How to Turn Business Data Into Recommendations
Move from observation to options, a bounded proposal and a measurement plan.
Part of the Business Analytics track: modules, projects and first roles →
By the GullyMentor team · 5 min read · Published 2026-10-18 · Updated 2026-11-06 · Sources listed at the end. Not a promise about mentorship outcomes.
A business recommendation connects evidence to an action under stated assumptions. It explains why the action is worth considering, what it may cost, how you will measure it and what would change your mind. A finding alone is not a recommendation, and neither is a vague instruction to "improve marketing".
How is a finding different from an explanation?
In a fictional shop, revenue rises 12% while gross margin falls from 40% to 32%. That is an observation if the calculations are sound. "Discounts caused the decline" is an explanation requiring additional investigation.
Check product mix, returns, purchase costs and discount changes. Reconcile the measures before proposing a fix. Write the alternative explanations next to the one you favour, so the reader can see you considered them.
What decision and constraints are you working with?
Ask what the manager can actually change: price, promotion, stock mix or supplier terms. Consider budget, time and the risk of losing customers. An action outside the manager's control is not a useful recommendation.
Write the objective clearly. Maximising sales volume and improving contribution margin may suggest different choices. If the objective is unclear, ask before you analyse.
How do you compare options?
Offer a small number of realistic alternatives, usually two or three. For each, explain expected benefit, cost, uncertainty and reversibility. A blanket price increase may be less informative than a limited test on selected products.
| Option | Benefit | Cost or risk | Reversible? |
|---|---|---|---|
| Reduce discount depth on three products | Margin may recover | Orders for those items may fall | Yes |
| Renegotiate supplier terms | Lower unit cost | Takes time, uncertain | Partly |
| Do nothing and monitor | No risk | Margin stays low | Yes |
Use estimates honestly. Label assumed values and show how the result changes if those assumptions are wrong.
What is a bounded next step?
Recommend a limited experiment with a defined owner and observation period. State the success measure and a guardrail, such as margin improving without unacceptable loss of order volume.
Do not invent an outcome. The recommendation is a proposal to test, not evidence that the improvement has already happened.
How do you write the memo?
Use six short parts: decision, finding, supporting evidence, options, recommendation and measurement plan. Keep it to one page. Attach the calculations and definitions so the reasoning is inspectable.
Example recommendation line: "Cut the festive discount from 20% to 10% on three low-margin products for two weeks, owned by the category manager. Success is gross margin on those items rising while order count falls by no more than the agreed guardrail."
What are the common mistakes?
- Presenting a correlation as a cause.
- Ending with a general theme instead of a specific action.
- Hiding missing data instead of naming it.
- Recommending something the reader cannot approve or change.
- Giving one option, which looks like a conclusion reached before the analysis.
How do you check you got it?
Give your memo to someone and ask: what am I being asked to decide, what happens if I say yes, and how will we know it worked? If they cannot answer all three in a minute, rewrite. Also confirm you have named one condition that would reverse your recommendation.
Worked scenario: from finding to recommendation
Suppose a fictional online bookshop sees returns rise in one category. You check the data and find that most returns come from one supplier's titles, with the reason code "damaged in transit". The finding is the concentration of returns. The explanation, packaging quality, is still a hypothesis.
A sound recommendation would be to ask the supplier for a packaging review, move a small batch to a different courier for a fixed period and compare the return rate for both groups. State the owner, the period and the guardrail, such as delivery time not worsening. If the return rate does not differ, say what you would investigate next.
What should you build next?
Create a one-page memo for the fictional shop. Include one alternative explanation, one missing field and one reversal condition. Define the numbers first using Business KPIs Explained. If you want feedback on a memo, the audit page explains how to request a review.
Keep confidence proportional to the evidence. A small, clearly bounded test is often a better recommendation than a confident claim.
Sources & further reading
Examples and practice recommendations are educational illustrations. Tool documentation and programme inclusions should be checked for your intended use.