Textile Energy Performance Indicators That Account for Product Mix

A mill improves its operating controls. Yet its electricity use per kilogram rises. The monthly report appears to show failure. Yet the mill also made far more of a product that needs several extra process stages.

The raw ratio cannot separate those changes. Textile energy performance indicators need a clear energy boundary. They also need a sound output measure and a fair basis for the product mix. Otherwise, a change in orders can look like a change in efficiency.

This guide builds an original example with two product families and a fixed monthly load. All figures and model coefficients are hypothetical. They teach the method. They are not textile benchmarks, design targets, or measured factory results.

Choose the question before the indicator

Different measures answer different questions. Total energy shows the amount consumed. Specific energy consumption, or SEC, relates energy to output. A normalized indicator compares actual energy with a reference adjusted for relevant conditions.

An energy baseline is the reference used for comparison. It can be a measured value or a model tied to a defined period. An energy performance indicator, or EnPI, is the measure used to assess the result.

The public scope of ISO 50006:2023 covers how to set up EnPIs and baselines. It explains their use and upkeep. It also covers monitoring performance and showing improvement. It does not supply a universal kWh-per-kilogram target for every mill.

Keep absolute consumption and the normalized result together. A plant can use more total energy yet perform better for its workload. Both facts matter to managers.

Draw the energy boundary

State which meters, buildings, processes, and energy sources belong to the indicator. A dyehouse measure may include process pumps and drying but exclude sewing. A whole-site measure may include offices, utility losses, and outdoor lighting.

Use the same boundary for the baseline and reporting periods. A new workshop may connect to an old meter. This can raise the reading without changing the original process. Record that event before interpreting the trend.

Keep a simple meter map. Show supply meters, internal meters, shared utilities, and any estimated allocations. Reconcile totals where possible. A detailed dashboard built from inconsistent meter boundaries is still unreliable.

The 50001 Ready data-collection guidance starts with the chosen scope. It identifies the energy sources and uses within it. It also calls for relevant-variable data and reliable measurements. Those items should form one collection plan, not separate files with different dates.

Define a production denominator that has meaning

“Kilograms of fabric” can refer to input fabric, processed fabric, or accepted output. It can also reflect different moisture conditions. State the mass basis and keep it consistent.

Choose a clear counting point for site output. It must avoid double counting. Fabric that passes through three machines is not three times the final product mass. A process-stage indicator may count each stage, but it answers a different question.

Keep rework visible. If a batch is dyed twice, its energy belongs in the record. Do not count the same accepted product twice. That can make poor quality look like better energy performance.

Align output and energy dates. Fabric may consume energy late in one month and be accepted early in the next. Where this timing is material, track work in progress or use a suitable longer period. Do not quietly move only the denominator to improve the result.

See how fixed loads distort a simple ratio

Assume a simple process uses 50,000 kWh per month for fixed operating needs. It also uses 1 kWh for each kilogram made. The process and its controls remain unchanged in this small example.

SEC falls by 25% when output doubles. No equipment efficiency changed. The fixed load was spread across more product.

The UNIDO energy-management guide explains this weakness of simple ratios. Base loads and product mixes can both distort them. Use that warning when a report rewards high output months and penalizes quiet months.

The fixed term is not necessarily unavoidable waste. It is a model description for a stated range. Teams may still reduce that load through better controls or equipment changes.

Group products by real process demand

Product families should reflect differences that matter to energy. Check the process route and drying needs. Fabric construction and required treatment stages may also matter. A sales category alone may not describe the work done in the factory.

Start with a small number of defensible groups. Too many groups can leave too little data to estimate each one. Too few can hide large differences and leave the mix problem unsolved.

Keep a mapping table from production codes to energy groups. State who owns it and how new codes enter the table. A blank code should become a data issue, not silently fall into the easiest group.

Use requirements that are known before performance is assessed. If an operator can move a poor batch into a high-energy group after seeing the result, the indicator is open to bias. Group definitions need stable technical reasons.

Build a baseline with explicit units

For the main example, assume a validated baseline has this monthly form:

Reference electricity, kWh = 50,000 + 1.0 × Q_A + 3.0 × Q_B.

Q_A is kilograms of product family A. Q_B is kilograms of family B. The coefficients are 1.0 and 3.0 kWh per kilogram. The fixed term is 50,000 kWh per month within the model’s stated operating range.

These coefficients are teaching assumptions. A real mill needs suitable measured data or engineering analysis. It may need both. They are not universal energy requirements for light and dark shades.

The DOE 2015 baselining guide explains how to track separate product outputs. It also shows comparisons based on modeled energy. Its method is useful technical background; current programme rules need their own check.

Keep the original baseline data and the approved formula. Do not let the coefficients drift each month. They could follow the very performance change the model should measure.

Calculate the product-mix example

In the baseline month, the mill makes 80,000 kg of family A and 20,000 kg of family B. Total output is 100,000 kg.

Baseline reference energy = 50,000 + 80,000 + 60,000 = 190,000 kWh.

Assume actual baseline energy also equals 190,000 kWh. The raw SEC is 1.90 kWh/kg.

In the reporting month, output is still 100,000 kg. The mix changes to 40,000 kg of A and 60,000 kg of B. Apply the original baseline model to this new mix:

Reporting reference energy = 50,000 + 40,000 + 180,000 = 270,000 kWh.

Measured reporting energy is 243,000 kWh. Raw SEC is now 2.43 kWh/kg. That is 27.9% higher than 1.90 kWh/kg, but actual energy is below the 270,000 kWh reference for the new mix.

The report should explain both results. The factory consumed more electricity than in the baseline month. It also performed better than the baseline model predicts for the reporting workload.

Express the normalized result clearly

Use a defined ratio:

Energy performance ratio = actual energy ÷ reference energy.

For the reporting month, the ratio is 243,000 ÷ 270,000 = 0.90. Under this convention, values below one indicate lower energy than the reference.

Normalized improvement = (1 − actual ÷ reference) × 100%.

The example gives 10% improvement. The energy difference relative to the model is 27,000 kWh. It is not a 27,000 kWh reduction from the actual baseline month, which used only 190,000 kWh.

Label the direction of every index. Some organisations set a baseline index of 100; others use an improvement percentage. A dashboard that mixes these conventions can show an upward arrow for both good and bad results.

Use equivalent output with care

The model assigns B three times A’s variable electricity per kilogram. One equivalent-output measure is Q_A + 3 × Q_B.

Baseline equivalent output is 140,000 equivalent kilograms. Reporting equivalent output is 220,000 equivalent kilograms. These are weighted units for analysis. They are not physical fabric mass and must not replace the actual production total.

The fixed 50,000 kWh still matters. Dividing total energy by equivalent output does not remove that fixed-load effect. A model that includes both the weighted output and the fixed term is clearer for this example.

Also keep the weights fixed for the comparison. Do not replace the weights with each month’s actual energy per product. That puts the performance change into the denominator. That can hide the very improvement or loss the indicator should show.

Consider weather and operating schedules

Product mix may not explain every meaningful change. Outdoor conditions can affect cooling. Operating hours can affect ventilation and lighting. Incoming material conditions can affect drying duty.

Check these relationships with process knowledge and data. Do not include every recorded variable just because software allows it. A variable needs a clear purpose and enough reliable variation to assess its effect.

Keep calendar effects visible. A month with fewer workdays may have lower total output but nearly the same overnight load. A daily or weekly model may describe that pattern better than a monthly fixed term.

The 50001 Ready indicator guidance describes several kinds of measure. They include totals, ratios, and numerical models. It calls for normalization when relevant variables significantly affect consumption. Choose the simplest form that fairly answers the factory’s question.

Avoid normalizing away the problem

Some variables describe required work. Others describe losses the team wants to reduce. Treating both as acceptable reasons for higher energy can weaken the indicator.

For example, a required product route may justify extra drying. Unplanned rework may be an outcome to improve. Suppose each rework kilogram raises the allowed energy. The site measure may then fail to reward better first-pass quality.

Use two views when needed. A process diagnostic can show energy per kilogram actually treated, including repeats. A broader business indicator can use accepted output. Relate it to total energy on a consistent basis. Each view needs a clear label.

The same issue arises with idle hours. They can explain consumption, but that does not make them desirable. Choose the purpose. Does the measure assess equipment at a given schedule? Or does it assess how well managers reduce idle operation?

Aggregate periods with the correct weights

Reference energy may differ between months. In that case, do not take a simple average of their improvement percentages.

Suppose month one has reference energy of 270,000 kWh and actual energy of 243,000 kWh. Its improvement is 10%. Month two has a 90,000 kWh reference and 99,000 kWh actual energy. It is 10% worse than reference.

An equal average of the percentages gives zero. The combined result is different:

Total actual = 243,000 + 99,000 = 342,000 kWh.

Total reference = 270,000 + 90,000 = 360,000 kWh.

Combined improvement = 1 − 342,000 ÷ 360,000 = 5%.

The first month carries three times the reference energy of the second. Sum comparable actual and reference values before calculating the combined ratio. Use the same energy basis and valid models for every included period.

Keep electricity, heat, and fuel bases clear

Track each important energy source before making a combined total. Electricity may be measured in kWh. Gas may arrive in volume units and purchased steam in mass units. Conversion needs the correct conditions and energy content.

Use a stated heating-value basis for fuel. Steam mass alone is not an energy quantity. Its pressure, condition, and chosen reference state matter. Keep the conversion method with the data.

Do not count both boiler fuel and the steam made from that fuel in one whole-site energy total. They describe input and an internal transfer. A steam-user indicator can use delivered steam energy. A boiler indicator can use fuel input. Those boundaries serve different purposes.

Site energy, source energy, and emissions also differ. A common unit does not erase the boundary. State each conversion factor and its date. Then compare fuel and electricity changes on that basis.

Separate price effects from physical performance

A lower bill does not prove lower energy consumption. Tariffs, demand charges, exchange rates, and fuel prices may change. Keep cost per kilogram beside physical energy measures rather than replacing them.

Assume electricity falls from 200,000 to 190,000 kWh. Consumption falls 5%. If the energy price rises from $0.10 to $0.12 per kWh, the energy charge rises from $20,000 to $22,800. That is a 14% cost rise despite lower consumption.

At a constant $0.10 price, the physical reduction is worth $1,000. This constant-price view isolates one effect for analysis. The actual bill remains the cash cost and must still be reported.

Likewise, energy divided by sales revenue can fall when selling prices rise. It may be a useful financial measure, but it needs care as a physical efficiency indicator. Use output measures that suit the decision.

Keep on-site generation visible

Solar can reduce purchased electricity without reducing the electricity used by factory equipment. For an electricity-consumption indicator, keep one boundary. Include the relevant on-site generation and exports.

Assume a simple site has no storage and no other source. Grid import is 180,000 kWh, solar generation is 50,000 kWh, and export is 10,000 kWh. The implied site load is 220,000 kWh, subject to the chosen meter and loss boundary.

Suppose the earlier site load was also 220,000 kWh. Lower grid purchases alone then do not show better end-use efficiency. They show a change in supply. Report renewable generation and energy performance as separate measures.

Combined heat and power needs a fuller balance. Track fuel inputs and useful outputs without double counting internal transfers. Do not add generator electricity to its fuel energy and call the sum total site input.

Test whether the baseline still represents the work

A model may work within its observed range and fail outside it. A new fabric family, major capacity change, or new operating mode can require a review.

Compare reporting conditions with the baseline data. A model may be based on small batches of A and B. Do not assume it describes a new product C. A high statistical fit also has limits. It does not prove the model works at half or twice the observed output.

Inspect whether errors follow product mix, season, or output level. A pattern can reveal a missing driver or a poor model form. Investigate meter and production records before treating the pattern as a process fault.

ApparelDiary’s guide to regression and dyeing recipe cost explains broader model checks. Here the outcome is energy relative to a baseline. The model inputs and meaning must suit that purpose.

Match precision to evidence

An indicator calculated to four decimal places may rest on much less precise data. Meter uncertainty, estimated output, missed intervals, and model error all affect the result.

Suppose a monthly comparison suggests a 1% improvement. Check how far plausible data and model errors could move the result. A shift of several percentage points would make that month weak evidence. Investigate the trend and the sources of uncertainty.

Use a data-quality flag with each period. State the share of energy or output that was estimated. Keep the basis for any correction. A corrected record should retain its earlier value and the reason for change.

The DOE discussion of energy-intensity indicators distinguishes efficiency from structural effects. It also explains limitations from incomplete and estimated data. Its framework is broader than one factory. The same caution still helps when reading a factory index.

Compare external benchmarks on a matching basis

Before using an outside figure, check its scope and energy sources. Check the output unit, product mix, and period too. A spinning benchmark cannot directly judge a finishing line. Put final-energy and source-energy figures on a common basis before comparing them.

The IEA indicator overview separates activity, structure, and efficiency when analysing energy trends. That provides a useful reminder: an aggregate ratio can change for several reasons.

Use an external benchmark to ask a better question. A large gap may justify checking meters, process routes, or equipment. It does not prove that the gap is all recoverable waste.

Record the source and its sample limits. A vendor may report an ideal operating test. A study may cover one mill or a broad industry sample. These are different kinds of evidence. Do not invent a percentile or savings potential from a single comparison.

Make the monthly review useful

Show total energy, actual output, and product shares. Add raw SEC, reference energy, and the normalized ratio. Add data-quality flags and major operating events. A short explanation should connect the numbers to a decision.

Assign action to the team that can investigate the cause. Keep a later check of what changed. An index is useful when it guides work, not merely when its colour changes on a dashboard.

Keep baseline revisions traceable

Set review rules before an inconvenient result appears. A major boundary change or a new process may justify a new baseline. A poor month alone does not justify erasing the old comparison.

Retain the old baseline, the new one, the reason for change, and the effective date. State which periods use each. If results link across baselines, document the method. Do not join unlike percentages without explanation.

Keep the fixed reference needed to assess past progress. A rolling model can help forecast future use. Adding recent improved performance changes the question it answers. Forecasting and measuring progress from an earlier state need clearly labelled views.

Frequently asked questions

Is kWh per kilogram always a poor indicator?

No. It can be useful when the process, product mix, and fixed-load effects are stable. Test whether it tracks the question fairly. Keep the raw ratio even when a normalized measure is also needed.

Can energy rise while performance improves?

Yes. More output or more demanding products can raise required energy. A valid baseline can show better performance for that workload. Absolute consumption may still rise. Report both outcomes.

How many product groups should a mill use?

Use enough to represent important differences, with enough reliable data in each. Start from process demand and test the grouping. There is no universal group count that suits every mill.

Should rework be a normalizing variable?

It depends on the purpose. It may help diagnose energy per treatment pass. A business measure may focus on energy per accepted output. Allowing more energy for every repeat can hide a quality loss. Keep both views distinct.

Does a better indicator prove one project saved energy?

It shows performance relative to the stated baseline. Several changes can contribute. Attributing savings to one project needs its own suitable evidence and measurement plan.

Use indicators that explain the work

A good energy indicator makes the comparison fair without excusing avoidable waste. Keep boundaries, units, product groups, and baseline rules visible. Report total consumption beside the normalized result, then investigate what changed. The aim is a measure that managers can use and another reviewer can reproduce.

References

ISO, public scope of ISO 50006:2023.

DOE and Lawrence Berkeley National Laboratory, 50001 Ready Task 8: Energy Data Collection and Analysis.

DOE and Lawrence Berkeley National Laboratory, 50001 Ready Task 11: Energy Performance Indicators and Energy Baselines.

US Department of Energy, Energy Intensity Baselining and Tracking Guidance, 2015 edition.

UNIDO, Practical Guide for Implementing an Energy Management System, 2015.

US Department of Energy, Energy Intensity Indicators Methodology, Caveats, and Cautions.

IEA, Energy End-uses and Efficiency Indicators: overview.