The SaaS / Software Procurement Playbook: The Cheapest Bid Was Not the Cheapest Supplier

An RFP is not a document for comparing suppliers.

It is the instrument that sets the terms on which you will later negotiate, and most buyers give that away in the first section without noticing.

Here’s what that costs.

Three bidders

All technically capable, all through evaluation.

	Bidder A	Bidder B	Bidder C
Year 1 subscription	£380,000	£450,000	£420,000
Implementation	£250,000	£120,000	£200,000
Annual uplift	8%	3% cap	5%
Right to reduce licences	none	15% at anniversary	none

 

On headline subscription, A wins by £70,000 and procurement writes it up as a saving.

Model three years properly, with the uplift compounding and implementation included, and A comes to £1,483,632 against B at £1,510,905. A still wins, by £27,000 across three years, which is inside the noise. But A wins.

Now apply one thing that happens in almost every software estate. By the first anniversary, roughly fifteen percent of seats are surplus. People left, the rollout was smaller than forecast, a department reorganised.

Only Bidder B lets you hand them back.

	Bidder A	Bidder B
Three-year total	£1,483,632	£1,369,769

 

Bidder B, who quoted £70,000 a year more, is now £113,863 cheaper across three years. And the gap widens every year afterwards, because A’s base keeps compounding at eight percent on seats you are not allowed to remove.

The most expensive-looking bid was the cheapest supplier. An evaluation model that scores headline price will never discover that.

Two fixes, both in the bid pack

Weight commercial flexibility at fifteen percent, and score it as separate items rather than as a judgement. Uplift cap out of five. Reduction rights out of five. Exit terms out of five. Notice period out of three. Audit scope out of two. Twenty points, itemised.

Itemised scoring is defensible if your award is ever challenged. A single “flexibility” score is an opinion with a number attached to it.

Mandate the pricing format. Issue a spreadsheet with the rows locked and state in the RFP that submissions in any other format will be returned unread. Then mean it.

Without that, every bidder submits in whatever shape flatters them. One quotes per user per month, one quotes a platform fee plus consumption, one loads implementation into year one so their recurring number looks low. You spend a week normalising and produce an answer you can’t defend.

Two rows in that schedule carry more money than all the others, and almost nobody asks for them. The annual increase percentage. And separately, what that percentage is applied to.

List price, or the amount you actually paid. Those are completely different numbers, they diverge every year, and over five years on a mid-sized contract the difference runs to around £230,000. Asked as a row in a spreadsheet at bid stage it costs nothing. Discovered in the master agreement after signature it is permanent.

Preconditions: entry criteria, not negotiating points

The most underused device in software buying is the precondition. A mandatory requirement a supplier must accept before they can bid at all.

It works because it inverts the normal sequence. Ordinarily the vendor sets the terms of engagement and you react. Preconditions establish that the negotiation happens on your terms, surface the expensive variables before anyone is emotionally committed, and disqualify inflexible vendors before you’ve spent three months discovering the inflexibility.

The ones that earn their place:

Pricing fully unbundled and itemised, because bundled submissions hide margin and prevent line-by-line negotiation.

Disclosure of fiscal year end and current sales incentives, because timing pressure is real and you should not have to guess at it.

Disclosure of standard renewal structures, including annual increases and notice periods.

The contractual definition of each licensing metric, stating explicitly whether service accounts, API calls, contractors and deactivated users are counted.

Acknowledgement that you will negotiate in parallel with multiple shortlisted vendors and compare final offers before award.

Agreement to work from your master agreement rather than theirs.

Removal of all automatic renewal provisions.

Put them in section one with a signature block. Each bidder returns the page signed before receiving the full pack.

Expect two or three to push back. That pushback is information. A supplier who refuses to work from your paper at bid stage will be immovable at contract stage, and you’ve learned it in week two rather than week fourteen.

When not to run one

An RFP costs six to twelve weeks of your time and a month of every bidder’s.

Don’t run one where there is only one viable supplier and everybody knows it. A single-source RFP is theatre, experienced sellers recognise it inside a week, and you get worse pricing than a straight negotiation would have produced, because you’ve spent your only real asset, which is their uncertainty about whether they’ll keep the account.

Don’t run one where the switching cost genuinely exceeds any achievable saving. If migration costs £600,000 and the best plausible saving is £80,000, you’re not running a competition, you’re running an expensive way of confirming your incumbent’s position.

For a renewal where you’d like tension but can’t credibly switch, run a genuine market test on one business unit instead. It costs a fraction, produces real competitor pricing, and the threat is credible because you could actually execute it.