Manufacturing ERP in the UAE: From Bill of Material to a Cost You Can Defend
· 13 min read · Faceela
Here is a test worth running this week. Pick a product you shipped last month. Ask what it cost to make. Not the standard cost on the price list — what that batch actually consumed in material, labour, machine time and scrap.
In most factories in this country, someone can produce a number within a day. It will be a number built from the bill of material and the standard labour rate, which means it is a restatement of what the product was supposed to cost. The question of what it did cost, and why the two differ, usually goes unanswered — not because nobody is capable of answering it, but because the data required has never been captured in one place.
That is the real reason manufacturers buy ERP. Not inventory visibility, which most companies have well enough. Not production scheduling, which most plants do adequately on a whiteboard. It is the inability to explain a margin movement, which becomes intolerable the moment you take on a customer big enough to negotiate on price.
The three questions a manufacturing system exists to answer
What do we have, and what is it committed to? Stock on hand is easy. Stock on hand net of what is already allocated to open work orders and confirmed sales orders is the number that decides whether you can accept an order, and it is the one that does not exist in a spreadsheet.
What should we make or buy next, and when? This is the demand and supply question. It is answered either by a planning run or by reorder points, and choosing between those two is a bigger decision than most buyers realise.
What did it cost, and where did the variance come from? Material, labour, machine, subcontract, scrap. Broken down enough that a variance points at a cause rather than at a department.
Everything a manufacturing ERP contains is machinery for those three. If a demonstration spends forty minutes on dashboards and has not answered them, you have watched an advertisement.
The bill of material is the system, not a list
The single largest determinant of whether a manufacturing implementation succeeds is the quality of the bills of material, and it is almost always underestimated because the BOMs "already exist" — in the engineering drawings, in the production supervisor's head, in a spreadsheet with a version number in the filename.
Multi-level BOMs, and knowing why each level exists
A multi-level BOM means a component of your product is itself manufactured. That is not an engineering choice; it is an operational one. A level should exist where something is stocked, counted, transferred between departments or sold separately. Every additional level adds a work order, a stock movement and a place for the count to go wrong.
The common failure is a BOM structure inherited from the CAD system, where levels reflect design hierarchy rather than production reality. You end up with work orders for sub-assemblies that are never stocked and exist for four minutes on a bench. The fix is a phantom BOM — a level that explodes into its parent at planning time without generating its own order — and knowing when to use one separates a configured system from an expensive one.
Variants, options and the products that are almost the same
If you make the same item in six colours and four sizes, you have a variant problem, and there are two ways to model it. Separate finished goods with separate BOMs gives you clean costing and an item master that grows without limit. Attribute-driven variants against one template keep the master small and push complexity into the BOM's conditional lines.
Neither is wrong. What is wrong is deciding it after go-live, because migrating from one to the other means re-costing history. This is one of the decisions that must be made by the business before anything is configured, and it is exactly the kind of thing that turns into the mundane data work that quietly kills ERP projects when it is deferred.
By-products, co-products and the ones nobody mentions
Process manufacturers — food, chemicals, plastics, metals — produce things other than the intended output. Offcuts that get remelted. A secondary grade that sells at a discount. Waste with a disposal cost. If the system can only record one output per work order, the value of everything else lands in variance and your product costs are wrong in a direction that flatters you. Ask about this early: it is a capability question with a yes or no answer, and it disqualifies products.
Routings, work centres and where the hours actually go
A routing is the sequence of operations, each on a work centre, each with a setup time and a run time. It is what turns a BOM into a cost that includes conversion, and it is what makes capacity planning possible.
Two decisions do most of the work here.
How granular are the work centres? One work centre per machine gives you precise costing and requires precise data capture from operators. One work centre per department gives you approximate costing and survives contact with a busy shop floor. Start coarser than you think you need. A costing model that depends on operators logging every operation accurately, in a plant where nobody has ever logged anything, produces worse numbers than a simpler model that people actually follow.
Does the operator confirm, or does a supervisor? Real-time confirmation at the machine is the ideal and it is a change-management project, not a configuration setting. The people you are asking to do it are usually the most experienced people in the building, and experienced staff resist for reasons that are normally worth listening to. Plan the rollout around that or the data will be entered in a batch at the end of the shift from memory, which is the same as no data with a worse audit trail.
The work-centre cost rate itself is a management decision: does it include depreciation, energy, indirect labour, maintenance? Whatever you choose, choose once and write it down, because a rate that changes basis halfway through a year makes every year-on-year comparison meaningless.
MRP or reorder rules: the decision most buyers get backwards
Material requirements planning explodes demand — sales orders, forecasts, work orders — through the BOMs and proposes what to make and buy, offset by lead times. Reorder rules are simpler: when stock falls below a minimum, buy up to a maximum.
Vendors demonstrate MRP because it is impressive. Many manufacturers are better served by reorder rules for most of their items, and by MRP for a minority.
| Reorder rules | Full MRP run | |
|---|---|---|
| What it needs to work | An accurate stock figure and a sensible min/max | Accurate BOMs, routings, lead times, and a demand signal worth planning against |
| Best for | Stable consumption, short lead times, low-value components, consumables | Long lead times, expensive components, make-to-order products, dependent demand |
| Failure mode | Excess stock on slow items; stockouts when demand shifts | Plans nobody trusts, so planners override every suggestion and the run becomes decoration |
| What it demands of you | Reviewing the levels once or twice a year | A planner who owns the exceptions daily |
| Realistic time to value | Weeks | Months, and only after BOM accuracy is proven |
The honest sequence is to run reorder rules on the long tail and introduce MRP for the items where lead time and value justify the discipline. A planning run that proposes a hundred orders a day, of which the planner accepts nine, has not automated anything — it has added a filtering job. Getting that boundary right is more valuable than any scheduling feature, and it is the kind of thing a competent ERP implementation scoped against how the plant really runs settles before configuration begins.
Make-to-stock, make-to-order, and the trap between them
Make-to-stock means you build to a forecast and sell from inventory. Make-to-order means the sales order triggers the work order. Most manufacturers are both, on different products, and the system has to hold both without a person deciding each time.
The trap sits between them: configure-to-order, where the customer specifies a combination that has never been made before. That needs a BOM generated at order entry, a cost calculated before you can quote, and a work order that references a specification rather than a stored product. Systems handle this very differently, and some handle it by having a consultant write it for you. If a meaningful share of your revenue is bespoke, this is the capability that decides your shortlist — and it is worth confirming in writing whether it ships in the product or is being built for you.
Fabrication businesses sit at the far end of this. If your output is a structure priced from a schedule of rates and paid against certified progress, you are running a project rather than a production line, and the contractor version of the same problem describes your billing side better than any manufacturing module will.
Scrap, yield and the quantities that never balance
Every plant has a gap between what the BOM says a batch should consume and what it actually consumes. The gap has three sources, and a system that lumps them together tells you nothing.
Scrap is material that entered production and was rejected. It should be recorded against the operation where it was found, because scrap discovered at final inspection and scrap discovered at the first operation are different problems with different costs.
Yield loss is expected and continuous — evaporation, offcut, trim. It belongs in the BOM as a consumption factor, not in variance. If your standard says one kilogram and you always use one point zero four, the standard is wrong and the variance report is lying about a fact everyone on the floor already knows.
Uncounted issue is material that left the store and was never recorded against a work order. It shows up as an inventory adjustment at stock count and gets absorbed into cost of sales. This is the largest of the three in most plants and it is not a costing problem. It is a store-control problem wearing a costing problem's clothes.
Separating these three is the first genuinely useful thing an implementation produces, and it usually happens in the first quarter.
Standard cost or actual cost
This decision determines what your monthly reporting can say, and it is frequently made by whoever configured the system rather than by the finance director.
| Standard costing | Actual costing (average or FIFO) | |
|---|---|---|
| What a product costs | A fixed rate set periodically | What it consumed, revalued as costs move |
| What the month-end shows | Variances, by cause | A moving margin with no explanation attached |
| Best for | Repetitive production, stable BOMs, meaningful volume | Volatile input prices, low-volume or bespoke work |
| The maintenance burden | Someone must re-set standards on a defined cycle | Little, until you need to explain a movement |
| The failure mode | Standards nobody has updated in two years, so every variance is noise | No baseline, so nobody can tell a price change from an efficiency change |
The practical answer for most UAE manufacturers is standard costing on the finished goods where volume justifies it, with a disciplined annual or semi-annual revision and a variance report that names the responsible function. Imported raw material with a fluctuating landed cost — freight, duty, clearing — is the case that most often breaks the model, because the landed cost is known weeks after the receipt.
A variance report that no one can act on is worse than none, because it looks like control. The same trap runs through management reporting generally: a number can be perfectly calculated and still be misleading if the basis under it went stale.
WIP, traceability and quality
Work in progress is material and conversion that has left the store and not yet become finished goods. It is a real balance and it has to be reconcilable to physical reality on the floor at any moment. The common defect is a system where WIP is only relieved when a work order is closed, so a long-running order accumulates cost invisibly and month-end shows a WIP balance nobody can walk out and point at. Ask how partial completion is handled before you ask about anything on the dashboard.
Lot and batch traceability is not optional for food, pharmaceutical, cosmetic or chemical manufacturers, and it is increasingly demanded by large customers everywhere else. The requirement is bidirectional and it has a time limit: given a customer complaint, name every raw material lot in that unit; given a supplier's recall notice, name every customer who received anything containing it. The second direction is the one systems fail at, and the honest test is to ask for it during the demo with a stopwatch running.
Quality checks belong where a defect becomes expensive to fix: incoming inspection, an in-process gate, final release. Attaching a check to every operation produces a plant where operators pass everything to keep the line moving, which is worse than no checks because it manufactures evidence. Fewer checks, at points that matter, with a hold status that genuinely stops the material moving.
Subcontracted operations
Sending material out for plating, heat treatment, machining or coating is normal, and it is where inventory records most often go wrong. Material at a subcontractor is still your stock. It is not in your warehouse, it has not been consumed, and it has to appear on the balance sheet somewhere you can count it.
The requirements are specific: a stock location representing the subcontractor, a document that moves material out without it becoming a sale, a receipt that brings back a different item, and a cost that combines your material with their service invoice. Systems that model this properly make it a routing operation with an external work centre. Systems that do not will have someone raising manual stock adjustments in both directions, which reconciles to nothing.
If a subcontracted step is on your critical path, ask specifically how the planning run treats its lead time. Many do not.
The UAE-specific parts
Landed cost. Most manufacturers here import a substantial share of raw material. Freight, insurance, duty and clearing charges arrive on separate invoices, often after the goods. If those costs are not allocated back to the receipt, your material cost is understated by a variable amount and every margin figure inherits the error.
Free zone and mainland. If you manufacture in a free zone and sell to the mainland, or hold stock in a designated zone, the movement has customs and VAT consequences that the system has to represent as transactions rather than as a note in a file. Establish this in scoping; it is a structural decision about entities and locations, not a report.
E-invoicing. The UAE regime is voluntary from 1 July 2026. Businesses with annual revenue of AED 50 million or more must appoint an Accredited Service Provider by 30 October 2026 and go live on 1 January 2027; below that threshold, appointment is due by 31 March 2027 with go-live on 1 July 2027, and government entities follow on 1 October 2027. Failure to appoint or implement on time carries a penalty of AED 5,000 per month. B2B and B2G are in scope; B2C is not. For a manufacturer this is normally a smaller job than for a contractor, because your invoice is already a structured document — provided your customer and item master data is complete enough to transmit.
Arabic. Bilingual documents and an interface your storekeeper can read are not a cosmetic layer. They determine whether the people entering the transactions can do so without a translator standing behind them.
Dates and thresholds above are from the UAE Ministry of Finance e-invoicing programme, checked on 12 August 2026.
How to test it before you buy
Take one real product with at least two BOM levels and run it end to end in the demonstration.
- Build the BOM with a sub-assembly, and show what happens if that level is made phantom.
- Add a routing with two work centres, setup and run times, and show the calculated cost breakdown.
- Confirm a sales order and run planning. Look at what it proposes and, more importantly, at what it does not.
- Release a work order, consume less material than planned on one component and more on another, and record scrap at an operation.
- Complete a partial quantity and show the WIP balance before and after.
- Send an operation to a subcontractor and bring the material back with the service cost attached.
- Fail a quality check and show that the material cannot move.
- Trace one finished lot back to raw material lots, then forward from a raw lot to every customer.
- Produce the variance report and ask what action each line implies.
The exercise takes an afternoon and it separates products that manufacture from products that do inventory with a manufacturing menu.
The order to do it in
Item master and units of measure first. Every downstream calculation depends on them. This is unglamorous and it is where the schedule is really won.
BOM and routing accuracy second, proven against actual production. Compare what the BOM says a completed batch consumed against what the store issued, for a month, before trusting anything. Until that reconciles, planning and costing are both built on sand.
Shop-floor transaction capture third, at whatever granularity people will genuinely sustain. Removing the paper before the process is agreed just automates a workflow that was never right.
Costing and variance reporting fourth, once the inputs are real.
Planning last. MRP is the capability everyone wants first and the one that requires everything else to be true.
If Odoo is on the shortlist, its manufacturing applications cover most of this natively, which makes the honest assessment of where Odoo fits in the UAE a better starting point than a feature comparison. Should you go that way, take the route that puts configuration ahead of custom code. A heavily modified manufacturing module is the hardest thing in any ERP to upgrade, and it is the reason plenty of plants are still running a version nobody dares touch.