How to Turn a Business Problem Into a Data Question
|

How to Turn a Business Problem Into a Data Question

A business problem into a data question often starts with a simple statement: “sales are down,” “customers are leaving,” or “costs are increasing.” But before opening a spreadsheet or creating a dashboard, the real challenge is turning that business problem into a clear data question that can actually be answered. Mondays have a familiar script. The boss says “sales are down, can someone look at the data?”, a few people nod, and someone scribbles “look at data” in a notebook.

 Look at the data is a terrible brief

A worry has no edges. A useful question does. It needs something that can be counted, a group it covers, a stretch of time, and something to compare against. Try it with “sales are down.” Down against what — last month, last year, or a target someone set in April and then forgot? Rupees or number of orders? Everywhere, or just one city? Until those things are settled, nobody knows which numbers to pull, and nobody knows when to stop. Open-ended analysis will happily expand to fill whatever deadline you give it.

Vague briefs create another problem: almost every finding feels a bit relevant. The report turns into a tour of whatever the analyst noticed, and people skim tours without acting on them.

Spend the first hour getting the question right. The analysis that follows usually takes far less time. That still surprises people every time it happens.

Kavya, her chai, and one sentence

Kavya isn’t real. She’s stitched together from a few people who are. She sells cotton kurtas online with a small team, a marketplace storefront, and a courier she’s a little wary of. One afternoon, over chai, she tells Rohit, who looks after operations: “Returns are killing us. Look at the data and tell me what’s going on.”

It sounds reasonable. There’s almost nothing in that sentence a spreadsheet can actually use. The rest of this piece follows that sentence until it becomes something a pivot table can answer. A couple of wrong turns stay in on purpose — the wrong turns are often the useful part.

Why is it hurting now?

Returns have been around since the first parcel went out. So why is Kavya upset this week and not last month? Rohit asks, and five minutes later they have it. The courier raised its reverse-pickup fee two months ago. On some orders, collecting a returned kurta now eats the entire margin. Cash is tight. The real pain isn’t the return rate itself — it’s the rupees leaking out every time a parcel comes back. If she’d been losing sleep over bad reviews or stock stuck in transit, the question would have looked completely different. Same three words, different problem underneath. People who skip this conversation often spend a week analysing the wrong thing very thoroughly.

What would you do with the answer?

If one question deserves to be written on the wall, it’s this one: what will you do differently depending on what the data says? If the honest answer is “nothing, we’ll carry on as before,” drop the problem and pick another. Analysis that can’t change an action is just a pastime, and an expensive one. Rohit had to push a bit, but Kavya’s list eventually arrived: rewrite the size guide, drop a few designs, tighten the return policy, or haggle with the courier. Four possible moves, each hanging on a different fact. A new size guide only helps if fit is the issue. Dropping designs only works if a handful of them are doing most of the damage. A stricter policy only matters if customers are gaming the current one.

Look at what happened. One shapeless worry turned into four rival explanations, each linked to something she could actually do on Monday. The data question gets built from those.

A useful phrase to keep: “we’ll pick between A, B and C, and we need to know X to decide.” Write it down. Everything else is just about sharpening X.

Chop it up

Big problems are usually bundles, and bundles can be taken apart. The old journalist questions work well here because each one points at a place the trouble might be hiding. How much comes back, and what does one return really cost once you add shipping, repacking and the lost sale? Where do the returns come from? Who sends things back — new buyers or regulars? When do they spike? Which products? Kavya covered the back of an electricity bill with these. Ten minutes in she had five smaller questions instead of one large groan, and she noticed (a little sheepishly) that she didn’t actually know the answer to any of them. She had feelings, plenty of them. That itself is useful — it shows where numbers could genuinely help.

It also keeps priorities honest. A return that costs ₹40 to handle, with one design behind most of them, tells you where to look first.

Read More: How to Improve Logical Thinking: A Practical Guide to Thinking Critically

Write your guesses down first

This sounds odd, but it works. Before opening a single file, write down what you think is going on. The formal word is hypothesis. In plain language it’s just a guess you can check. Kavya’s four were: sizes run small; two designs have a fabric or colour problem; some customers order two sizes and send one back, treating the site like a changing room; cash-on-delivery orders get refused more often. Each of those can be wrong, and that’s what makes them useful. “Returns have many causes” can’t be wrong, so it can’t be tested, so it doesn’t help anyone. A guess that might fail tells you which numbers to pull and which result would sink it.

Guessing first also stops the fishing trip. Open a dataset with no theories and every little wobble starts to look meaningful, even when some of them are pure chance. Go in with four named suspects and you test four specific things.

Ask the team as well. Rohit mentioned that one fabric bleeds in the first wash. Meenal on the support desk had heard “too tight” so often she’d stopped writing it down. Free hypotheses, both of them, and better than most dashboards manage.

Read More: How to Improve Logical Thinking: A Practical Guide to Thinking Critically

Every word needs a number attached

This is the real conversion — where a business problem becomes a data question. Each loose word gets tied to something you can actually count.

Start with “returns.” Do exchanges count? Parcels refused at the door? Things sent back after thirty days? Kavya’s team discovered they’d each been using the word differently, so any figure quoted could be argued over. They settled on one clear meaning: any item the customer sends back within fifteen days of delivery, for a refund or an exchange.

“Too many” needs a yardstick. She chose her own previous six months — the data she trusted most. “Customers”: everyone, or only repeat buyers? “Kurtas”: all designs, or just the top twenty?

Nearly every good question needs four things pinned down: what’s being counted, who or what it covers, the time window, and the comparison. Miss one and you’ve still got a worry, just wearing a nicer shirt.

Kavya’s version after all that: over the last six months, what share of delivered items came back within fifteen days, split by design, size and payment method? Longer than the original sentence, and a hundred times more usable, because someone can now go and fetch exactly those numbers.

Read More: How to Think Like a Data Analyst (Even If You’ve Never Touched a Spreadsheet

Is the data even there?

A lovely question is worthless if nobody wrote the answer down. So go looking. Orders are probably in the store platform. Return reasons might be in the courier portal, a support inbox, or nowhere at all. Payment method is usually easy. Size ordered, less so. Who owns each system, and can it be exported without a three-week email chain? Kavya hit a snag. Return reasons were free text, typed by customers in a mix of English and Hindi, and about half were blank. So “sizes run small” couldn’t be tested by reading the reasons. It could still be tested sideways, by comparing return rates across sizes.

When something is missing you have three options. Bend the question to fit the data that exists. Borrow a stand-in measure that moves with the thing you care about. Or start recording properly now and come back to the question in a few months. Kavya used the second and third. Return rate by size became her stand-in for fit problems, and a compulsory dropdown for return reasons went live that same week. Finding the gap early is better than discovering it after a fortnight of analysis. It can also be revealing on its own. A business that has no record of why things come back has been making decisions half-blind.

Read More : Online vs Offline Data Analytics Course Which Is Actually Better?

Squeeze it into one sentence

By now the table is crowded: a pain point, some decisions, a few sub-problems, guesses, and firm definitions. Boil it down to one sentence. Then hand it to someone who wasn’t in the room and watch what they do. If they know straight away which numbers to fetch, and you can say what you’d change depending on the result, it’s ready. If they frown, the frown usually tells you what’s still missing — often a time window or a comparison. Kavya’s final version: which designs and sizes account for most returns in the past six months, and are returns higher on cash-on-delivery than on prepaid orders? It names the thing being counted, the groups, the window and the comparison. It maps to three decisions she can actually make. An afternoon with a pivot table covers it.

One more useful habit: decide beforehand what each possible answer would mean. If three designs cause most of the returns, she reviews fabric and photos for those. If returns are even across sizes, the size guide stays. If cash-on-delivery returns are double, she nudges people toward prepaying with a small discount. Settling that early protects against a common trap — finding something interesting and then having no idea what to do about it.

Read More : Building Your First Power BI Dashboard (The Order Matters)

Same trick, different shops

Nothing here is special to kurtas. A coaching institute owner says “students keep leaving.” Put through the same steps, that becomes: of the students who joined in the last two batches, how many left before month three, and what did the leavers have in common — batch timing, faculty, or fee plan?

A kirana owner says “margins are thin.” That turns into: which ten products take the most shelf space but earn the least profit this quarter?

A small clinic hears “patients wait too long.” Sharpened, it’s: on which weekdays and at what hours does the median waiting time cross thirty minutes?

Each one started as a grumble and ended as something a person could go and count.

Where people trip

Most of the wrong turns are familiar. Someone forwards a shiny data export and asks what’s in there. Sometimes that pays off. More often it just gives you charts and shrugs. Someone asks “how do we grow?” That one can’t be answered from a spreadsheet, so shrink it until the answer fits in a sentence and the work fits in a week. Someone asks why a rival’s prices are unfair. The answer is already baked into the question, so there’s nothing left to discover. A good data question leaves room to be surprised.

Then there’s the classic: two things move together and someone declares that one caused the other. Cash-on-delivery returns may well be higher, but those customers might live in towns with slow couriers, where people change their minds while waiting. Write down the other possible explanations and cross them off one at a time. And don’t skip the packer or the support agent. They often already hold half the answer. Last, don’t wait for a perfect question. Run a rough one, see what comes back, then fix it. The second version is nearly always better.

What Kavya did next

By the end of that afternoon Rohit had the first answers. Three designs, all pale colours, caused most of the returns. Sizes L and XL came back more often than S and M, which pointed at fit rather than a general quality problem. Cash-on-delivery orders were returned or refused noticeably more than prepaid ones. Nothing earth-shattering. All of it is usable. She reworked the size chart using real garment measurements, reshot the three designs in daylight with the model’s height printed next to the photo, and added a small prepaid discount at checkout. Then she wrote the date in her notebook and checked the return numbers again six weeks later. Problem, question, data, answer, action, result. Each loop makes the next question a little sharper.

When another pair of eyes helps

Plenty of people get through all this on their own, especially on contained problems. A few situations are worth a second view: the problem is big and tangled, with several departments involved; the data sits in four systems that don’t agree with each other; a large decision rides on the answer; or everyone has stared at the same numbers for so long that nobody sees them clearly any more. That’s the work Ankashram does. We’re a data analysis firm, and a surprising amount of what we do happens before any analysis starts: sitting with a team, untangling a vague worry, and shaping it into a question the data can actually answer. Often the most useful thing we hand back is a clearly worded question, because everything after that gets faster and cheaper.

To close

Every data project is only as good as its question. The tools can be a plain spreadsheet and the analysis can be modest, as long as the question is sharp and a decision is waiting on the answer. So the next time somebody says sales are down, customers are leaving, or costs keep creeping up, resist the urge to open the file straight away. Ask why it’s hurting now. Ask what decision is waiting. Break the problem into pieces, guess before you look, pin the words down, and write it all as one clear sentence. And if you’d like a hand turning a messy business problem into a question worth answering, Ankashram is here for that.

Frequently Asked Questions

Q1.What counts as a data question?  

Anything you could answer by counting or comparing records that somebody kept. “Which designs came back most last quarter?” counts. “Why are we struggling?” doesn’t — not yet.

Q2.Can’t I just open the data and look around?  

You can. Occasionally something useful jumps out. But without a clear question you won’t know a real pattern from a fluke, and you won’t know when to stop, so the work drags on.

Q3.How will I know my question is ready?  

Give it to a colleague. If they can go and pull the numbers without coming back to ask you clarifying questions, it’s ready. If they return with three questions of their own, those three are the gaps.

Q4.The data I need isn’t recorded anywhere. Now what?  

Change the question to fit what you have, find a stand-in measure, or start recording today and answer it in a couple of months. Kavya did a bit of each.

Q5.Do I need to code?  

No. Deciding what to ask is the hard part, and that needs curiosity more than code. A spreadsheet handles most of what comes after.

Q6.Who should help shape the question?  

The person who raised the problem, the person who will act on the answer, and someone who knows how the data actually gets recorded. If a packer or a support agent can join for twenty minutes, even better.

Q7.How long should it take?  

An hour or two, usually. If it’s still fuzzy after that, the problem is probably too big — split it.

Q8.When does it make sense to call in a firm?  

When the data lives in several places that disagree, a big decision hangs on the answer, or everyone has stared at the same numbers for too long. Ankashram helps with exactly that — starting with the question, then the analysis.

Similar Posts

Leave a Reply

Your email address will not be published. Required fields are marked *