The 5 Questions Your Spend Data Must Answer Before You Source Anything
Most sourcing failures happen before the first RFQ goes out. Here are the 5 questions to answer first. With AI doing the legwork, answering all five now takes weeks, not months.
I have watched sourcing waves fail at consulting firms, at e-sourcing platforms, and inside organisations I ran. Almost never because of bad negotiation. Almost always because the groundwork was skipped.
Before you run any sourcing event, your spend data must answer five questions.
1. What do we actually spend, by category?
Not the budget. Not the GL summary. Actual invoiced spend, categorised by what was bought. Most companies discover their “category view” is really a vendor-name view, and the same item hides under five different cost lines. Messy data is fine; uncategorised data is not.
2. Who are we concentrated with?
In my experience the 80/20 rule holds with remarkable consistency: roughly 80% of spend sits with about 20% of suppliers. Know your 20%. They are your leverage, your risk, and your first conversations.
3. Where has competition died?
Pull the contracts that auto-renewed. The suppliers who have not been challenged in 3+ years. The single-source categories that became single-source by accident, not strategy. Stale supply relationships are where the 15–25% gap concentrates.
4. What does the supply market say the price should be?
Internal data tells you what you pay. Only market data tells you what you should pay. Supply market research used to take a junior analyst weeks per category. With AI doing the research legwork, a solid category brief now takes hours. There is no excuse to negotiate blind anymore.
5. Which stakeholders can kill this?
Every category has an internal owner who can veto a supplier switch. Identify them before the wave, not during it. A saving that operations refuses to implement is worth exactly zero.
Answer all five and your first sourcing wave is aimed at the right categories, armed with market facts, and pre-cleared internally. Skip them and you will run a competitive event on a category that did not matter, with stakeholders who were never on board.
The good news: with AI doing the categorisation and market research, getting from raw AP data to all five answers now takes weeks, not the months it used to. Speed matters. Opportunities do not wait for a six-month data-cleansing project.