Training ROI for Garment Factories: Measure the Benefit That Reaches the Business

Start with the business change the training should make

A sewing team can finish a course, pass a test and still produce no extra saleable garments. Perhaps the next operation remains overloaded. Perhaps the factory has no orders for the added capacity. Training can improve skill without producing the financial return that the budget promised.

A useful training cost-benefit analysis follows the whole chain. Did people learn and use the skill? Did factory results change, and did that change create value? Each link needs evidence.

Training ROI then compares a defined financial benefit with the cost of achieving it. It should have a stated period and viewpoint. A factory may earn a return over four months. A worker's career gain or a wider social benefit has a different scope.

This guide uses an invented garment-factory example. The figures show the method, rather than an industry benchmark. They also show why a lower cost per piece is not always a reduction in the wage bill.

Check that training addresses the actual constraint

Low output does not by itself prove a skill gap. An operator may be waiting for cut panels. A damaged guide or unclear quality rules can also hold up the work. A course will not restore missing material or repair a machine.

Look at the work before choosing the course. Talk with the people doing it and observe the task. Find out when the problems occur. Distinguish what the worker does not yet know from what the system prevents them from doing.

NIST's manufacturing workforce guidance starts with an assessment of goals, challenges and gaps. Its service approach supports choosing training around a defined need. It does not establish a guaranteed productivity gain for a particular course.

Suppose a factory finds that one cell handles and checks its work in different ways. The agreed training covers a defined approved method and practical checks. The cell has enough materials and sound equipment. The next processes have room for its extra good output.

Those assumptions matter to the business case. If they change, the expected benefit changes. Record what else must change for the skill to be useful. Training should not receive all the credit for work done by other teams.

Define the decision and the period

Before calculating a percentage, state what management is deciding. Is it whether to run a pilot, extend a successful programme or continue refresher support? The relevant costs will differ.

A completed pilot is a past cost when deciding on a new rollout. It remains part of the cost when reporting the return on the entire programme. Switching views can make the same project look much better. Explain which view the figures use.

The example covers one complete training package. It counts benefits for four months after training ends. The cost includes its evaluation. No benefit beyond that period is counted, and no grant, tax effect or financing charge is assumed.

The model uses one currency, USD, for convenience. A real factory can use its working currency. It should treat exchange rates and inflation consistently where they affect the comparison.

This is a simple undiscounted comparison over a short stated period. It is not a full investment valuation. For a longer programme, show when cash flows occur and choose a suitable discount rate. Timing matters when one option produces benefits much later than another.

Separate three questions about cost

Managers often need three views of the same programme. One shows the extra cash required. Another shows the staff time and resources assigned to training. The third shows what the business gives up. It compares training with the best other option the factory could actually use.

These views overlap, so they should not be added indiscriminately.

ACCA's relevant-cost guidance asks which cash flows a decision changes. It also considers benefits the business gives up. A share of an existing cost is not automatically an extra cost. Some paid workers may have spare time. Others may have to leave work that would otherwise earn money.

For a training budget, recording paid participant time is sensible. It shows the resources used. Those wages may stay the same with or without training. If so, the course has not caused another cash payment of that amount.

Likewise, lost production is not automatically a cost equal to sales value. Some production can be recovered later. Some has no buyer. When a sale is truly lost, deduct only the costs that are actually avoided to determine the value forgone.

Build the example's cost ledger

Eight operators attend six sessions of two hours each. Total participant time is:

8 people × 6 sessions × 2 hours = 96 person-hours.

At an illustrative payroll rate of USD 4 per hour, that time has a recorded wage value of USD 384. A supervisor also spends twelve hours supporting practice and review. At USD 6 per hour, the recorded value is USD 72.

Regular wages remain unchanged in this example. There is no paid replacement cover or additional overtime. The trainer's fee covers preparation and delivery. The supervisor's time is a separate resource within the factory.

The resource budget is USD 1,200 + USD 456 = USD 1,656. That figure is useful for showing the programme's use of paid resources. It is not the incremental decision cost used below.

No extra room rent is incurred. Another factory may need to hire a room or buy special equipment. It may also pay for cover. Include those costs in the relevant view. An omission cannot be defended by calling the programme “internal.”

Value the production given up during training

Assume that the training schedule causes 120 fewer finished garments to be sold. These sales are lost for good within the order window. There is no spare capacity elsewhere to supply them.

Each sale would bring USD 8. The costs actually avoided when a garment is not made and sold are USD 5.50. These include the relevant material and other variable cash costs. They exclude regular wages, which continue to be paid.

The lost incremental contribution is therefore:

120 × (USD 8 − USD 5.50) = USD 300.

The decision cost becomes USD 1,200 of extra expenditure plus USD 300 of contribution forgone, or USD 1,500.

The calculation allows for wages that continue: it does not deduct them as an avoided cost. The USD 456 values that same paid time. Adding it again would mix cost bases and overstate the extra loss.

A normal contribution figure may deduct labour costs. If those wages will still be paid, add them back before using that figure here. Otherwise, the lost benefit is understated. Follow the actual change between the training and no-training cases.

Suppose the factory could recover the output for an extra USD 180 and keep the sales. That option needs a closer look. It may change the decision cost. The correct answer depends on what the factory would really do.

Measure learning before pricing the outcome

The first check is whether people can perform the intended task. A signed attendance sheet only confirms participation. A written quiz may test knowledge while missing a practical difficulty.

The CDC training-evaluation guide distinguishes learning from transfer to the workplace. It supports assessing both and checking later whether people apply the skill. Satisfaction with a course does not establish its effectiveness.

For the fictional cell, define an observed demonstration against the approved method. Use the same criteria before and after training where that comparison is suitable. Record the conditions and any support provided.

One learner may need a prompt at each step. Another may complete the task unaided. Record that difference without shaming the learner. A need for more training is not misconduct.

Follow-up belongs in the plan from the start. People may understand a skill in class yet struggle to use it during a style change. Normal production pressure may reveal other gaps. Ask what support is missing before assuming the person has forgotten everything.

Compare output carefully

Suppose the trained cell averages 200 accepted finished garments per shift before training. Its later average is 225. A similar comparison cell moves from 190 to 200 over the same periods.

The cells have the same scheduled hours and staffing within each comparison. For this teaching example, their product mix, quality rules and material supply are close enough to compare.

The simple adjusted change is:

(225 − 200) − (200 − 190) = 15 extra good garments per shift.

This is the arithmetic of a difference-in-differences comparison. It removes the common ten-garment improvement observed in the comparison cell.

It does not prove that training caused the remaining fifteen. HM Treasury's evaluation methods guidance explains the need for a sound comparison. Without training, the two groups must be assumed to follow parallel trends.

One before-and-after pair cannot test that assumption. Review several periods, staffing changes, machine work, material quality and other interventions. Perhaps only the trained cell received a new guide. The calculation cannot separate the guide's benefit from training.

The fifteen is a teaching estimate for planning. A stronger claim would need better evidence. A larger programme may warrant a more rigorous design and statistical help.

Keep output units consistent

The benefit calculation counts accepted finished garments at the defined cell boundary. It does not add together every operation completed on the same garment.

One operator attaches a collar; another closes a sleeve. They have completed two operations on the garment, not made two garments for sale. Multiplying both by the selling price would count the same product more than once.

The same risk occurs with rework. A garment returned and repaired should not become a second finished unit in the output total. Keep first completion, rework movement and final accepted quantity distinguishable.

A faster trained operation can also build a queue without raising final output. Check where the extra work ends up. It must reach the measured boundary and pass the remaining processes.

ApparelDiary's line-balancing guide explains how resource constraints affect line output. Use that wider view before turning an operator's time saving into extra sales.

The training evaluation should retain quality and safety checks alongside volume. Skipped checks or an unsafe pace may raise output. That does not make the programme a success.

Limit the forecast to demand that can be served

The planning estimate is fifteen extra accepted garments per eligible shift. Assume it holds for eighty shifts over the four-month benefit period. One eligible production shift occurs per working day.

Potential additional output is:

15 × 80 = 1,200 garments.

The commercial team can support only 1,000 additional sales in that period. These sales are extra to the no-training case. They are not orders brought forward from another month.

The forecast therefore values 1,000 garments, not 1,200. Extra stock without an expected sale is not the same benefit as an extra delivered order.

At USD 8 selling price and USD 5.50 of incremental cash costs, the contribution is USD 2.50 per additional garment. ACCA's contribution explanation provides the basic sales-less-variable-cost relationship.

The example assumes existing paid labour and fixed costs do not rise with these sales. Check all extra costs before relying on that assumption. Include production, packing, transport and selling costs.

Forecast benefit = 1,000 × USD 2.50 = USD 2,500.

This is contribution before recovering the programme's decision cost. Calling the whole USD 8,000 of extra revenue a benefit would ignore USD 5,500 of extra costs.

Calculate ROI and benefit-cost ratio separately

For this article, net training ROI is defined as:

ROI = (benefit − programme decision cost) ÷ programme decision cost × 100.

Using the four-month figures:

(USD 2,500 − USD 1,500) ÷ USD 1,500 × 100 = 66.7%.

The net benefit is USD 1,000. The percentage is a simple four-month forecast under the stated assumptions. It is not an annual return, a guaranteed result or a measured impact from a real factory.

The benefit-cost ratio uses a different numerator:

Benefit-cost ratio = benefit ÷ cost.

USD 2,500 ÷ USD 1,500 = 1.67, rounded.

The CDC economic-evaluation glossary defines the ratio as benefits divided by costs. A value above one means benefits exceed costs on the chosen basis.

A ratio of 1.67 means about USD 1.67 of benefit for each USD 1 of cost. It does not mean a net ROI of 167%. The cost must be recovered before the remainder becomes the net return.

Publish the benefit, cost and period beside either metric. A percentage without those details is difficult to review.

Use payback for timing, with its limits visible

For the timing example, spread the 1,000 extra sales evenly over four months. Assume their contribution becomes available in those same months. Monthly contribution is USD 2,500 ÷ 4 = USD 625.

Assume the USD 1,500 decision cost occurs before the benefit period begins. Simple payback is:

USD 1,500 ÷ USD 625 = 2.4 months after benefits begin.

This includes the USD 300 forgone contribution as well as the extra expenditure. It measures recovery on the chosen decision basis, not only the trainer invoice.

If customers pay later, the cash recovery date will differ. Training may take a month before any benefits start. Include that month when counting time from the programme's launch.

Payback ignores benefits after recovery and does not discount later cash flows. It is useful for understanding exposure, but it should not replace the full period calculation.

A programme with slow initial learning may have uneven benefits. Then add each month's amount to the running total. An average may hide the early delay.

Test the assumptions that could reverse the answer

The largest uncertainty may be demand rather than learning. Even a clear skill improvement has limited financial value if the factory cannot use the capacity.

The table holds the USD 1,500 decision cost and USD 2.50 contribution constant. All scenarios are invented and cover eighty eligible shifts. Additional sales are capped by both output potential and demand.

The break-even number of additional sales is:

USD 1,500 ÷ USD 2.50 = 600 garments.

Below 600, this financial model does not recover the decision cost. At 600, net benefit is zero. This threshold belongs to the example. It is not a training target for each worker.

The model may need more work. Trainer fees could rise, people may need more coaching, or each sale may earn less contribution. Keep costs and benefits connected to the same scenario.

A precise expected return needs evidence about how likely each scenario is. Without that evidence, present the range and leave those odds open.

Avoid counting one improvement twice

Suppose output rises while total regular wages remain unchanged. Wages divided by good output will fall. That can be a useful productivity indicator, but the wage bill has not fallen.

The model may already count the value of extra sales. Adding the apparent unit-wage saving could count the same capacity gain again. Trace each benefit to a change in sales, spending or another defined result.

Likewise, fewer repairs may free time that creates the additional output already counted. Do not add all freed minutes at a labour rate unless they create a separate benefit on the chosen cost basis.

Lower material purchases or outside repair bills may be separate savings. Check whether the contribution estimate already includes them. The extra good-unit count may also capture part of the gain.

The CDC cost-benefit overview states the basic comparison in common monetary units. In a factory, the hard part is often deciding which benefits overlap.

Use a short benefit register with an owner, evidence source, period and overlap check. It should let finance explain each amount without relying on the training team's headline percentage.

Check whether the benefit lasts and who receives it

Do not assume that a short observation continues unchanged for a year. Staff moves, product changes, skill decay and weak follow-up can alter the result. Recheck the skill where it is actually used.

The study The Skills to Pay the Bills evaluated soft-skills training among female garment workers in India. The researchers used random assignment. They tracked productivity, wages, retention and other outcomes.

That study shows the value of a clear evaluation design and multiple outcomes. Its return percentage cannot simply be moved to another factory, course or workforce. The present example does not borrow its effect size.

Factory return and worker benefit should also be examined separately. Workers may gain confidence, safer work or a chance for promotion. Fair access to skills also matters. These benefits need attention even when the ROI does not price them.

Required training and worker protections still apply when the financial forecast is weak. The business case can help choose how to deliver training. It does not give permission to skip duties.

Keep evaluation respectful. Explain what is measured and use relevant facts. Avoid public rankings that treat a person's learning needs as a measure of their worth.

Keep the participant record complete

Keep a record of who was offered training, who attended and what follow-up was available. Looking only at the strongest finishers can overstate the result. Someone may learn and then move to another line. Their later output may fall outside the original cell comparison.

Explain missing observations rather than filling them with a favourable average. A missing quality check is not a passing check. A missing production entry is not proof of zero output. Those problems need investigation before the figures enter the return calculation.

The course record should also identify changes in delivery. Later groups may receive extra coaching or different practice materials. That changes the package they receive. The added cost and support may explain a different result.

This record need not expose personal scores to every manager. A private learner record can guide coaching. Finance may need only grouped figures for its decision.

Make a rollout decision from the evidence

A pilot can justify a larger test without proving that every line will achieve the same result. The next group may have different skills, products or support needs. Trainer capacity can also become a constraint.

Before expanding, bring the evidence together. Review learning, work on the floor, quality and the financial assumptions. A positive forecast with weak evidence may call for another measured stage. A modest return with strong non-financial benefits may support a different decision.

ApparelDiary's capacity investment decision-tree guide can help frame choices under uncertainty. Training evaluation still needs its own costs, outcome evidence and follow-up plan.

The final record should show what was spent and what changed. Distinguish benefits already achieved from those still forecast. It preserves the cost basis and shows how demand limits affected the result.

The next decision-maker can then trace the figures behind the percentage. They can see how training could help the business and what must be true for that gain to occur.

References

• NIST MEP: manufacturing workforce assessment and development.

• ACCA: relevant costs and opportunity costs.

• CDC: evaluating training effectiveness and workplace transfer.

• HM Treasury: evaluation methods, including difference-in-differences.

• ACCA: contribution and cost-volume-profit analysis.

• CDC: economic-evaluation glossary.

• CDC: cost-benefit analysis and net benefits.

• Adhvaryu, Kala and Nyshadham: The Skills to Pay the Bills, December 2017 author paper.