Purpose
This diagnostic exists to answer one question: which factor is currently limiting trust in this analytics organization?
It is deliberately not a scored assessment. A composite number would average away the very thing the trust model is designed to expose — that one weak factor sets the ceiling regardless of how strong the others are. What follows is a set of operating questions and a way to read the answers.
How to run it
- Answer for a specific client, account, or business partner. Answers averaged across a portfolio are unreadable.
- Answer from evidence, not intention. If the answer is 'we would, if asked', the answer is no.
- Run it twice — once with the delivery team, once with the person who consumes the work. The gap between the two readings is itself a finding.
- Stop at the first factor that reads clearly weak. There is little value in diagnosing the second constraint before the first is addressed.
Reliability — is the work dependable?
- If two people answered the same question independently, would they produce the same number?
- Is there a named owner for every metric definition in regular use?
- When a number is disputed, is there a documented reconciliation path, or does resolution depend on a particular person?
- Do commitments on timing hold without escalation, and is a missed date visible before the client notices it?
- Does the client re-check the work before using it?
Relevance — does the work serve a decision?
- For each recurring deliverable, which decision does it inform, and who makes that decision?
- Which decision changes if the analysis says the opposite of what everyone expects?
- Does the delivery cadence match the decision cadence, or the reporting calendar?
- When was a recurring deliverable last retired, and what triggered it?
- Can the team name the three most consequential decisions this client will make this quarter?
Judgment — does the work carry a view?
- Do findings arrive with an interpretation and a recommendation, or with options and no position?
- Is confidence stated explicitly, including when it is low?
- How many people on the team can make a defensible recommendation without escalating?
- When the evidence is ambiguous, does the team say so and still recommend, or does it defer?
- When a recommendation turns out to be wrong, what happens to the person who made it?
Accountability — does ownership survive delivery?
- Is the unit of ownership the deliverable, or the question?
- Who is accountable for what happens after delivery, by name?
- How would the team know if a recommendation was received and then ignored?
- When a problem originates outside the analytics function, who carries it?
- Is follow-through visible to the client without the client having to ask?
How to read the answers
- Weak Reliability
- Every other factor is discounted. Recommendations will be re-litigated as data disputes. Fix definitions, ownership, and the reconciliation path before investing anywhere else — work on relevance or judgment will not be believed.
- Weak Relevance
- The classic high-reliability trap. The function is accurate, busy, and peripheral. Adding accuracy or capacity will make this worse. Start by writing down the decisions and testing that list with the client, then retire what serves none of them.
- Weak Judgment
- The team is trusted to be correct but not consulted before decisions. Usually this is a permission problem before it is a skill problem. Require a stated view in writing, review the reasoning rather than only the outcome, and make the cost of a defensible wrong call survivable.
- Weak Accountability
- Trust erodes quietly, because nothing visibly fails. Work is delivered and nothing happens. Move the unit of ownership from deliverable to question, and make post-delivery follow-through part of the standard rather than a courtesy.
- Two or more read weak
- Treat this as an operating-model finding rather than four separate problems, and sequence from Reliability outward. The factors are ordered the way they are because each one depends on the ones before it being credible.
- All four read strong, and trust is still low
- Re-run the diagnostic with the person who consumes the work. A large internal-versus-external gap is the most common cause, and it is a translation finding, not a capability finding.
What this diagnostic deliberately omits
- A composite score. Averaging hides the constraint, which is the only thing worth finding.
- Benchmarks. There is no defensible external norm for these questions, and an invented one would manufacture certainty.
- A maturity grade per factor. Stage language belongs to the organization as a whole, not to individual factors.