The report’s here. What do you do with it?

What do you need to know?

A report should help you understand what happened in your marketing and what is worth doing next. That sounds obvious. Yet a list of accurate results can still leave your most useful question unanswered.

Suppose sales have fallen. This is an example, not a client report. You can see the fall in the table. To make a decision, you need to know where to investigate. Did fewer people reach the site? Did fewer visitors buy? Did the products they chose change?

Each answer leads to a different conversation. Traffic falling after a budget reduction is a different situation from spending the same amount while people stop short of ordering.

Put the figures beside what changed.

Perhaps your most popular product went out of stock during the same period. Perhaps a promotion ended or prices changed. The people looking at the advertising need to know. Otherwise, they may search for an explanation where nothing has changed.

Before discussing the report, note what happened in the business during that period. It doesn’t need to be a document. A few specific observations about stock, offers, delivery or customer questions can change how you interpret the figures.

Ask the team for the same clarity: what they changed, when and why. If several things changed at once, you may not immediately know which made the difference. Knowing that is useful too.

Follow the example through to a decision.

Take an imaginary store so we can follow the reasoning without presenting a client result. Traffic is similar to the previous month, but orders have fallen. The owner asks whether the ads should change. It is a reasonable question. We do not yet have the answer.

First we check that the data is comparable: the same period, definition of an order and measurement method. Then we use the store like a shopper. Does the advertised product exist? Is it available? Does the page show the promised price? Can we reach payment?

In this example, delivery cost appears at the final step without being explained earlier. The customer team says people have been asking about shipping. Those are two observations worth investigating together. They are not proof that delivery explains the entire decline.

We propose showing delivery information earlier on the product page and checking that the ad still describes the offer accurately. We agree who will implement it. We record when it changes and what else happens in the store so we know what we are comparing later.

Where does AI help?

We use AI to help gather and read data from different places. It helps us bring together information from advertising, the website and email, and investigate the questions it raises.

A convincing explanation still needs checking, especially when the data leaves out something you know about the business. That’s why talking to you remains part of the work. A report won’t know about a stock problem if that information never reached it.

We take responsibility for the decision after reading and checking the findings. Technology helps us see more and make time for the questions that matter.

What you ask AI changes what comes back.

Ask it why sales fell and it may produce a coherent explanation from insufficient information. I would first ask what changed in the data, what information is missing and which hypotheses conflict. I want to inspect the basis of the conclusion before being persuaded by the writing.

For example: separate changes in order count from changes in average order value, identify the categories contributing to the difference and flag unavailable products. Then verify those observations against the original sources. Column definitions, dates, currencies and order statuses must be consistent.

AI was not on the supplier call unless someone provided the information. It cannot know that deliveries were delayed or a product was withdrawn after complaints. Bringing those details into the conversation can change the explanation. That is analytical work, not something a better-sounding prompt removes.

Sometimes the right decision is to wait.

If the price changed yesterday and people take longer to decide, the outcome has not matured. If tracking changed, the comparison may be broken. Agree a review date and the information needed for the next decision. Waiting is useful when you know what you are waiting to learn.

An unavailable product, a failed form or contradictory pricing is different. You do not need weeks of observation before repairing something demonstrably broken. Separate operational faults from hypotheses about customer behaviour because they call for different responses.

What do you leave the call with?

A proposed change, its reason, an owner and what you will review next. If suitable enquiries stall before the proposal, for example, the team can follow the response and terms discussed. The report should distinguish what is known from the information still missing.

If reading the report leaves you unable to say what happens next, ask for that connection. Numbers show where to look. The people who know the business help decide what to do.

You can send this edition to the person who reviews the report with you. Pick one result and try to get from that number to a decision. If you get stuck, you have found a useful question for your next call.