Counting answers most questions
Which customers are worth most, which products actually make money, where orders come from, what changed this quarter. Those are counting and grouping questions, not modelling ones.
The analysis that changes decisions in a small business is nearly always simple. The sophisticated work is usually a way of avoiding a conclusion the simple work already gave you.
Questions worth answering
- Which customers generate the most profit, not the most revenue?
- Which products are actually profitable after returns and handling?
- Where do good customers come from, as opposed to all customers?
- What is the repeat rate, and how has it moved?
- What changed between this period and the same period last year?
Small data needs care
- One large order distorts a monthly average
- Seasonality dominates in most businesses
- Small samples produce large apparent changes
- Correlation with fifty records is usually coincidence
- Compare against the same period last year, not last month
Where the data actually is
| Source | Contains |
|---|---|
| Accounting system | Revenue, costs, customers |
| Order or job system | What was sold and when |
| Website analytics | Where visitors came from |
| CRM | The sales process |
| Support system | What goes wrong and how often |
Combining two of those usually produces the interesting answers, and it is frequently the first time anyone has looked at them together.
Present it so it changes something
An analysis that produces a chart and no decision was not worth doing. State what you found, what it implies and what you would do differently.
One clear finding acted on beats five interesting observations filed away.