Data-driven work for small businesses: start with what you have

What data-driven work is, and what it is not

Data-driven work sounds like something for large companies with a data team. In practice it means something simpler: basing recurring decisions on current figures. Do you hire because it feels busy, or because utilization has been above 85 percent for three months? Do you call a customer because you happen to remember, or because the invoice is twenty days overdue and the customer is among your top five debtors?

So it is not a project with an end date, but a habit. Microsoft describes it in its adoption roadmap as a data culture: an organization where people naturally reach for data to support a decision.

What it is not: building as many dashboards as possible, commissioning a data warehouse before you know your questions, or postponing every decision until the data is perfect.

Why it often stalls in small businesses

The three most common reasons:

1. Data is scattered. Invoices in accounting, hours in time tracking, projects in project software, customers in the CRM. Anyone with a question that spans two systems has to export and combine in Excel.
2. Figures differ per person. Sales reports revenue by order date, finance by invoice date. The meeting then debates who is right instead of what to do.
3. Reports arrive too late. A monthly report that is ready halfway through the next month describes the past. You can no longer steer.

None of these problems is solved by collecting more data. They are solved by making existing data easier to use.

Step 1: start with decisions, not data

List five decisions you make every week or month. For example: which projects get extra capacity, which customers we call, whether we accept this job at this price, whether we hire someone. For each decision, write down what information you are missing now or get too late.

This is the most important step. Skip it and you build dashboards nobody uses.

Step 2: choose one or two metrics per decision

Turn each decision into a measurable metric with a threshold. "Projects that overrun" becomes "projects that have used more than 80 percent of their hour budget at less than 60 percent progress". See 12 KPI examples for small businesses for inspiration.

Step 3: find out which software already holds the data

For almost every metric in a small business, the source data already exists somewhere:

The missing piece is almost never the data itself, but the connection between systems.

Step 4: connect once instead of exporting every month

A manual export is fine for one analysis. For data-driven work it is a trap, because someone has to redo it every week. By connecting the source software, data is synchronized daily to one place. From there you can use it in dashboards, in Power BI or Excel through OData, or in questions to an AI agent. Upload budgets and targets as CSV.

You do not need to commission a data warehouse for this. If you later want your own model in Power BI, a star schema helps separate facts and dimensions cleanly.

Want to see what this looks like with your own software? Request a 20-minute demo. We connect one source and show your own figures, without a sales process.

Step 5: record definitions

Agree what revenue, margin and an active project mean, and write it down. Small differences in definitions are the main reason people start to distrust dashboards. Read more about a semantic layer for KPI definitions.

Step 6: make it a rhythm

Data-driven work only becomes a habit when the figures are on the table at a fixed moment. Three rhythms that work well in small businesses:

• Monday morning: an automated email overview with five metrics for the management team.
• Weekly project meeting: the list of projects above the threshold.
• Month-end close: monthly figures against budget, with a short commentary.

In between, ask follow-up questions to the AI agent instead of requesting a new export.

How you know it works

After three months, measure three things: how much time producing reports takes, how quickly a deviation is noticed and whether decisions were made earlier or differently. If the answer to the last question is no, the problem is not the data but step 1.

Not sure which questions to ask yet? Start with 50 questions to ask your business data. Read also how to centralize business data.