Does Training Line Supervisors Raise Garment Output? What the Evidence Says
Does training line supervisors raise garment output? Yes, in the best Indian evidence. In a randomised trial across Shahi Exports factories (NBER, June 2023), sewing lines whose supervisors were all trained ran 3.3 efficiency points higher for six months after training ended. Supervisors that middle managers ranked highest for training gained nothing.
Last checked: 20 Oct 2026
By the Mama Editorial Team.
Key takeaways
- The strongest Indian study is a randomised trial of 25 weekly one-hour soft-skills sessions for Shahi Exports line supervisors. Lines with all supervisors trained ran 4.1 points (7.3%) more efficient during training and 3.3 points (5.8%) more efficient in the six months after (Adhvaryu, Murathanoglu and Nyshadham, NBER w31335, June 2023).
- Who gets trained decides the result. In the same trial, supervisors that middle managers ranked highest for training showed no productivity gain, while low-ranked supervisors' productivity rose 12%.
- Better Work's often-quoted "22% productivity" from its Supervisory Skills Training comes from supervisors' own answers about how long their line takes to reach its hourly target (Babbitt, Better Work Discussion Paper 22, Sept 2016). It is a randomised study, but the output figure is self-reported.
- Training alone is not the same thing as a management system. The Bloom et al. trial in Indian textile plants bundled consulting on many practices and raised productivity 17% in the first year (QJE, 2013).
- In an assumed 500-operator unit, the measured 3.3-point effect is worth about ₹30 lakh a year in wages already paid. That is many times an assumed ₹4.5 lakh programme cost, but only if the right supervisors are trained.
The four studies an owner should know
Most claims cite one of four research programmes. They differ in who was trained and how output was counted.
| Study (year) | Where | Who was trained, how long | How output was measured | Effect on output |
|---|---|---|---|---|
| STITCH, Shahi Exports (NBER w31335, June 2023) | Indian garment factories, 1,849 supervisors randomised, 561 lines | Line supervisors, 25 weekly 1-hour sessions | Daily line efficiency from factory records (pieces ÷ target) | +4.1 pts during training, +3.3 pts (5.8%) for 6 months after |
| Better Work SST (Discussion Paper 22, Sept 2016) | Cambodia, Haiti, Indonesia, Jordan, Lesotho, Nicaragua, Vietnam; 776 supervisors at baseline | Line supervisors, about 3 days | Supervisor's own report of minutes to reach the hourly target | Target reached 11.44 min faster, computed as +21.5% hourly output, driven by female supervisors |
| Bloom et al. (QJE 2013; NBER w16658) | Large Indian woven cotton-fabric firms, 28 plants of 17 firms | Whole plant management, 1 month diagnosis + 4 months implementation | Plant output, quality and inventory records | +17% productivity in year 1 (published version) |
| PACE, Shahi Exports (JPE, Aug 2023) | Five Bengaluru-area garment factories | Sewing operators (not supervisors), 80-hour programme | Operator-level production records | +13.5% productivity; 256% net return to the firm 8 months after |
Only the STITCH trial trained supervisors and measured line output from factory records. The largest headline number, 22%, rests on self-reports.
What the Shahi STITCH trial found
Shahi Exports, described by the authors as India's largest ready-made garment manufacturer with roughly 70 factories, ran the trial with researchers from the University of Michigan and UC San Diego (NBER w31335, June 2023).
The curriculum covered communication, planning, problem-solving, stress management and motivating workers, over about nine months on average.
For comparison with your own floor: before training, from January to March 2017, the average line in the study ran at 54.96% efficiency against a budgeted 60.60%.
During training, lines with all supervisors trained ran 4.1 percentage points above lines with none trained. That is 7.3% of the control-group level.
For six months after training ended, the gap was still 3.3 percentage points, or 5.8% of the control level. The effect peaked about four months after training began.
Workers shared in the gain. Treated lines were 3–4 percentage points more likely to earn any production bonus, which the paper puts at 38–51% of the baseline rate.
Supervisor retention improved as well. Trained supervisors were more than 15% less likely to quit than untrained ones.
The paper found no effect on supervisor attendance, worker attendance or worker retention.
On money, the authors used the firm's revenue and margin data to estimate a net profit of $4,460,669 to the firm up to six months after training. That is $4,807 per trained supervisor, after counting training costs, faster salary growth for trained supervisors and higher bonuses.
Why the supervisors your managers pick may gain nothing
The researchers asked middle managers to rank their supervisors by who would benefit most from training. Access was then randomised within each ranking group (NBER w31335, June 2023).
Highly recommended supervisors showed no significant productivity gain. Low-recommended supervisors' productivity rose 12% against controls.
The managers were not wrong about their people. They were solving a different problem. In the untrained group, highly recommended supervisors were 15% more likely to quit, a gap the paper reports as not statistically significant (p = 0.29). Training cut quitting among highly recommended supervisors by 22% against 3% for the low-ranked group.
A floor manager who loses a supervisor must cover the line and train a replacement, so managers used training to keep flight risks.
The authors calculate that, for the managers' choice to make sense for the firm, each avoided supervisor exit would need to be worth $31,000–$62,000. That is about 11–23 times a supervisor's average annual earnings in the study.
The practical lesson: if your production manager draws up the training list, the list may be optimised for keeping people, not for output.
What the Better Work SST evaluation does and does not show
Better Work, the ILO and IFC programme, ran Supervisory Skills Training in seven countries and commissioned a randomised evaluation from Tufts University (Babbitt, Discussion Paper 22, Sept 2016).
Supervisors queued for training were randomly assigned to be trained immediately or a few months later (surveys June 2014 to November 2015).
The productivity measure was a survey question. Supervisors with hourly targets were asked how long their workers took to reach the target on the last full working day, adjusted for changes in target size.
Training received several months earlier cut this adjusted time by 11.44 minutes against an average of 53.24 minutes. The paper converts that into a 21.5% rise in hourly output, and the effect was driven by female supervisors.
Perceived manager support mattered for other outcomes: supervisors who felt the least support for the training were the most likely to see higher turnover several months after it.
The Better Factories Cambodia course sheet rounds the result to "a 22 % increase in productivity" for lines with trained female supervisors (Better Work SST flyer).
The design is sound, but the outcome is a supervisor's own estimate of yesterday's line, not a production record. Treat 22% as an upper signal, and 3.3–4.1 points from Shahi's records as the planning number.
As of 20 Oct 2026, India is not among the country programmes linked from the Better Work home page, so SST is not offered through a Better Work India team.
Training is not the same as a management system
The Bloom et al. experiment is often cited in the same breath, but it tested something larger. A consulting firm spent a month diagnosing and four months implementing a set of practices at randomly chosen plants: quality tracking, maintenance, inventory rules, daily review and incentives (NBER w16658).
The firms averaged about 270 employees and were family-owned. The published paper reports a 17% productivity rise in the first year (QJE 128(1), 2013); the working paper's estimate was 11%.
The longest-run evidence on supervisor training is historical. Bianchi and Giorcelli found that US firms in the 1940–45 Training Within Industry programme, which taught supervisors to instruct, improve methods and handle people, were 6% more productive within a year and 27% after ten years (NBER w28833, 2021; published in JPE 130(6), 2022).
No randomised trial of supervisor training in Indian knitwear, Tirupur or Ludhiana units was found in NBER, Better Work or J-PAL publications (checked 20 Oct 2026). The Shahi evidence comes from the garment lines of one large exporter.
What it could be worth on your floor (₹)
The arithmetic uses one rule from our 1% of sewing efficiency piece: one efficiency point recovers 1% of the attended minutes you already pay for, so it is worth about 1% of the sewing wage bill.
Assume a 500-operator sewing floor, 10 lines, 20 supervisors and assistants, and ₹15,000 per operator per month all-in. That is a ₹9 crore annual sewing wage bill. Replace the wage with your own state's notified rate and market pay.
Assume a programme of 25 one-hour weekly sessions like STITCH, run in-house by an external facilitator at ₹15,000 per session (₹3.75 lakh), plus 500 supervisor-hours away from the line valued at about ₹150 an hour (₹75,000). Total assumed cost: ₹4.5 lakh.
| Scenario (assumed 500 operators, ₹9 crore wage bill) | Efficiency gain (points) | Wage-minutes recovered (₹ lakh/year) | Assumed programme cost (₹ lakh) | Recovered ÷ cost |
|---|---|---|---|---|
| Shahi average, 6 months after training | 3.3 | 29.7 | 4.5 | 6.6× |
| Half the effect (some trainees gain nothing) | 1.65 | 14.9 | 4.5 | 3.3× |
| Manager-picked list, high-recommended only | 0 | 0 | 4.5 | 0× |
The table counts only payroll; output valued at your cut-and-make rate is worth more.
The last row is the risk. In the Shahi trial the effect came from supervisors managers did not prioritise, so a hand-picked list can cost the full ₹4.5 lakh and move nothing.
The National Productivity Council's self-paced e-LEAP course "Supervisor and his Job" is listed at Rs.250 (NPC e-LEAP catalogue, checked 20 Oct 2026). A short online module is not the same dose as 25 facilitated sessions, so do not expect the STITCH effect from it.
How to run it so you can tell if it worked
Train in two waves. Put half the supervisors in wave one and half in wave two four to six months later, chosen by lottery within each floor, not by the production manager. Better Work's evaluation used the same waitlist idea.
Fix the measure before you start. Use daily line efficiency from production records for each line, for three months before training, as Shahi did.
Compare wave-one with wave-two lines over the same months, so season and buyers cancel out.
Watch quits as a second outcome. In the Shahi trial training cut supervisor quitting by more than 15%, which has its own value if supervisors are among your hardest people to replace.
Check floor time too. Training does little if supervisors still spend half the shift in the cabin; our piece on how many supervisors a 400-worker plant needs prices that leak.
What cameras and data cannot tell you here
No camera measures communication, problem-solving or how a supervisor treats a worker. These are the skills STITCH and SST taught.
Line data alone cannot prove why a line improved; only the wave design separates training from a new style or season.
What machine-level observation can add is an objective version of the measures Better Work asked supervisors to self-report: how often work piles up at stations, and how much time machines sit idle waiting for work. Counted the same way before and after, those are harder to argue with than a supervisor's memory.
Any such measurement should count lines and machines, not rank people. Workers need a clear notice under DPDP; see the worker CCTV notice template.
Where Mama fits
Mama reads a factory's existing cameras and sends the owner a short morning note on where machine hours were lost yesterday, by line, not by person. For a training wave, that gives a before-and-after count of idle and waiting time on trained and untrained lines. It fits inside the wider picture of making more from the lines you already pay for. To see what that would look like on your floor, leave an email or phone in the form below and Mama will get back to you.
FAQ
Does supervisor training increase garment factory productivity? In the largest Indian randomised trial, at Shahi Exports, lines with all supervisors trained were 3.3 efficiency points (5.8%) more productive for six months after training ended (NBER w31335, June 2023). The gain came almost entirely from supervisors that middle managers had not ranked highly.
Is the Better Work 22% productivity figure reliable? It comes from a randomised evaluation (Better Work Discussion Paper 22, September 2016), but output was measured by supervisors' own reports of minutes to reach the hourly target. Use it as an upper signal and production-record results such as Shahi's 3.3–4.1 points for planning.
How long should a supervisor training programme run? The Shahi STITCH programme ran 25 weekly one-hour sessions, about nine months on average, and its effect peaked around month four. Better Work SST ran about three days, with effects measured several months later.
Should managers choose which supervisors get trained first? In the Shahi trial, the supervisors managers recommended most showed no productivity gain, because managers used training to keep likely quitters. A lottery within each floor, in two waves, gives a fairer test and a cleaner result.
Is there a Better Work programme in India? No India country programme is linked from the Better Work home page (checked 20 Oct 2026). Indian units can use the published SST and STITCH course outlines, or National Productivity Council e-LEAP supervisor modules.