Ask business data questions in plain English: a guide

Questions instead of reports

For years, getting a number out of your business meant opening a report, picking filters and reading a table. Now you can ask business data questions in plain English, the way you would ask a manager standing next to you, and get the answer back in a sentence.

That sounds simple, and it mostly is. But the quality of the answer still depends on the quality of the question. A clear question gets a clear answer. A vague one gets an answer to something you did not quite mean.

This guide covers what happens to your question, the four parts of a good one, and side-by-side examples of vague and clear questions for a shop or restaurant.

What happens to your question

When you type "How were sales at the mall outlet last month?", the system has to turn your words into something exact. It works out which measure you mean (sales), which period (last month), and which slice of the business (the mall outlet).

Once those three are clear, ordinary code fetches the records and does the sums with a fixed formula. The AI's job is to understand your words at the start and explain the result in plain words at the end. It should not do the maths itself.

So most of the work in a good answer is in that first step: understanding exactly what you asked. Anything you leave out, the system has to assume. The fewer assumptions it makes, the closer the answer is to what you wanted.

The four parts of a good question

The measure. Say which number you want. Sales, number of bills, average bill, gross profit, discounts given, stock on hand. Gross profit, for example, is sales minus what the goods you sold cost you. "How are we doing?" names no measure at all.

The period. Say which days. Yesterday, last week, last month, the year so far. "Recently" leaves the system to guess whether you mean three days or three months.

The comparison. A number alone rarely tells you much. RM 50,000 is good or bad only against something: the month before, the same month last year, or another outlet. Say what you want it compared with.

The slice. Say which part of the business. One outlet or all of them, one product category or everything, one company or the whole group. If you run several outlets, this is the part people most often forget.

You do not need all four every time. But when an answer surprises you, check which of the four you left out.

Vague questions and clear ones

Vague: "How are sales?" Clear: "What were total sales last week, compared with the week before?" The clear version names the measure, the period and the comparison.

Vague: "Is the new outlet doing well?" Clear: "What were sales and the number of bills at the new outlet last month, compared with the average of our other outlets?" "Doing well" can mean many things. Sales and bills are things code can count.

Vague: "Are we making money?" Clear: "What was gross profit and net profit last month, and how did they change from the month before?" Net profit is what is left after every expense, including rent and staff, not just the cost of the goods.

Vague: "Which products are bad?" Clear: "Which products had the lowest gross profit last month?" or "Which products have not sold in 60 days?" "Bad" could mean low sales, low margin or stuck on the shelf. Each is a different question with a different answer.

Here is a made-up example of a clear question and its answer. You ask: "What were sales at Outlet B last month, compared with the same month last year?" The answer: "RM 50,000, against RM 40,000 in the same month last year, up RM 10,000 or 25%." The rise is RM 50,000 minus RM 40,000, and RM 10,000 divided by RM 40,000 is 25%.

Ask one thing, then follow up

Long questions with three parts joined by "and" often get a muddled answer. It is better to ask one thing, read the answer, then ask the next.

A good order is "what" first, then "why". Start with "What were sales last month, compared with the month before?" If they fell, ask "Why did sales change last month?" Then narrow it down: "Which outlet changed the most?" and "Was it fewer bills or smaller bills?"

Each answer tells you what to ask next, and a few questions take you from a single total to the actual reason.

Two more habits help. Use the same words for the same thing, so "sales" always means sales. And check the date on the answer. A good system says which day the figures are "as of", meaning the last day they cover, so you know whether last night's trading is in it.

When the answer is "I don't know"

Some questions cannot be answered from your data, however clearly you phrase them. "What does the shop across the road charge?" is one. "Will it rain on Saturday?" is another. Your sales and accounts simply do not hold that information.

A good system says so plainly instead of making up a number. That is not a failure. It tells you the answer you do get, when you get one, came from your own records.

If a month's books are not closed yet, a good system should also say that, rather than show a loss that is not real.

Ask it in plain words

Every clear question in this guide is one you can ask iSenang, your AI data analyst. Type it in your own words, tap a ready question if you are not sure where to start, or open a ready report.

The figures come from code with one fixed formula for each measure, every answer says which day it is "as of", and when a question is not covered by your data, it says so instead of guessing.