5.2ERP, WMS, TMS & forecasting
The essentials in 30 seconds
- ERP = single base: purchasing, inventory, sales and accounting share the same data — end of re-entries.
- Process → data → tool: clean up and frame before computerising, otherwise you freeze disorder.
- Forecast wrong but useful: measure the error, forecast together (S&OP) and gain responsiveness.
Mini-case thread — Nadia makes her data reliable before the tool
Nadia dreams of software that will fix re-entries and stockouts at her agrifood SME in Algiers. But her items are duplicated and her history is distorted by stockouts. This section shows how she cleans items and units, tests a simple forecast on ten products and launches a mini-S&OP before any ERP choice.
Sequel to case 5.1 — Nadia moves from dashboard to tooling, without skipping steps.
1. ERP, the data core
ERP (Enterprise Resource Planning) unifies all processes (purchasing, inventory, sales, production, accounting) on a single database: the purchase order becomes the expected receipt then the accounting entry — with no re-entry and no gaps between departments.
Market: SAP/Oracle/Microsoft for large accounts, SME ERPs including open source, mostly cloud/SaaS (subscription, no servers).
Master-data quality (items, suppliers, units — 0.4) conditions everything: an ERP fed with dirty data industrialises chaos.
2. Satellites: WMS, TMS, e-procurement
WMS: drives detailed warehousing (locations, paths, waves, scanning, cycle counting) — detail the ERP ignores. TMS: optimises transport (carrier choice, routes, tracking, invoice control). E-procurement/e-sourcing: catalogues, workflows, online tenders/auctions.
Architecture rule: the ERP is the data core (official inventory); specialist tools connect to it.
Warning signs: chained re-entries, conflicting truths, key-person dependency, overflowing reconciliations, EDI/traceability requirements — then the tool pays for itself.
3. Demand forecasting and S&OP
The whole chain bets on the future: forecast from cleaned history (correct stockouts, promos, exceptions) by identifying trend, seasonality and randomness (from spreadsheet moving averages and exponential smoothing to ML models).
Three truths: every forecast is wrong — measure its margin (MAPE): it sizes safety inventory (3.3); forecasting is done together (sales + supply) through monthly S&OP (one shared set of figures); responsiveness (short lead times) reduces dependence on distant sourcing.
Mandatory sequence: process → data → tool. Right-size to your scale (a spreadsheet makes an excellent first inventory ERP in a micro-business) and budget the full cost (configuration, data, training).
🔎 Concrete example — A micro-business moves to an SME ERP
Spreadsheet vs till vs actual inventory: three truths. After cleaning references and units, deployment of a cloud SME ERP (inventory + purchasing + invoicing): re-entries removed, P2P automated, replenishment computed every night.
Key: write the target processes before the tool; the tool freezes whatever you give it.
Forecast: test on 10 major items with smoothing + seasonal profile, measured error and 30-min sales/purchasing S&OP.
Section vocabulary
- ERP
- Integrated management software: all processes (purchasing, inventory, sales, accounting) on a SINGLE database — end of re-entries and conflicting spreadsheets; SAP is the large-account reference.
Simple example: Nadia enters the order once: receipt, inventory and accounting follow with no re-entry.
What it is not: buying an ERP while keeping three parallel spreadsheets "just in case".
Reflex: where is my official data for each item, in a single place?
- WMS
- Warehouse Management System: fine steering of the warehouse (locations, picking paths, waves, scanning, cycle counting) — the level of detail the ERP ignores.
Simple example: Nadia stores by locations with scanning: the picker follows the path, inventory rotates without blocking.
What it is not: managing hundreds of moving references from the storekeeper's memory.
Reflex: does my warehouse need this level of detail, or is organisation enough?
- TMS
- Transport Management System: transport optimisation — carrier choice, routes, shipment tracking, transport invoice control.
Simple example: Nadia compares carriers, tracks shipments and checks invoices in the same place.
What it is not: choosing the carrier by phone, with no record and no invoice check.
Reflex: who chooses the carrier, on what basis, and who checks?
- Demand forecasting
- Projection of future demand from cleaned history (trend, seasonality, randomness); always wrong by nature — its measured error margin sizes safety stock.
Simple example: Nadia forecasts ten star products on history corrected for stockouts, with a seasonal profile.
What it is not: extending raw sales with stockout zeros as if it were demand.
Reflex: is my history cleaned before any calculation?
- S&OP
- Sales & Operations Planning: monthly ritual aligning sales, supply and management on ONE shared set of figures (demand, capacities, inventory) — forecasting as a collective process.
Simple example: Nadia brings sales and purchasing together for 30 minutes: one demand figure, one inventory plan, alerts.
What it is not: everyone forecasting in their corner with their own never-reconciled file.
Reflex: do we have a single shared figure each month?
In practice — Assess your tooling and your forecasts
- Map your tools: where do purchasing/inventory/sales data live? How many re-entries? Where are the two truths?
- Test a simple forecast on 10 major items (moving average/smoothing + season) and MEASURE the error vs the next actual.
- Set up a 30-min monthly mini-S&OP: sales + purchasing around one table — demand, inventory, orders, alerts.
- If planning a tool project (SME ERP, WMS): write the target processes and clean master data before any software choice.
Ready-to-say script — launch the mini-S&OP (1 min)
"30 minutes a month: sales and purchasing around a single table — demand, inventory, orders, alerts. Sales: what are customers seeing? Purchasing: what are lead times saying? We leave with a single figure and one action each."
Say to invite sales and management, then keep the monthly pace without lengthening it.
Copy-paste template — tool framing sheet
TARGET PROCESS (in 3 lines): _______________
WHERE DATA LIVES TODAY: _______________
RE-ENTRIES TO REMOVE: _______________
MASTER DATA TO CLEAN: _______________
NEED (ERP / WMS / simple spreadsheet): _______________
FULL BUDGET (tool + data + training): _______________Small business / tight budget
A maintained spreadsheet beats an empty ERP: Nadia first makes references and inventory reliable before any subscription.
Algeria SME
Before the tool, make units, codes and inventory reliable: inventory gaps distort any automatic order.
Agrifood import
Test the forecast on ten major products with seasonality before extending: the measured error sizes safety.
Growing warehouse
Dedicated WMS only when movements and references exceed what simple organisation absorbs.
✓ Best practices
- Clean history before forecasting: a stockout at zero sales is not zero demand.
- Forecasts are more reliable in aggregate (family/month) than in detail (SKU/week): forecast aggregate, then break down.
- Budget training and data migration: ERP failure is rarely technical, usually human and data.
✗ Common pitfalls
- Computerising a fuzzy process: you freeze fuzziness for years.
- Believing software will create rigour for you: it amplifies it, both ways.
- Neglecting seasonality: the 2-3 year average annual profile is the simplest and most powerful lever.
Key takeaways
- ERP = SINGLE base (end of re-entries); WMS/TMS/e-procurement orbit around it, ERP stays the data core.
- Clean master data conditions everything: a dirty ERP industrialises chaos.
- Every forecast is WRONG: measure its error, build it together (S&OP), gain responsiveness.
- Mandatory sequence: process → data → tool; right-size to your scale.
48h challenge — apply tomorrow
Map where your data lives (purchasing, inventory, sales) and count re-entries, then test a simple forecast on 3 items with cleaned history. Note the gap with actual: by how much are you wrong?
- I know where my data and re-entries are.
- I tested a forecast and measured its error.
- My monthly mini-S&OP is scheduled.
Frequently asked questions
Does an SME need an ERP/WMS?
Yes when symptoms cost more than the tool: re-entries, conflicting truths, key-person dependency, overflowing reconciliations, required EDI/traceability. SME cloud in weeks, progressive modules; dedicated WMS when the warehouse handles hundreds of moving SKUs despite solid 4.1-4.2 organisation.
My forecasts are always wrong: what is the point?
Be less wrong (cleaned history, right aggregate level, seasonality, field input via S&OP), know by how much (MAPE) to size safety, and organise so error costs little (short lead times). Progress comes from these three fronts, not from a crystal ball.
Which sequence makes a tooling project succeed?