It is Halloween, so let me tell you about something genuinely haunting your business: dark data. Dark data in business is all the information you collect, store, and never look at again. Years of sales transactions, customer chat logs, complaint records, delivery notes, cashier receipts. It sits in exports, old spreadsheets, and admin panels, quietly holding answers to questions you keep guessing at.
The reason this matters is not philosophical. Most business owners I meet believe they need to start collecting data before they can make data-driven decisions. Almost none of them do. If your business has operated for more than two years, you are already sitting on more usable evidence than you will ever finish reading. The gap is not collection, it is digging.
Here is what dark data typically looks like for an Indonesian SME, and three digging exercises that reliably surface money. None of them require new software.
What Counts as Dark Data
Take a quick inventory. You almost certainly have:
- Transaction history in your POS, accounting software, or marketplace seller dashboards, going back years.
- Chat history in WhatsApp Business, marketplace chats, and Instagram DMs. Thousands of conversations where customers told you, in their own words, what they wanted and why they hesitated.
- Complaint and return records, formal or informal, scattered across email and admin notes.
- Operational logs: delivery times, stock movements, staff shift records.
The defining feature of dark data in business is that it was recorded for one purpose, usually compliance or operations, and never reused for decisions. The receipt existed to complete a sale. Nobody ever asked what ten thousand receipts say when read together.
Exercise 1: Top-Customer Concentration
Export 12 months of sales, group by customer, sort descending. This takes one afternoon in a spreadsheet, and it changes how owners see their own business almost every time.
At a building-materials distributor I worked with, the owner was confident revenue was "spread across many contractors." The export said otherwise: 6 customers out of roughly 240 produced 58 percent of revenue. Two of those six had been drifting to a competitor for months, visible as shrinking order frequency in the data, invisible in daily operations because each individual order still looked normal.
Questions this exercise answers:
- What share of revenue comes from your top 5 and top 20 customers?
- Is any top customer's order frequency or basket size trending down over the last two quarters?
- Which top customers have never been contacted outside of transactions?
If your top 5 customers are more than 40 percent of revenue, you do not have a marketing problem, you have a retention priority. A single lunch with a drifting key account is worth more than a quarter of ad spend, and the data tells you exactly whose calendar to book.
Exercise 2: Complaint Theme Counting
This one is uncomfortable, which is why nobody does it. Gather every complaint, return, and negative review from the last 6 months. WhatsApp threads, marketplace reviews, refund notes. Then do the crudest possible analysis: tally them by theme. Late delivery, wrong item, damaged packaging, price dispute, unresponsive after-sales. A tally chart on paper is fine.
A snack producer selling through resellers did this and found that 71 of 104 logged complaints were variations of one theme: broken product on arrival. Not taste, not price, not service. Packaging. They had been debating a rebrand and new flavors for months while the actual problem was a Rp 1,800 difference per unit in bubble wrap and carton quality. The fix cost less than one month of the ad budget and cut complaints by more than half in the following quarter.
Complaints are the highest-signal dark data you own because customers wrote them for free, unprompted, and honestly. Counting themes converts a pile of unpleasant messages into a ranked to-do list. The same logic applies before you buy any customer-facing technology; if the top complaint theme is slow replies, that changes the math I laid out in Chatbots vs Live Agents: The Real Cost Comparison.
Exercise 3: Dead-Hours and Dead-Days Analysis
Your transaction timestamps know your rhythm better than you do. Export sales with date and time, then pivot by day of week and hour of day.
What surfaces, reliably:
- Dead hours you are fully staffed for. A cafe owner in Serpong found Tuesday and Wednesday 2 to 5 pm produced under 4 percent of weekly revenue while carrying full staffing. Shifting one staff member's hours and running a weekday afternoon promo turned the dead zone from a pure cost into a break-even experiment.
- Peaks you are understaffed for. The same data showed Saturday 10 am to noon queues, visible as a cluster of transactions followed by a suspicious gap, which was customers walking out.
- Seasonality you plan around by feel instead of fact. Payday spikes, Ramadan curves, school-holiday dips. You know these exist. The data tells you their exact size, which is what you need for stock and staffing decisions, especially heading into campaign season, a topic I cover in Harbolnas Season: Prepare Your Store for 11.11 and 12.12.
Why You Should Not Buy a Dashboard Yet
The instinct after reading this is to shop for a business-intelligence tool. Resist it for now. Dashboards automate questions you already know how to ask. If you have never manually done the three exercises above, you do not yet know which questions matter for your business, and you will end up with a beautiful dashboard displaying metrics nobody acts on. I have seen SMEs spend Rp 100 million on BI tooling that gets opened twice a month.
The right order is: dig by hand once, act on what you find, notice which questions you want answered every month, and only then automate those specific questions. Manual first, tooling second. The spreadsheet you already own is enough for round one.
The Practical Takeaway
Pick one exercise and put two hours on the calendar this week. My recommendation for most businesses is the top-customer concentration analysis, because it takes the least effort and most often triggers immediate, obvious action.
Dark data in business is not a technology problem to solve someday. It is money you already paid to collect, sitting unread. The scariest thing about it, fittingly for today, is not what is hiding in there. It is how long you have been making decisions without looking.