ERP and operations

What an ERP actually is

One system holding the operations a company runs on: orders, stock, purchasing, finance, sometimes production, projects and people. The point is a single record rather than five systems arguing about the same number.

That single record is the benefit and the cost. Everything connects, so everything has to agree, and the discipline that requires is why these projects are hard.

Accounting alone belongs in finance and accounting. Stock alone is inventory. Customer relationships sit in CRM and sales.

When one system beats several

Connected specialist tools are usually better for a small company. Each does its job well, each is easy to replace, and the connections are somebody’s monthly maintenance rather than a project.

ERP starts paying when reconciliation becomes the job. Stock counts that disagree with the ledger. Orders retyped between systems. A month-end close that takes two weeks because three sources have to be aligned by hand.

The other trigger is structural: several legal entities, several currencies, several warehouses, or a manufacturing process where costing depends on material movements. Those are hard to hold together with connectors.

Modules, and buying only what you run

• Finance, the ledger everything else posts into.
• Purchasing and suppliers, with approvals and receipt matching.
• Inventory and warehouse, including locations, lots and serial numbers where relevant.
• Sales and fulfilment, from quote through invoice.
• Production, if you make things rather than move them.
• Projects, service or field work, depending on the shape of the business.

Vendors sell suites and price modules separately. Map the modules you will genuinely use in year one, get those quoted, and treat the rest as a later decision rather than a discount to chase now.

Implementation is the real product

Software is a fraction of what an ERP costs. The rest is configuration, data migration, integration and the time your own staff spend away from their jobs.

Data is where projects die. Stock records with no owner, a customer list with four spellings of the same company, and a ledger with historic adjustments nobody can explain will all have to be cleaned before they can be loaded. Start that work early, because it does not depend on which product you choose.

Ask the vendor who does the implementation. A partner network means variable quality and a separate contract, and the partner matters more to your outcome than the logo on the software.

Customisation, upgrades and the long term

Every customisation is a tax paid at every upgrade.

Distinguish between processes that give you an advantage and processes that are simply how it has always been done. The first are worth building around. The second should bend to the software, and the resistance to bending them is the most predictable source of overrun.

Ask how upgrades work: automatic, scheduled or a project each time, and what happens to your configuration when they run. Hosted products upgrade on the vendor’s calendar, which removes a decision and adds a dependency.

Cost, contracts and getting out

Licences are usually per user per month with tiers by role, since a warehouse scanner does not need a finance seat. Implementation is quoted separately and commonly runs one to three times the first year of licences.

Ask three commercial questions before signing. What the price is at renewal rather than at signature. What data extraction looks like if you leave, in full and in a readable format. And what support is included, since named support is frequently an extra tier.

The habits behind quoted numbers in software generally are described in what pricing pages hide.

Signals that a project is going wrong

These implementations rarely fail suddenly, and the warning signs are consistent enough to be worth naming.

Data cleaning that keeps being postponed until later in the plan. A growing list of customisations, each defended individually and never counted together. Key staff attending workshops while still holding their full workload. A go-live date that has not moved even though three earlier milestones did.

Any two of those together mean the plan has stopped describing reality.

The response is unpopular and effective: cut scope rather than move dates. A first phase covering finance and one operational area, working properly, beats a complete deployment that arrives late and is trusted by nobody. The remaining modules go in afterwards, from a position where the system already works.