Inside the Operations Dashboard That Tells a Bangladesh RMG Factory What It Can Actually Promise Next (Part 2)

An operations dashboard should show what a factory can make and what stops it. Track output, faults and machine losses on a clear time basis. Use the records and the supply chain dashboard in Part 1 to check the next shipment promise.

Start with the order that needs a decision

Imagine a sewing department reporting 68% efficiency while one line runs at 45%. The department figure may be correct. It can still hide an order that needs attention. Open the line record and check the style, available labour, output and remaining work. Then test the route to shipment. Another stage may still be the true constraint.

The first question is not who caused the red number. It is what has changed and whether the change affects the promise. Check for a changed seam, missing attachment or late trim. An old time standard may also need review. Each cause needs its own remedy. The same low efficiency figure cannot identify which one applies.

Each alert should lead to the affected order and quantity. Show its event time, expected result and recorded exception. A supervisor can investigate the immediate cause. Central Planning can estimate the effect on later stages. Sales can then discuss a supported recovery option with the buyer when needed.

Show when the event occurred and when the screen was refreshed. A machine count captured at noon but uploaded at four o'clock is not a live noon-to-four view. If the connection fails, retain the last valid reading with its timestamp and make the missing period visible.

Give every metric a clear scope

The public description of ISO 22400-2 links manufacturing KPIs with defined formulas, elements, time behaviour and units. That is a sound discipline for a factory dashboard. A percentage becomes useful only when people agree what its numerator and denominator contain.

A sewing labour measure does not measure fabric yield. A dyeing machine's batch output cannot be compared directly with pieces from a packing cell. Keep the units appropriate to the process, then connect each result to the order's remaining load.

Keep a calculation definition beside the report. Include how breaks, overtime, repairs, transfers and missing records are handled. Changing a definition can improve the reported result. Production may not have changed at all. Preserve the earlier rule when comparing historical periods.

Calculate line and department efficiency correctly

For an agreed good-output basis:

Efficiency = earned standard minutes ÷ available operator minutes × 100

For one style, earned minutes equal qualifying pieces multiplied by the approved SMV. For several styles, calculate each product separately and add its earned minutes. The SMV guide explains the work standard that supports this numerator.

Available minutes should match the people and time in the report. Ten operators may work a full period while two join halfway through. Use their actual included minutes. A headcount at the end of the shift does not show the whole day's labour input.

Consider two illustrative lines. Line A earns 6,000 standard minutes from 10,000 available minutes, giving 60%. Line B earns 12,600 from 18,000, giving 70%. Department efficiency is (6,000 + 12,600) ÷ (10,000 + 18,000) × 100 = 66.43%. The simple average of 60% and 70% is 65%, which gives the smaller line too much weight.

For mixed output, suppose a line completes 300 good pieces at 12 SMV and 200 at 18 SMV. It earns 3,600 + 3,600 = 7,200 standard minutes. With 10,000 available minutes, efficiency is 72%. Do not multiply all 500 pieces by one standard. That would distort the result.

Count each qualifying piece once. A garment that returns from repair and passes another check is not a second finished garment. State whether the chosen efficiency measure earns work at a particular operation or at completed-garment release. Use a clear method for partial progress. Do not add the same work again at the next reporting point.

Use the result fairly

A line comparison should account for style, method, material, staffing and stage of learning. A difficult first day and a stable repeat order are different operating conditions. Keep those conditions with the record before using a rate in an operator or supervisor review.

The dashboard can reveal where support is needed, but it does not establish individual fault. Let people challenge an incorrect assignment or count. Keep machine faults, poor incoming parts and changed methods visible. An unexplained personal score can hide these causes.

Keep cutting yield separate from labour output

Three compatible orders of 800 pieces contain 2,400 pieces in total. A shared cutting plan may reduce repeated setup and allow a more useful size mix. More plies alone do not improve marker efficiency, which depends on how pattern area fits within the marker area.

Before combining orders, check fabric width, shade and lot rules. Also check pattern direction, size ratios and due dates. Safe lay height and equipment limits still apply. One combined plan may require several physical lays. The RMG cutting process guide follows these controls through production.

Suppose separate plans require three setups of 25 minutes, while a compatible combined plan requires two. The illustrative setup saving is 25 minutes. If marker utilisation remains 84% in both cases, the evidence supports a setup benefit but no marker-yield gain. Measure each benefit on its own basis.

Release the cutting plan only after the inputs are checked. A reused marker may fit yesterday's width. It may not fit today's usable width. A higher marker percentage is of little help if the resulting parts violate shade or direction requirements.

Read DHU alongside defective units and repair work

DHU = defects found ÷ units inspected × 100

DHU counts defects, while defective percentage counts garments with at least one defect. A garment can contain several defects, so the two measures can differ greatly. DHU can exceed 100 without an arithmetic error. The detailed DHU guide develops this distinction.

Suppose inspectors check 250 garments and find 30 defects on 20 garments. DHU is 30 ÷ 250 × 100 = 12. The defective share is 20 ÷ 250 × 100 = 8%. Neither result tells the team how many garments will be repaired successfully or how long the repairs will take.

Keep first inspection and repeat inspection records distinct. A repaired garment may return for another check. State how that repeat check affects the denominator and trend. The same care applies when comparing inline checks with final checks, because their purposes and sampling coverage may differ.

A count-based Pareto chart ranks open seams first. The other two categories each need 40 repair minutes. They rank equally on that measure. Severity requires another view: a rare critical defect may need immediate action even when its frequency and repair time are low.

Treat a fishbone diagram as a list of possible causes to test. Observe the operation, inspect the material and check the machine condition. Do not turn a category such as method or operator into a finding without evidence. Use a comparable sample after the change to see whether the problem has improved.

Inspection coverage belongs beside the trend. A lower DHU may reflect a better process. It could also reflect a different style mix or weaker inspection. The statistical process control guide explains why stable definitions and comparable groups matter when interpreting a chart.

Calculate OEE from the underlying equipment record

Overall equipment effectiveness has three factors: availability, performance and quality. In the conventional calculation, availability is run time divided by planned production time. Performance compares the ideal time for total output with actual run time. Quality is first-pass good output divided by total output.

Siemens' OEE formula documentation shows these relationships. Define the planned production window and ideal cycle before collecting the data. The ideal cycle needs a sound process basis. Do not choose it just to improve the percentage.

OEE = availability × performance × quality

Use fractions in the product and multiply by 100 to display a percentage. For example, 95% availability, 70% performance and 90% quality produce 0.95 × 0.70 × 0.90 = 59.85% OEE. Reporting only 95% uptime would miss two losses. The machine also ran slowly and produced some non-good output.

A second worked example starts with raw counts. A machine has 480 planned production minutes and stops for 30, leaving 450 run minutes. It produces 750 units at an ideal cycle of 0.5 minute per unit. Of those, 720 are good on their first pass.

The cross-check is first-pass good output times ideal cycle divided by planned time: 720 × 0.5 ÷ 480 = 75%. Use full precision in the calculation. Round the display at the end. Rounding each factor first can alter the final product.

Turn the percentage into a loss explanation

The example starts with 480 minutes. Stop loss takes 30 minutes. Performance loss takes 75 ideal-time minutes, and non-good output accounts for another 15. Fully productive time is 360 minutes. These four values add to 480, which makes the loss structure easier to understand.

That accounting does not diagnose the cause. Slow running can involve feeding, material, methods or equipment condition. Stop records may include missing inputs as well as breakdowns. Review the reasons before choosing a maintenance action. The maintenance programme guide helps connect the failure pattern with an appropriate response.

A performance result above 100% needs a check. Examine the ideal cycle, unit conversion and counts. It may indicate an outdated reference. Do not simply cap the number and discard the underlying discrepancy. For mixed products, use each product's own ideal cycle. An unweighted average may give the wrong result.

Reuse line layouts without freezing old assumptions

A base layout records machine order, tasks and staffing. It also shows the expected flow of work. It can reduce repeated planning for a familiar style family. The useful part is the known method and resource requirement, not merely the drawing.

For a new style, identify added, removed and changed operations. Check seam construction, fabric handling, attachments, machine type and standard time. A small design change can affect the operation that already limits the line. Check the old SMV in a trial. Otherwise the booking estimate may look too certain.

The industrial engineer needs a valid method and balance. The planner needs an early load estimate and a learning allowance. The supervisor needs the machines, people and first-off checks ready. Operators need clear instructions and an explained standard. All four roles should use the same approved version. Give each role the detail it needs.

Keep task precedence and station limits intact when testing a new balance. The line-balancing guide discusses how those constraints affect allocation. Moving a task on a chart does not make the physical sequence feasible.

Before bulk output, confirm labels, trims, settings and the approved first-off piece. Record what remains unresolved. A short start check can prevent much larger repair work, but the actual saving needs evidence from the case. The dashboard should show whether the check occurred, not claim a fixed return for every sign-off.

Show short-order changeover and learning separately

A fixed changeover consumes more time per piece on a short order. Suppose setup occupies 120 line minutes and stable output is 100 pieces per hour. A 500-piece order needs five stable production hours, so setup is two of the seven elapsed line hours, or 28.57%. For 5,000 pieces, it is two of 52 hours, or 3.85%.

This simplified example excludes ramp-up and interruptions. It illustrates why a quantity change can affect the commercial case even when the standard sewing content per piece stays the same. Do not add the setup allowance again if the cost or capacity model already includes it.

Measure changeover from a defined ending event on the old style to a defined starting event on the new one. Define the stable rate used to measure ramp-up. The first good piece is one event. The first full hour at the planned rate is another.

A reusable layout may reduce preparation. A compatible marker may reduce cutting work. Neither guarantees that every small order will be profitable. Use the measured setup, learning pattern, contribution and route constraints when deciding which short orders fit the factory.

Link nine operating views through shared records

The operations team needs detail, but it need not place every detail on one screen. A manager should see exceptions. Each should lead to the evidence needed for a decision. The nine views below can share order, style, work-centre and event keys.

Use consistent keys when joining the views. A garment counted at sewing, finishing and packing is moving through stages; it is not three separate saleable garments. A repair return is another event for the same item. Preserve that identity so the dashboard can reconcile movement without inflating output.

A WIP count also needs age and status. Fifty pieces awaiting a routine transfer differ from fifty held for unresolved shade approval. Group the queue by its next action, then show who can release it. A total alone may encourage the team to move work without resolving the hold.

Reconcile one shift before joining the whole plant

A stage balance can reveal missing or repeated events. Suppose sewing opens with 120 pieces, receives 800 cut sets, transfers 760 completed pieces and records ten pieces as scrap. With no other movement, the closing quantity is 120 + 800 − 760 − 10 = 150 pieces. Compare that calculated balance with the physical count before explaining the difference as lost production.

Keep repair work inside the correct boundary. A garment moved from one sewing station to another has changed location, but it has not left the sewing department. If the departmental report treats both moves as an external transfer, its closing balance will be wrong. State whether a record represents a piece, a bundle or a set of components, and use a valid conversion before adding them.

Late capture needs a clear rule too. An offline device may upload yesterday's events this morning. Use the event date to place work in the right period, while retaining the upload time to explain the delay. A repeated upload should not create a second movement. A user may correct a mistaken count. Keep a link between the old entry and the correction.

A shift reconciliation is particularly useful during a pilot. It checks the physical meaning of the data before the same events feed efficiency, WIP and shipment forecasts. A smooth-looking chart can otherwise spread one counting error across several reports.

Allow for the next style's learning pattern

A stable efficiency rate may be too high for a new style's first days. Consider three days with 8,000 available labour minutes each. Suppose the supported planning rates are 50%, 65% and 75%. Expected earned minutes are 4,000, 5,200 and 6,000, giving 15,200 over the three days.

Applying 75% to all 24,000 available minutes predicts 18,000 earned minutes. The difference is 2,800 standard minutes, or 46.67 SAH. These are illustrative rates, not a standard learning curve. Use comparable styles and the actual planned team to estimate the pattern.

Update the forecast during the run. If the first day misses its rate, check why. The same reduction may not apply to every later day. A missing attachment that has now arrived differs from a seam that still needs training. The revised forecast should reflect the remaining problem and the evidence for recovery.

Let Central Planning revise the next promise

Central Planning should reconcile the accepted order with the current floor plan before offering the next slot. Use the same style, quantity, due date and version on both sides. When a forecast becomes firm, replace the matching reservation rather than adding the order as a second load.

Floor results can change the expected capacity. Suppose the next five days provide 20,000 labour minutes per day. At 65% expected efficiency, the plan offers 65,000 standard minutes. If comparable evidence now supports 60%, it offers 60,000. The difference is 5,000 standard minutes, or 83.33 SAH.

The planner must check whether this evidence fits the coming work. A temporary material hold may not justify lowering every future style's expectation. A method problem that persists may justify a change. Keep the reason and affected period with it.

Do not multiply an effective labour-capacity estimate by DHU or OEE simply because all three appear on the dashboard. They may describe overlapping losses on different bases. DHU is a defect frequency, not a yield fraction. Past good-output efficiency may already reflect repairs and downtime. Another deduction can count the same loss twice.

Test the whole route by date. Cutting capacity cannot replace a finishing shortage, and spare sewing hours cannot release a held fabric lot. A revised schedule should show which order moves and which resource changes. State what recovery depends on. A broad monthly capacity total is only the first check.

Keep packing, quality release and dispatch as distinct events. An improved efficiency forecast must not mark the order as shipped. Only the appropriate shipment evidence should close that part of the commercial promise.

Read Bangladesh productivity evidence with care

The following external context was checked on 22 October 2026. The BIDS study on firm-level upgrading reports an average 63% increase in subprocess productivity between 2014 and 2023. It describes output per hour and uses 2023 data with recalled 2014 data, alongside interviews and discussions.

That finding gives a useful reason to examine individual processes. It does not establish that a monitoring package will raise every factory's efficiency by 63%. Product mix, skills, equipment and organisational changes can all affect the comparison. Earned-minute efficiency also has a different basis from simple output per hour.

A 2024 study of ICT implementation examines Square Fashions and Northern Toshrifa Group through before-and-after gross sales and net profit. The approach highlights the need for a baseline. It does not isolate the software's contribution from every other business change.

Before a factory pilot, define the intended benefit. Measure output, labour, defects and response times on a comparable basis. Record investment, running cost and added data-entry work. Explain changes in style mix or staffing during the pilot. Show the original level and measurement basis with each gain. A percentage alone is hard to assess.

For example, output rising from 80 to 100 units per hour is a 25% relative increase. Efficiency rising from 60% to 75% is also a 25% relative increase, but the change is 15 percentage points. Preserve both the original and final levels so a headline cannot blur these meanings.

Use RFID as part of a reconciled shipping process

RFID can connect physical cartons with packing and movement events. The tag identity must match the read location and business event. GS1's EPCIS overview describes a common model for sharing visibility events across systems and organisations.

Consider an illustrative manifest containing 40 unique carton IDs. The gate system records 43 reads, but three are repeat reads of cartons already counted. Removing repeat reads may leave 40 unique IDs. That still does not prove the set is correct. One expected carton could be absent while a carton from another shipment is included.

Compare the actual set with the manifest. Report expected IDs not observed and observed IDs not expected. Investigate missed reads, stray reads, damaged tags and physical loading before closing the exception. A count of 40 against 40 is weaker evidence than a complete identity match.

Record the event time, location and shipment reference. A packing event establishes packing, while a dispatch event establishes the recorded outbound movement. A tag by itself does not prove the garment's fibre origin or confirm arrival at the buyer's warehouse. Those claims need the appropriate evidence and handoffs.

Prepare for Digital Product Passports on the correct timetable

The European Commission's textile apparel page lists Q4 2027 as the planned adoption date for the textile delegated act. This is a planned legal step. It is not a deadline for every garment to carry a passport during 2027. The final act will define the textile-specific requirements.

The Commission's general DPP guidance distinguishes the registry from product rules and describes a transition period of at least 18 months after ESPR delegated acts. Before claiming compliance, check the final product act and its start dates. An operational registry does not complete the rules for every sector.

Factories can prepare reliable records now. Map product and material identities, composition evidence, production events and supplier handoffs. Keep the evidence behind an origin claim. Define who can read, correct and export each record, and how a correction remains traceable.

A QR code provides a route to information; it does not validate that information. A blockchain can preserve an entry. It cannot prove that the entry matched the physical material. Neither technology removes the need to reconcile quantities and assess supporting evidence.

BGMEA announced its agreement with AWARE on 10 May 2026, followed by an introductory pilot session on 30 September. These are readiness initiatives. Use the Commission's legal timetable when evaluating broader deadline claims in industry announcements.

In the announced approach, material batches receive digital tokens and later producers add production information. A useful pilot should test whether those handoffs remain complete and whether a finished order can be traced to its supported inputs. Check correction rights, access, exportability and continuity if a service provider changes. Joining a pilot does not prove compliance. The future product rules are not yet final.

Choose the first digital investment from a real gap

If material holds repeatedly surprise sewing, connect receipt and release events first. If the factory overbooks specialist equipment, improve dated capacity and resource records. If cartons are misallocated, test manifest reconciliation at packing and dispatch. These are different problems and may need different capture methods.

A controlled manual pilot can establish the required events before sensors or software are selected. Test normal work and awkward cases: an offline reader, a duplicate event, a repaired item, a partial order and a cancelled booking. Confirm that a user can reach the underlying record and correct an error without losing its history.

The manufacturing analytics stack guide connects source capture with reporting. An ERP can reduce repeat entry, while specialist systems may supply machine or inspection events. Check whether the combined records agree. They must also arrive in time for the decision.

Questions to settle before the next booking

Ask whether the accepted order version matches the floor plan, whether the required material is usable and whether the next work centre is actually free. Check the skills and attachments, the changed layout and the remaining finishing window. Review the latest comparable efficiency evidence without counting the same losses twice.

For a changed or cancelled order, identify the time and stock that are truly released. Buyer ownership, quality holds and other reservations may prevent immediate reuse. Record the effect of each override on neighbouring work so the planner can explain the new promise.

A daily exception meeting can resolve immediate output and quality issues. A weekly planning review can update future assumptions and short-order decisions. After shipment, compare the forecast with what happened. Use the actual route and outcome to improve the next estimate.

The operations dashboard is most useful when it connects a measured event with a practical decision. It shows what the floor has completed, what still needs attention and what the next promise depends on. Central Planning can then carry that evidence into the commercial conversation with confidence that the underlying work has been checked.

See also the buyer’s journey in Bangladesh (Part 2).

References and further reading

• ISO 22400-2 — Manufacturing operations management KPIs.

• Siemens — OEE standard formulas.

• BIDS — Choice of techniques and firm-level upgrading in Bangladesh RMG.

• Nipa — ICT implementation case studies in Bangladesh garment factories.

• GS1 — EPCIS overview.

• European Commission — Textile apparel and the Digital Product Passport.

• European Commission — Digital Product Passport framework.

• BGMEA — May 2026 AWARE agreement.

• BGMEA — September 2026 DPP pilot session.