Regional ERP vs SAP for Gulf Mid-Market Manufacturers: Fit, Cost and Time to Value

Use a five-part reality test. The twin must identify a specific physical scope; receive governed operational state at a useful frequency; contain a model that represents relevant behaviour or constraints; compare observed and expected state; and support a named decision such as maintenance timing, parameter optimisation, scheduling or quality intervention. Record latency, missing-data behaviour, model validity and decision owner. Start with one expensive constraint—an oven profile, bottleneck machine, energy-intensive utility or quality-critical process. If a dashboard and threshold rule solve the decision, use them. A twin is justified only when the model changes outcomes enough to repay integration, instrumentation and model maintenance.

The fastest way to expose a fake twin

A vendor displays a polished 3D machine, a live temperature tile and an OEE percentage and calls the screen a digital twin. Ask what physical object it represents, how its state is synchronised, what behaviour the model predicts, how model error is checked and which operating decision changes. If those answers are missing, the plant may have a useful visualisation—but not the maintained operational representation the name implies.

Another plant schedules preventive maintenance outside “planned production time”. OEE rises because the denominator shrinks, but maintenance cost and available customer capacity have not improved.

A third line runs faster, creates more scrap downstream and reports good performance because rejected output is detected after the OEE measurement point.

None of these examples proves OEE is bad. They prove it is being asked to answer the wrong question.

OEE is a structured way to describe equipment loss through availability, performance and quality. Margin is the financial result of price, mix, volume, material, conversion cost, logistics, service and working capital. The two connect through a causal chain; they are not the same metric.

ISO 22400-1, confirmed current in 2025, provides an industry-neutral framework for manufacturing-operations-management KPIs. Its value is disciplined definition and use—not declaring one KPI universal.

Define OEE before debating it

The common structure is:

OEE = Availability × Performance × Quality

Where, under the organisation's controlled definitions:

• Availability compares actual operating time with planned production time.
• Performance compares actual output rate with an ideal or reference rate during operating time.
• Quality compares good output with total output.

The exact data boundaries matter more than the arithmetic. Define:

• asset or line boundary;
• planned production time;
• excluded time and reason;
• when operation starts and stops;
• ideal cycle or rate by product;
• good quantity and quality decision point;
• rework treatment;
• unit of production;
• data source, refresh and owner;
• effect of changeover and planned maintenance;
• effective date of every definition.

Without a metric contract, two plants can report 80% OEE while measuring different realities.

Why OEE does not equal profit

It does not know product margin

One hour of scarce capacity used for Product A may create much more contribution than Product B. OEE sees time and output; it does not know customer price, material cost or mix.

It can reward unnecessary production

Running a machine to keep utilisation high can create stock with no current demand. Output consumes material, labour, energy, space and cash.

It is local

Improving a non-constraint machine may create more queue before the real bottleneck. The line or plant ships no more.

It treats losses multiplicatively, not economically

A one-point quality improvement on expensive material may be worth more than several availability points on a low-cost process. OEE points have no fixed currency value.

Its denominator can be managed

Changing planned time, ideal rate or quality boundary changes the result without changing customer value.

It omits many costs

Premium freight, overtime, tooling, energy peaks, excess WIP, warranty, late delivery and working capital may move while OEE remains stable.

Start with the margin equation

At a useful operational level:

contribution = net sales − truly variable material, processing, freight and selling costs

Then consider conversion resources, fixed and semi-variable costs, depreciation, support and other financial-statement elements under the organisation's accounting policy.

For improvement decisions, ask:

1. Will this increase saleable throughput for demand we can serve?
2. Will it improve product or customer mix through the constraint?
3. Will it reduce actual material, labour, energy, subcontract or quality cost?
4. Will it release working capital or avoid investment?
5. Will it protect price or revenue through better service and quality?

OEE can help explain why throughput was lost. It cannot answer all five alone.

Find the constraint before optimising equipment

The constraint is the resource, market or policy that limits the system's ability to create more value. It can change by product mix, shift or period.

Look for:

• persistent queue before a resource;
• downstream starvation;
• overtime or expediting concentrated in one area;
• customer demand exceeding demonstrated capacity;
• products competing for the same specialised process;
• subcontracting used to bypass capacity;
• quality hold or release limiting shipments;
• engineering, material or labour rather than machine capacity.

Improving constraint availability or yield can move shipments. Improving a resource with spare capacity may only produce WIP faster.

Measure constraint minutes lost by reason and economic impact. One hour on a bottleneck should be prioritised by contribution opportunity and customer commitment, not average machine rate alone.

The metrics that connect operations to margin

Throughput contribution per constraint hour

Calculate expected contribution after relevant variable cost divided by constraint time required. Use it to understand mix decisions where demand and capacity genuinely compete.

It is not a permanent ranking. Customer commitments, strategic products, minimum runs, shelf life and sequencing matter. Document assumptions.

Saleable throughput

Measure good, released output that can satisfy demand—not gross machine count. A unit awaiting inspection or rework has not created shipment capability.

Schedule attainment at the constraint

Did the constraint produce the planned mix and quantity in the planned window? High utilisation on the wrong product is not success.

Material yield and scrap value

Measure good output relative to material input and value the loss. Piece scrap can hide the economics of expensive grades, components or early-stage loss.

First-pass yield

What percentage passes without rework at the intended operation? Final quality rate can look good after costly rework.

Cost of poor quality

Include scrap, rework, sorting, retest, concession, return, warranty, premium freight, line disruption and customer claim where measurable.

Conversion cost per good unit

Track labour, machine, energy, subcontract and relevant overhead per saleable unit, segmented by product and volume context.

On-time-in-full

Customer service links factory output to revenue. Define requested and committed dates separately and prevent promise-date changes from rewriting history.

WIP days and flow time

Excess WIP consumes cash and lengthens feedback. Measure release-to-completion time, queue by operation and ageing.

Cash conversion and inventory

Track raw, WIP and finished inventory tied to production decisions, including slow-moving and customer-specific stock.

Use an economic loss tree

Translate operational loss into decision categories:

Loss	Operational measure	Economic connection
Breakdown at constraint	lost constraint minutes	missed contribution, overtime, service risk
Breakdown off constraint	recovery time and downstream effect	maintenance cost, future constraint risk
Slow cycle	rate loss by product	constraint capacity or labour/energy cost
Changeover	duration and frequency	mix flexibility, batch inventory, capacity
Scrap	material and conversion value	margin and cash loss
Rework	hours, queue and retest	capacity, labour, delay
Material shortage	starved time	expediting and throughput loss
Overproduction	excess output and ageing	inventory, space and obsolescence
Late release	finished stock on quality hold	revenue delay and working capital

Not every lost minute has the same value. Apply economic priority while preserving safety, quality and regulatory controls.

Availability: ask whether downtime cost value

Classify downtime with a reason hierarchy: breakdown, planned maintenance, changeover, material, labour, quality, tooling, utilities, upstream starvation, downstream block and no demand.

“No demand” is not an equipment loss in the same sense as breakdown. Excluding it from OEE may be appropriate under the chosen definition, but management still needs to see unused capacity.

For each major event capture duration, asset, product, shift, cause, action and recurrence. Estimate economic impact at the system level. A ten-minute stop on the constraint during a full order book can matter more than hours on idle support equipment.

Use mean time between failure and repair measures where they support maintenance decisions, but connect them to criticality and lost throughput.

Performance: ideal rate is a governed standard

If ideal cycle time is set to the best moment ever observed, performance appears permanently poor. If reset to current average, improvement disappears.

Set rate by product, equipment, tooling and operating condition through an approved engineering method. Record version and effective date. Separate ramp-up, reduced-rate approval and temporary material limitation.

Analyse small stops and speed loss, not just average rate. Then ask whether recovering rate creates useful flow. Faster output before a blocked downstream process increases queue.

Energy and wear can rise nonlinearly at maximum speed. The economically optimal rate may be below technical maximum if it improves yield, maintenance and energy without constraining shipments.

Quality: measure where the defect is created

Final inspection can detect a defect long after the responsible operation. Assign defect, quantity and cost to the point of creation where evidence allows.

Track first-pass yield, rolled throughput yield across multiple operations, scrap, rework and escape. A line producing 99% good output at each of ten stages has a much lower probability of a unit passing all stages without defect.

Do not improve OEE quality by moving inspection outside the line boundary. Preserve total cost and customer consequence.

Quality decisions need traceability to material lot, machine, tool, recipe, operator, measurement and change. The purpose is controlled learning, not blame.

Changeovers illustrate the trade-off

Long campaigns improve OEE by reducing changeover loss. They also create inventory and delay other products. Very small batches improve responsiveness but can consume constraint capacity.

Choose campaign and sequence using:

• demand and due dates;
• contribution per constraint time;
• changeover duration and matrix;
• shelf life and obsolescence;
• minimum process quantity;
• cleaning and quality requirements;
• raw-material availability;
• finished-goods and WIP targets.

Measure changeover duration and adherence, but evaluate improvement through smaller viable batches, flow time, inventory and service—not OEE alone.

Labour productivity without gaming

Units per labour hour can improve by producing easy items, delaying indirect work or reducing staffing until queues grow.

Use good units or earned standard hours per paid hour alongside overtime, absenteeism, rework, service and safety. Separate direct touch time, waiting, travel, setup and support where actionable.

Automation business cases should show labour redeployed or cost avoided, throughput created, quality improved and maintenance or technology cost added. A theoretical headcount saving is not realised margin until the operating model changes.

Energy and utilities

Track energy per good unit, per operating hour and by production state. Baseload during idle, startup peaks, leaks, poor power factor and compressed-air loss may not appear in OEE.

Normalise for product and conditions. A higher energy-per-unit product mix does not necessarily indicate deterioration. Connect tariff periods and demand charges to scheduling where practical.

Report energy cost and emissions measures separately where required; do not assume they move identically.

OEE by asset can mislead across the plant

Do not average percentages from unlike machines. A simple arithmetic average gives equal weight to a minor tool and the system constraint. Multiplying aggregate availability, performance and quality from inconsistent denominators is also weak.

Use OEE at the defined equipment or line boundary for loss analysis. At plant level, report flow, throughput, schedule, cost, quality, inventory and service. If a composite index is used, publish its weighting and limitations.

Benchmarking OEE across plants is risky unless definitions, products, planned time, ideal rates and quality boundaries are comparable. Internal trend under stable definitions is usually more valuable.

A margin-linked daily board

The shift team needs operational signals:

• constraint schedule and attainment;
• saleable output versus demand;
• lost constraint minutes by reason;
• material shortage and quality hold;
• first-pass yield and scrap value;
• changeover versus standard;
• orders at service risk;
• WIP queue and blocked flow;
• action owner and due time.

Show OEE components for the assets where they help diagnose loss. Do not make “raise OEE” the action. The action should remove a specific cause with an expected effect.

A weekly margin bridge

Bridge expected to actual result through:

• sales volume and price;
• product and customer mix;
• material purchase price;
• material usage and yield;
• labour and overtime;
• machine and energy;
• subcontract and premium freight;
• scrap, rework and warranty;
• inventory and absorption effects;
• delivery penalties or lost sales where evidenced.

Link each material variance to production orders and operational loss. Avoid attributing all margin variance to “efficiency”.

Evaluate improvement projects in money and capacity

For every proposed OEE action, define:

1. loss being removed and baseline;
2. asset and whether it is a current or foreseeable constraint;
3. minutes, yield or rate expected;
4. customer demand that uses released capacity;
5. contribution or cost avoided;
6. inventory and service effect;
7. implementation and recurring cost;
8. safety, quality and maintenance risk;
9. verification period and owner.

If demand cannot use the capacity, value may come from overtime reduction, maintenance window, smaller batches or avoided investment. State which one.

Common OEE gaming patterns

• excluding downtime by reclassifying planned time;
• lowering ideal speed;
• counting reworked units as good without cost;
• moving inspection outside the boundary;
• running unneeded production;
• delaying downtime entry across shifts;
• selecting the best-performing asset only;
• ignoring product mix;
• closing stops with “other”;
• resetting baselines after deterioration.

Protect trust with controlled definitions, reason-code governance, automated event capture where justified, audit trail and reconciliation to production quantity and quality records.

A balanced metric hierarchy

Executive

Contribution, operating margin, cash conversion, on-time-in-full, inventory and major risk.

Plant

Saleable throughput, schedule attainment, cost per good unit, quality loss, WIP, energy and service.

Value stream or line

Constraint output, flow time, first-pass yield, changeover, queue and OEE loss components.

Asset and shift

Downtime reason, cycle loss, defect, maintenance and action.

Every level should drill to the same production events. Do not create separate executive and shop-floor truths.

A 30-day reset for an OEE-heavy dashboard

Week 1

Audit definitions, denominator changes, ideal rates, quality boundaries and data completeness.

Week 2

Identify constraint by product flow and map OEE losses to throughput, cost, inventory or service.

Week 3

Add saleable throughput, schedule attainment, first-pass yield, scrap value, WIP and on-time-in-full. Remove tiles with no decision owner.

Week 4

Select improvement actions using economic impact, run verification and publish the first operational-to-margin bridge.

FAQ

Is 85% a universal world-class OEE target? No. Context, definition, process and economics differ. Use a controlled baseline and loss analysis tied to the system objective.

Should non-constraint machines have OEE targets? They can use OEE for reliability and loss diagnosis, but maximising their utilisation can create excess WIP. Their role is to support flow.

Can OEE be converted directly into money? Not with one universal rate. Value depends on constraint status, demand, product mix, cost and how released capacity is used.

What metric should replace OEE? None alone. Use a hierarchy linking throughput, schedule, quality, cost, inventory, service and cash, with OEE as a diagnostic measure.

Can OEE improve while margin falls? Yes—through unfavourable mix, overproduction, material inflation, excess overtime, quality cost, price decline or improving a non-constraint.

Where a system helps

A manufacturing platform can calculate governed OEE from production events while connecting each loss to work orders, material, quality, labour, energy, schedule and actual cost. Leaders see whether an improvement released constraint throughput, reduced cost or merely changed a percentage.

Explore OptiForge for manufacturing.

Related reading: From Job Cards to Real Costing (KB-449) and Duqm and Sohar Manufacturing Logistics (KB-451).

Sources

• NIST, Digital Twins: Definitions and State of the Art
• NIST, Digital Twins