How to find out what changed in your sales data

Someone asks "why are sales down?" and the honest answer is usually "it depends where you look". The total is the last place the answer shows up. This guide is the order of questions that gets you to it fastest, with or without software.

1. Put the numbers on a time axis

Total by month (or by week, if you have enough rows). Draw it as a line. A table of figures hides the shape; a line shows you in seconds whether you are looking at a sudden drop, a slow slide or one odd month.

2. Decide which kind of change it is

• A step: the level changes at one point and stays changed. Something happened on that date: a price change, a new competitor, a tracking fix.
• A trend: it keeps moving the same way. Look for something gradual, such as a market shifting or a product ageing.
• A blip: one period is off and the next is back to normal. Often a late payment, an outage, or a one-off order.

3. Compare like with like

If your business has a busy season, compare March with last March, not with February. A quiet January after a busy December is not a decline. You need at least two full years of data before a seasonal comparison means much.

4. Split it by segment

Break the same line by region, product, channel or customer type. A flat total can hide one segment falling and another rising. A fall that sits in one segment points at a cause; a fall spread evenly points at the market or at how you measure.

5. Check what you are counting

• Did the number of records change, or the value of each? Fewer orders and smaller orders are different problems.
• Is all of the data there? A file that stopped loading halfway through a month makes that month look like a collapse.
• Did a definition change? A new way of recording refunds will move a total on its own.

6. Confirm with a second source

Before you act, check the finding against something that was collected separately: bank deposits, a payment provider report, customer counts from another system. If two independent sources agree, you can act. If they disagree, you have found a data problem, which is also useful.

Doing this in AnalyzeIt

Ask "what stands out in my data?" and AnalyzeIt runs these checks across your tables. Each result is a card with a plain title, a chart, one sentence on what it means, and a "Show the working" panel with the figures and the exact query. Findings are labelled by how sure the evidence is, and it says how many checks it ran and what part of your data it could read.

Read more: How discovery works, or try the live demo with no account.