The best source is rarely at the top of the search results

There's one sentence that makes me prick up my ears these days: "It should be somewhere in the project folder." What usually follows isn't knowledge work. It's archaeology.

In there are the old tender, three supplier quotes, the consultancy's presentation, a benchmark, two versions of the bid comparison, an Excel file called final_v7_new.xlsx and the award recommendation, which is missing the actual reasoning. Every one of these is a source. But they're not all worth the same.

This gets expensive at the latest when the same category comes back on the table two or three years later. The new category manager goes looking for the should-cost model from back then. The business unit remembers "roughly 12% savings potential." The supplier, remarkably, only remembers the arguments that worked in its favor. And the consultancy back then did good work, except the project team has long since moved on.

So the problem isn't that we have no sources. The problem is that we have no hierarchy of sources.

Fraunhofer IPK reaches a very similar conclusion on systematic problem-solving: a problem is only considered solved once the knowledge gained has been documented, indexed and made findable for others after implementation. (Source: Fraunhofer IPK, Systematic problem-solving in companies) But findable doesn't mean reliable. A PDF doesn't become more accurate because a semantic search finds it in 0.8 seconds.

I'd sort sources into four levels.

First, internal primary sources: framework agreement, price sheet, purchase order, invoice, quote, bid comparison, negotiation minutes, award recommendation. If I want to know what price was agreed, I'm not initially interested in a benchmark from some study, but in the signed contract, and after that in what was actually called off and paid.

Second, context sources: market conditions and price indices, demand forecast and volume plan, supplier risk, specifications and approvals, internal capacity at the time. The price from 2024 is a fact. Whether it's still a reliable baseline in 2026 is a different question. A source without a date and the context it was created in is only half a source to me.

Third, reliable external evidence: official price and commodity indices, standards, BME or consultancy benchmarks, suppliers' annual reports. Indispensable for checking the internal view against the market, but a benchmark answers "What's happening out there?", not automatically "What should we decide tomorrow?"

Fourth, interpretation: press articles, consultants' opinions, LinkedIn posts, AI answers. Extremely useful for finding hypotheses. Dangerous when hypothesis and evidence get mixed up. That's why I'd never ask an AI "What's the truth about this category?" but rather "Which sources support which hypothesis, and which primary source are we still missing?" A small difference in wording. A pretty big one in method.

I've essentially seen three responses to this challenge:

• "We'll give the AI access to SharePoint." I think it can then only search poor documentation faster.
• "We'll build a central knowledge base." I think it makes sense, but if nobody knows which version is authoritative, all you centralize at first is the ambiguity.
• "Our category manager knows that." I think that's probably true. Until they move to another category, or even another company.

What I haven't seen is a procurement function that consistently ranks its sources by evidence, context and reasoning rather than by findability. ISO 30401 doesn't describe knowledge management as a one-off filing project, but as a system that has to be run and continually improved. (Source: ISO 30401:2018, Knowledge management systems) To me, the perfect source isn't the perfect document but the combination of evidence, context and reasoning: What was decided, on what basis, under what conditions, and does it still hold today?

My test would be simple: If the responsible category manager were gone tomorrow, could their successor tell within ten minutes which three sources carried the last award decision? If the answer is no, we probably don't have a search problem. We have a knowledge problem. In my experience, that's the case in 9 out of 10 situations.

This week's question for my network of procurement experts: If you had to hand over one of your most important categories to a new colleague tomorrow, which three sources would they need to read first, and has that already been documented? Surprise me!

I'm less interested in the system behind it than in the honest answer. I welcome every comment! If you don't want to miss the Elevatry video and upcoming Friday notes, you can sign up for free here.

— Heiko, founder of Elevatry. 25 years of strategic procurement. My next note will be published on October 16: Perfect wording.