Buying Judgment by the Hour: A Due Diligence Scorecard for Intelligence Vendors
Photo: Ash Carter, Public domain, via Wikimedia Commons
A regional industrial distributor signs a six-figure annual contract with a competitive intelligence provider. Weekly briefings arrive on time, formatted impeccably, dense with share estimates and competitor moves. Eighteen months later a rival opens a facility in a market the briefings had characterized as stable, and the executive team learns that the primary research behind that assessment amounted to eleven phone calls, a stack of trade press clippings, and a syndicated dataset the vendor had relabeled as proprietary.
The instructive detail is not that the vendor was weak. It is that the buyer never asked a question that would have surfaced the weakness before the first invoice cleared.
Outsourcing collection is a defensible management decision. No mid-market firm can staff researchers across a dozen product categories and forty states. Outsourcing judgment is a different transaction entirely, and it is the one most procurement processes fail to examine. A vendor that gathers facts and hands them over with visible seams is an asset. A vendor that gathers fragments, smooths them into narrative, and attaches a confidence score is selling something closer to theater.
The seven questions below are the ones that separate the two.
The Line Between Collection and Judgment
Before the questions themselves, a framing point. Intelligence work has four stages: collection, synthesis, interpretation, and recommendation. Buyers routinely scrutinize the first stage and accept the last three on faith. That is backwards. Collection errors are usually visible in the output. Interpretation errors are not, because they arrive dressed as expertise.
Question One: Where Does the Data Originate, and Can You Trace It Back?
Ask for a source inventory, not a sample report. Insist on a breakdown of primary research, licensed third-party data, public filings, and modeled inference. Then ask for source lineage on a single claim of your choosing, chosen at random, and watch how long it takes to produce. A vendor with genuine provenance can answer in an hour. A vendor without it will offer a methodology appendix written in the passive voice.
Question Two: What Is the Methodology, Stated in Verbs?
Most methodology statements are pages of nouns. Demand verbs. How many interviews were conducted, with whom, recruited how, at what response rate? What was the sample frame and how was it weighted? Which forecasts are extrapolations and which are judgments? What is the stated error band? A provider that cannot explain its method in plain English to a skeptical chief financial officer does not have a method worth paying for.
Question Three: Where Exactly Does AI Sit in the Workflow?
The question is not whether artificial intelligence is used. It is where. AI applied to transcript sorting and translation is a cost efficiency. AI applied to drafting conclusions is a different risk profile. AI generating figures is disqualifying unless clearly labeled. Ask for a written disclosure clause that names each stage of the workflow and the human accountable for it. In 2026 this is a reasonable request; a provider that resists it has told you something useful.
Question Four: Who Signs the Analysis, and What Have They Covered?
Contracts are often sold on the strength of a platform and staffed by whoever is available. Ask for the named analyst, their sector tenure, their coverage history, and how many accounts they hold. Ask whether that person will be on your engagement in month nine. Rotating analysts through accounts produces reports that read well and know nothing.
Question Five: What Conflicts Sit Behind the Recommendation?
Intelligence providers increasingly consult, sell data to the companies they cover, and publish sponsor-funded research. None of that is disqualifying, but all of it should be disclosed. Ask which of your competitors are clients, whether any subject of a report has paid for its production, and how the vendor walls off its consulting arm. An unexamined conflict is not a scandal; it is a slow drift in what gets emphasized.
Question Six: How Is Your Data Protected and Segregated?
What you disclose to a vendor becomes an input. Ask how your information is segregated from other clients, whether anonymized findings enter syndicated products, how long records are retained, who can access them, and which subcontractors touch them. Confirm security attestations and breach notification terms in writing. Intelligence relationships fail quietly in this dimension until they fail loudly.
Question Seven: What Are You Actually Paying For, and What Incentivizes the Vendor?
Per-report pricing rewards volume. Retainer pricing rewards presence. Neither rewards accuracy. Ask directly: what does the vendor lose when the intelligence is wrong? If the answer is nothing, propose back-testing against outcomes at six and twelve months, with a contractual review trigger. A provider confident in its method will accept that conversation. A provider selling narrative will negotiate around it.
Putting the Scorecard to Work
Score each question from one to five, weight them by your exposure, and set a minimum threshold below which no contract is signed regardless of price. Re-run the score annually. Keep a small internal capability, even one experienced analyst, whose job is to interrogate vendor output rather than duplicate it.
None of this argues against outsourcing. It argues for buying with the same discipline you apply to an acquisition. The firms that get genuine advantage from third-party intelligence are not the ones paying the most. They are the ones who know precisely which part of the work they have bought, and which part remains theirs.