What Ariba Guided Buying rollouts get wrong

Guided Buying looks simple in the demo. It rarely stays simple once real catalogues, real suppliers and real approvers get involved.

Guided Buying is one of the easier parts of an Ariba programme to sell internally. The demo is clean: a shopping-cart interface, a curated catalogue, a clear approval path. Everyone in the room nods. Then the project starts, and the gap between the demo and the first live catalogue becomes obvious fast.

The catalogue is the project

Most of the effort in a Guided Buying rollout isn't configuration, it's catalogue management: getting suppliers to provide clean, structured data, agreeing who owns keeping it current, and deciding what happens to the requester experience when a catalogue is incomplete or out of date. Skip this and you end up with a beautifully configured front end pointing at thin, unreliable content — which pushes buyers straight back to email and PDF quotes.

On the London Fire Brigade programme, catalogue and front-end optimisation ended up being where most of the sustained effort went, well after the initial build was technically complete. That's normal. Plan for it rather than treating catalogue quality as a one-off data load.

Approvals need real owners, not just a workflow diagram

It's easy to configure an approval hierarchy in Ariba. It's harder to get the organisation to agree, in writing, who actually approves what value of spend, and to keep that current as people change roles. Guided Buying makes approval gaps visible in a way manual processes never did — which is a good thing, but only if someone owns fixing them before go-live rather than during it.

Training has to include the exceptions, not just the happy path

The demo shows the happy path. Real procurement is full of exceptions — non-catalogue spend, urgent purchases, suppliers who haven't onboarded yet. If training only covers the ideal flow, users hit the first exception, get stuck, and quietly revert to whatever they did before the system existed. Build the exception handling into training from day one, not as a follow-up session after go-live.