MIXEDData strategy for SMEs: from Excel to decisions
Every SME has data. The problem is that it lives in twenty Excel sheets, three management systems and the memory of two people. "Data strategy" sounds like something for multinationals, but for a company of 20 or 80 people it means something simple: having the right, reliable numbers when you need them to decide. Here's how to build it, in four steps.
Why data in Excel never becomes decisions
It isn't Excel's fault: it's an excellent analysis tool. The problem starts when it's used as a database: several people update different copies, formulas break silently, the "right" version is the one on somebody's laptop. After a few months something very specific happens: nobody trusts the numbers any more, and decisions go back to being made on gut feeling. At that point the data exists but doesn't count.
The symptoms are always the same: the same KPI has different values depending on who calculates it; preparing the monthly report takes days; to answer "what margin do we make on customer X?" you have to ask three people.
Step 1 — Start from the questions, not the data
Write down the 3-5 questions you'd like to answer every week and currently can't. Real examples: "which jobs are losing margin?", "which customers are buying less?", "how many labour hours did I sell and how many did I pay for?". Each question tells you what data you need and how often. Everything else, for now, isn't needed.
Step 2 — One source for each data point
For each data point you need, establish where it originates and who is responsible for it. Attendance originates on site, not in the office. Orders originate in the management system, not in a "summary" sheet. If a data point today starts on paper or in a personal spreadsheet, that's the first thing to fix: it must be captured where it's generated, once, with a tool that makes it immediately available to everyone else.
Attendance as the first "certain" data point. In the construction companies we work with, the most retyped data was worker attendance: a sheet on site, a photo on WhatsApp, transcription into Excel at the office, sending to the payroll adviser. Four steps, four chances for error. With V-Site attendance originates from the worker's QR clock-in and is never touched again: it's the same for the site manager, the office and payroll.
It was those companies' first reliable data point — and from there deadlines, vehicles and site costs were connected to the same system. Clients tell us about hours of work recovered every week, but the biggest value is something else: for the first time the owner looks at a number and believes it.
Step 3 — A few KPIs, calculated by a machine
For each question define one indicator with a written, agreed formula (what goes into "margin"? does overtime count in "labour cost"?). Then make sure it's calculated automatically from the source, not rebuilt by hand every month. A dashboard that updates itself, even a simple one, is worth more than a beautiful report that arrives three weeks late.
| Question | KPI | Source | Frequency |
|---|---|---|---|
| Which jobs are losing margin? | Estimated margin per job (cost-to-cost) | Costs from attendance, materials, vehicles | Weekly |
| Which customers are buying less? | Revenue per customer vs same period last year | Sales system | Monthly |
| Hours sold vs hours paid | Ratio of invoiced hours / clocked hours | Clock-ins + invoices | Weekly |
| Deadlines at risk | No. of deadlines within 30 days without an owner | Deadline tracker | Daily |
Step 4 — Quality rules (few, but real)
Data gets worse on its own if nobody looks after it. Three rules are enough: mandatory fields where data originates (no clock-ins without a site, no orders without a customer); a 15-minute weekly check for anomalies (missing values, duplicates, outliers); one owner for each source. In large companies this is called data governance. For an SME it's a checklist.
And artificial intelligence?
It comes later, and it works better. Forecasts, assistants that answer in natural language, automatic alerts: they all work only on reliable data. Those who skip the four steps and start with AI get very convincing, wrong answers. We cover this in 7 concrete uses of AI for SMEs.
Rule of thumb: if answering a business question takes more than 10 minutes and more than one person, that data doesn't have a home yet. Finding it one is your data strategy.
Frequently asked questions
What is a data strategy for an SME?
One source for each data point, a few KPIs with agreed formulas, an automatic dashboard and minimal quality rules. Not a document, a practice.
Isn't Excel enough?
It's enough for analysis, not for collecting and sharing: different copies produce different numbers and distrust.
Where do you start?
From the 3-5 questions you'd like to answer every week and can't.
What are your 3 unanswered questions?
Bring them to us: in a free 30-minute consultation we'll tell you what data you need, where to get it and with which tool.