Framing the problem and consulting contract
9 min read
Lesson objectives
- Frame the decision, scope the engagement, and separate evidence from hypotheses across channels.
Please sign in to use all the tools
Once you sign in, the reading tools are yours: larger or smaller text, typeface, line spacing, focus mode, highlighting and bold, and notes saved to your account and gathered in My study.
A consulting engagement often begins with a sentence that already sounds like a solution: ‘Reduce contact-center response time.’ Converting it immediately into a staffing or chatbot project is premature. A customer may have waited because a request moved between teams without a shared record, because policy prevented the agent from finishing it, or because a status message never arrived and prompted a call. Each explanation implies a different intervention, cost, and risk of harm. The consultant's first task is to turn the sponsor's wish into a bounded decision: which customer outcome should improve, which journey, segment and channel are in scope, what resources can change, and who owns the decision? Separate request, need, explanation and proposal. The request is the sponsor's statement; the need is a user outcome; the explanation is a causal hypothesis; the proposal is a testable intervention. The first explanation is not a fact.
A decision brief contains five connected elements. First, frame a question that allows comparison, such as ‘Which change reduces repeat contacts about returns without increasing incorrect returns over eight weeks?’ Second, mark the process boundary: from the customer's request to confirmation of resolution; the app, phone and warehouse are included, product pricing is excluded. Third, record a baseline with its source, date and denominator, rather than a bare number. Fourth, specify data rights and constraints: complaint records with personal details minimized, interviews with understandable consent, and no customer identities in deliverables. Fifth, agree on an acceptance criterion and decision date: what evidence would make the sponsor pilot or reject a change? Operations must know whether the consultant only advises or may change a procedure. If authority is undefined, defer operational changes until it is clarified; diagnosis can continue with authorized information.
Do not confuse the unit of analysis with a metric's denominator. If a manager reports a seven-minute average response, ask whether it covers calls during staffed hours or all service requests including email and visits. Were abandoned calls excluded? The average may improve when complex cases move to another channel even as the whole journey worsens. Record which cases enter and leave the measure, request a breakdown by channel and problem type, and look for an outcome such as a completed return or the need to try again. A metric is not appropriate merely because it is available on a management dashboard. Response time can be an early operational signal, but it does not prove that the customer obtained the desired outcome. Every figure needs its source, period, denominator, computation and exclusions. Without these, classify it as an initial signal, not a baseline suitable for a commitment.
To avoid letting one sponsor monopolize the definition of success, collect at least three perspectives: the customer who wants completion without repetition, the agent who needs information and authority, and the process owner who carries cost and error risk. Gathering perspectives does not give them identical decision rights; it exposes conflict before investment. Marketing may want a seamless experience, compliance may require an extra check, and the contact center may want shorter conversations. The consultant states non-negotiable constraints and then the choice space inside them. Record what the review does not own: no legal ruling on returns policy, no disclosure of individual records, and no guarantee of a satisfaction increase. Refer legal, security, or access-rights questions to the authorized specialist while keeping the operational question open. This separation preserves the recommendation's independence and limits, and prevents the consultant from claiming an approval role.
Engagement boundaries are not a ceremonial exclusion list. Suppose a sponsor asks to ‘raise customer satisfaction’ but controls only service-message wording. If dissatisfaction stems from delayed financial decisions owned by another division, promising an overall satisfaction rise exceeds that sponsor's authority. Reframe a nearer attainable aim: reduce uncertainty about request status along the refund journey, while naming the other process owner's decision on actual delays. Narrow scope must not hide the larger problem; record the dependency, its decision owner and how it may confound pilot evaluation. Define which groups are in the review: cases open before launch or only new requests? Is a ‘case’ one request or one conversation? Missing this decision lets teams shift denominators later to improve the reported result. Set an acceptance condition at customer level, such as accurate information at an agreed time plus an actionable next step, guarded by message-error limits and access for people unable to use the channel. Do not promise that wording alone will settle payments. The engagement remains honest and governable, and the sponsor knows when new authority is needed instead of demanding a result the consultant cannot deliver. Without agreed decision rights, consulting may deliver diagnosis, options and pilot conditions, but cannot alter the payment system or refund policy on its own.
Before signing the brief, test who can say no. The sponsor may approve a speed goal, but only the data owner can authorize linked records, operations can change shifts, and the policy owner can define exceptions. Give every critical assumption an owner, small check and date. If personal-record analysis lacks authorization, do not extract data first and seek permission later; use an approved aggregated design or an authorized route. If operations cannot pilot on sensitive cases, choose a lower-risk group for learning and document limits on generalization. Early dependency mapping lets the sponsor widen authority or narrow the question before receiving an undeliverable report. Ask the team: what decision follows if another cause emerges; what evidence warrants a small pilot; what blocks it regardless of apparent benefit; and who informs customers when a promise changes? Unanswered questions indicate a governance gap that belongs in the consulting diagnosis. The first recommendation may be to assign a missing decision right rather than buy a tool. Record stakeholder disagreement without assuming bad faith; each sees a different service slice. Once decision and authority are clear, evidence collection has a purpose: testing competing explanations that lead to different actions. Gathering everything available without a decision question consumes time, raises privacy risk and creates an illusion of knowledge through abundant charts. This is the practical reason to start with a bounded contract rather than another dashboard.
Stress-test the brief by reversing its assumptions. If the team discovers in two weeks that response time is not the problem, can the question support a different recommendation without changing the outcome? If response improves while repeat contact rises, will the acceptance rule detect failure? If the data team cannot release identifiable records, can a less revealing analysis answer the question? A sound brief has clear answers to all three. Ask the sponsor to confirm scope, definitions, date and pilot decision owner, and record objections separately; approval of the brief is not advance approval of a particular solution. When a material data gap appears, specify a small verification step before committing cost or benefit claims. The brief then acts as a reviewable decision control rather than a document forgotten after the meeting.
Lesson summary
Frame the decision, scope the engagement, and separate evidence from hypotheses across channels.