When a large technology program runs late or over budget, the post-mortem usually points to delivery: the vendor was slow, the integration was harder than expected, the requirements changed. Look closer and the cause is often older. It was written into the contract months before delivery started.
Where the overrun starts
Three gaps show up again and again:
- Scope described as features, not outcomes. The contract lists modules and interfaces but not the business processes that must work end to end. Every gap becomes a change request.
- Acceptance criteria left for later. If “done” is defined during testing, it is defined by the party that wants to be paid.
- SLAs without consequences. Service levels are listed, but penalties are capped so low, or measured so loosely, that they never change behaviour.
What to fix before award
The cheapest moment to close these gaps is before the contract is signed, when the client still has leverage. In practice that means:
- Describing scope through end-to-end scenarios the platform must support, not only through components.
- Writing acceptance criteria and test responsibilities into the contract.
- Linking a meaningful part of payment to acceptance, not to delivery of software.
- Defining how change requests are priced before anyone needs one.
- Scoring bids on the realism of the delivery plan, not only on price and features.
Why independence matters here
Vendors help write many RFPs, and their advice is often good. But an RFP shaped by one bidder tends to favour that bidder’s product and terms. An independent technical evaluation before award usually costs a fraction of one significant change request, and it is the last point at which the client sets the rules.