 days | 15,080 |
| Ordinary hourly rate (₹/h) | 15,080 ÷ 26 ÷ 8 | 72.50 |
| Overtime hourly rate (₹/h) | 2 × 72.50 | 145.00 |
| Overtime hours a month (h) | 150 × 1.5 × 26 (assumption) | 5,850 |
| Overtime wage bill a month (₹ lakh) | 5,850 × 145.00 | 8.48 |
| Overtime wage bill a year (₹ lakh) | 8.48 × 12 | 101.8 |
| Overtime per worker per quarter (h) | 1.5 × 78 days | 117 |
Now split the hours. Nobody knows the true share of habit until the four tests are run, so the table shows three assumed shares.
| Share of overtime that replaces lost normal-shift hours (assumption) | Habitual hours a month (h) | ₹ a year paid at 2× for those hours (lakh) | ₹ a year if the same output came in normal hours (lakh) | Avoidable premium a year (₹ lakh) |
|---|---|---|---|---|
| 20% | 1,170 | 20.4 | 10.2 | 10.2 |
| 33% | 1,931 | 33.6 | 16.8 | 16.8 |
| 50% | 2,925 | 50.9 | 25.4 | 25.4 |
The "avoidable premium" is the second half of the 2× rate. If Sign 4 is also true, the real gap is wider, because you pay double for less.
The 117 hours a quarter sits under the 144-hour reference in a proviso to the Central Rules, r.69(4) (G.S.R. 345(E)), but your state sets the ceiling that binds your factory. Run your own numbers in the overtime cost calculator.
What to do once you know which it is
If it is the work, overtime is a legitimate tool for a short peak. When the peak lasts more than a quarter, compare it with a second shift or more hands. Overtime pays 2×; a shift pays 1× plus hiring, training and supervision.
If it is the habit, don't cut overtime overnight; workers have budgeted for that income. A steadier path:
- Fix the edges first. Start the line on the bell and stop it on the bell, with material and first-piece approval ready before the shift.
- Approve overtime per line, per day, against a named order. "Line 3, 2 hours, for order 4471" can be checked. "All lines, as usual" can't.
- Watch output per overtime hour for a month. If it rises toward the normal-shift rate, the overtime that remains is real.
- Talk to the workforce openly. If take-home pay is the real issue, that is a wage conversation, not an overtime one.
Keep it at line and shift level, comparing the punch record with machine hours, as in biometric hours versus machine hours, rather than chasing individuals.
What cameras can and can't tell you here
What a camera can see. Whether a line is running during the overtime window, minute by minute. When the first and last piece come off in the normal shift. Whether machines stand idle in paid hours. A head count in aggregate, by area, without names.
What a camera can't see:
- Consent. Whether a worker agreed to overtime lives in your records, not in the picture.
- Wages and the overtime register. Rates, hours booked and payments live in payroll.
- Individual pay decisions. Footage should never be used to dock anyone's pay automatically. Deductions follow the Code on Wages process.
Footage of recognisable workers is personal data under s.2(t) of the DPDP Act, 2023. Tell workers in writing what is recorded and why; see the worker CCTV notice under DPDP.
Where Mama fits
Mama reads a factory's existing cameras, or cameras it adds, and every morning sends the owner a short plain-language note on where hours leaked the day before: lines that started late, stopped early, or stood idle during paid overtime. It watches machines and flow, counts people only in aggregate, and never names or ranks anyone. Mama can draw a camera plan for the lines that run the most overtime. Leave an email or phone in the form below.
FAQ
How do I know if overtime is genuinely needed? Chart weekly overtime hours against the order backlog for six months. If overtime follows the backlog up and down, it is driven by work. If it stays high when orders fall, or rises before they do, habit is part of it.
What is the overtime rate for factory workers in India? Twice the ordinary rate of wages (basic plus DA, not gross salary), under OSH Code, 2020, s.27, calculated daily or weekly, whichever favours the worker; Code on Wages s.14 sets the same floor. It needs the worker's consent, and your state's rules set the overtime ceiling.
Is lower output in overtime normal? Some fall-off is expected: Pencavel (IZA DP No. 8129, 2014) found output rose more slowly once weekly hours passed about 49. A large gap on the same line usually points to how the day is planned.
Should I stop overtime altogether? No. Overtime is a legitimate tool for real peaks, and workers rely on it for income. Tie each approval to a line and an order, fix the shift edges, and let the hours follow demand.
Can cameras tell me which workers are stretching hours? That is not what they are for. They show at line level whether work happened during paid overtime. Keep it at line and shift level and give workers a clear CCTV notice.