Making Decisions on Data You Can Actually Check

Almost any business a few years old already holds enough information to decide well. The information is just scattered across a spreadsheet, a point-of-sale app and a group chat, then summarised by hand by one person the morning of the meeting.

Start by naming which decisions you have been making on guesswork, then record only the numbers those decisions need, in the place where the work actually happens.

What "data-driven" usually means in a growing business

Monday morning, an admin pulls figures out of the app, pastes them into the summary sheet and sends the file to the group. The ten o'clock meeting opens with someone asking whether Saturday's sales are in there yet.

Half the meeting goes on whether the numbers still hold. The decision gets made anyway, on a summary that is a few days old, and nobody in the room can check a line of it without opening the source file.

Businesses arrive at this point when the spreadsheets and generic tools that used to be adequate stop fitting how they work. From then on, decisions start getting made on stale data.

Start from the decision, not the data

A useful question: this month, which decisions did you make without a clear basis? The list is usually short and looks similar from one business to the next.

Often it is which items to reorder this week and in what quantity. Then which customers to chase for payment first, and whether it is time to add another person in the warehouse.

Take one item off that list and ask what number would settle it. For reordering, two are usually enough: sales per item over the last four weeks, and stock on hand today. For collections, a list of overdue invoices with their ages.

The rest is reporting out of habit. An hourly chart of website visitors is pleasant to look at, but no decision changes because of it. Once you have picked the numbers that matter, our guide to measuring digital ROI covers how to weigh what they tell you.

Why the numbers are stale

The cause sits in the process: data retyped after the work is finished will always run behind the work.

The sequence usually goes like this. An order arrives by message in the morning and gets written in a notebook or passed straight to the warehouse. The entry reaches the spreadsheet when the admin has a moment, sometimes that afternoon, sometimes tomorrow. Goods leave today and the stock figure drops in the file two days later. The gap becomes obvious the first time a large order lands.

Buying software to draw charts from that spreadsheet leaves you paying for a tidier version of the same problem. The figure on the chart is still two days old, and the tidy presentation makes it feel more certain than it has any right to.

A second risk is harder to see. When the process lives in one person's head, nobody else can verify the summary. You cannot tell whether this week's figure fell because sales fell or because a column went unfilled while they were on leave.

Three places the data usually hides

Before deciding to build anything, trace where the data sits now. There are rarely many hiding places.

• The spreadsheet only one person understands, with formulas built up over years, a couple of hidden sheets, and filling rules that are written down nowhere.
• The off-the-shelf tool that holds the data and will not report it the way you need. The reports cannot be adjusted, and every export has to be reshaped by hand, so nothing else can be built on top of it.
• The group chat, where the decision actually got made. A special price approved in a message, an order cancelled in one line, a complaint resolved without ever touching a system.

The third gets missed most often during planning. The decision happened there, and the only way to find it again is to scroll up.

If one of the tools you already pay for can report the numbers you need, using it is the right call. We do send prospects back to an off-the-shelf tool when it solves the problem. Check the reports menu before you assume something has to be built.

What a system of record changes

The delay goes when the recording happens at the same time as the work. The person taking the order enters it while the customer is still on the line, and the stock figure moves within the minute.

There is still a cost to be honest about. Orders arrive by message before anyone is at a screen, so someone has to enter them once they get there, and until they do, those orders are missing from the figure. The gap is an hour, and it closes one order at a time instead of everything waiting for Monday's summary.

The group chat does not go away either. People will keep approving a special price in a message and cancelling an order in one line. What changes is that the price and the cancellation get entered against the order, so the record of the decision no longer lives only in the scroll.

Once that runs, the numbers become a by-product of the day's work. Nobody pulls and pastes on Monday morning, because four weeks of sales and today's stock are already there. The meeting stops arguing about whether the data still holds.

The figures can also be traced, which matters more than the speed. If sales for one item look strange, you open the transactions one by one until you find the cause. On a manual copy, the only way to check is to ask the person who made it.

None of this has to be done at once. Take the process that goes wrong most often, and if two are equally error-prone, take the one that eats the most hours. The next process follows if it earns its place. Our digital transformation roadmap for SMEs sets out the order of stages.

Whose data is it

Decisions based on data only hold up if you can get the data out whenever you want. Years of complete records are worth little if opening them requires someone else's permission.

One question tells you where you stand: if payment stops next month, can I still retrieve the last three years of records? Put it to whoever looks after your system now, and put it to us.

Our answer is that the code, servers, domain and every account are in your name from day one. Handover at the end confirms ownership you already had. The next developer can open the database, export what is in it and carry on without asking anyone for approval.

If the answer you get about your current system is no, our post on the hidden cost of poor software choices sets out what that position costs.

A first step that costs one stage

Begin without buying anything. Write down the three decisions you most often make on guesswork, and beside each one write the number that would settle it and where that number ought to be recorded, which a single page covers.

That list defines the scope of the first piece of work. Without it, a conversation with any vendor widens into features that change no decision at all, and the price widens with it.

With the list in hand, the work can be split by stage. The first stage is discovery: one to two weeks, producing a written plan and a fixed price. You can stop there, because the plan and the price are yours either way, and the second stage is only agreed after you have read them.

Bring the two or three decisions you still guess at

You do not need a requirements list or a view on which system fits. Just describe two or three decisions you are still making on guesswork, and where the numbers usually come from. The first call is free and there is no deck.

If the report you need already exists in a tool you are paying for, we will tell you. If something does need building, discovery gives you a written plan and a fixed price within one to two weeks, and that document is yours whatever you decide afterwards. Use the contact form, which opens WhatsApp, or email hello@arktik.id.