The consultant’s field guide

Write “By [date], [owner] must choose [options] to improve [metric], subject to [constraint].” If stakeholders disagree on the metric, resolve that before building a model.

Name a threshold before requesting data. For example, find the minimum annual volume that makes a finishing line cheaper than outsourcing, then verify whether committed orders clear it.

Split the problem into causes you can test. For falling profit, start with realised price, units, mix and costs. Check for overlap before assigning analyses.

Record what you think, the evidence you would expect to see, and the evidence that would make you abandon it. Revisit the hypothesis after each major finding.

Agree the market, period, customer segment, options and deliverable. Keep a record of out-of-scope questions so they can be addressed when their value is clear.

Specify fields, definitions, period, level of detail, source owner and deadline. Ask for a small sample first so you can catch join keys, missing fields and format problems.

For each material input, record the source, date, scope, units and whether it is observed, estimated or assumed. Mark the assumptions that could reverse the recommendation.

Ask about the last real decision, what the person considered, what they did and why. Ask for evidence of the sequence before asking how they might behave in future.

Cross-check demand with customer purchases, operating capacity and another independently collected estimate. Trace reports back to their original inputs before treating them as corroboration.

Record the strongest evidence against the current recommendation. Decide whether it changes the main answer, narrows its scope or deserves a separate test.

Write the unit beside every factor. Households × purchases per household per month × ₹ per purchase gives ₹ per month. Convert to annual figures only after checking seasonality.

If a model splits profit by segment, sum the segments and compare with the source P&L. Check duplicates, join losses and unallocated overhead when the totals differ.

Verify who could have taken the action and for how long. Compare conversion among equivalent eligible populations and retention at the same cohort age.

Calculate the threshold at which the two leading options tie. Test whether demand, capacity and costs put the threshold within a credible range.

Use low, base and high cases supported by assumptions. Rank the inputs by how much they move the answer, then test the most sensitive input first.

Separate inputs, calculations and outputs. Add units, dates, a reconciliation check and a short change log. Ask someone unfamiliar with the model to trace one output back to its source.

Give each slide a sentence that states what the audience should learn. Make the chart, table or quotation prove that sentence.

Choose a line for time, bars for comparable categories and a waterfall for a reconciled change. Label units, time, source and any material exclusions.

Include the decision requested, two or three reasons, the best rejected option, the main risk and the next action. Put supporting calculations in an appendix.

Discuss the draft recommendation with decision owners and implementation leads before the meeting. Ask what evidence they would dispute and what makes the action difficult.

In a verbal recommendation, lead with the call, give the most decision-relevant fact, state the main uncertainty and name the next step.

Set the review date and work backwards through the analysis, dependencies and data collection. Request the input with the longest lead time early.

Write who decided what, when and why. Record unresolved issues with a named owner and date so they cannot disappear between meetings.

Check incentives, decision rights, capacity and tools before recommending training. Ask the people doing the work to describe the exception they face most often.

Agree the pilot scope, target, guardrail and review date before launch. Specify what happens if the result is inconclusive or the guardrail is breached.

Use approved storage and tools. Minimise personal or commercially sensitive details in interview notes, remove identifiers from shared exhibits and check who can access the material.

Confirm the objective, scope, period and any ambiguous unit. Then outline a few decision-specific branches and say which you would test first.

State what is measured and in what units. Describe the strongest pattern with a number. Explain what it changes about the case.

Size demand from customers and usage, then cross-check against operating capacity or transactions. Explain the largest uncertainty and avoid false precision.

Name the wrong unit or assumption, correct the calculation and explain whether the recommendation changes. Keep the interviewer oriented as you revise.

Use Situation, Task, Action and Result. Spend most of the time on what you personally did, why you chose it and what the evidence says about the outcome.

Score structure, arithmetic, exhibit reading and synthesis separately. Repeat the weakest segment with a fresh prompt before repeating a full case.