Step 1: Is the change real?
Before anyone builds a theory, rule out the boring explanations. A surprising number of “drops” are data, not business:
- Incomplete period. Comparing a month that has 24 days of data with a full month always looks like a fall.
- Late data. One branch’s sales did not load overnight; the total is down by exactly that branch.
- Changed definition. Someone excluded refunds, or VAT, from “revenue” last week.
- Calendar effects. Four weekends this month against five last month; Ramadan, Diwali or Christmas moving between periods.
Check the update time on the dashboard and compare like with like — same number of trading days, same definition. If the change survives, move on.
Step 2: How big is it, really?
Every KPI wobbles. Before explaining a change, decide whether it is bigger than the usual wobble. Put the KPI on a 13-period trend: if weekly revenue usually moves between −5% and +5% and this week is −3%, there may be nothing to explain. If it is −12%, there is.
Size it in money as well as percent. “Margin fell 1.5 points” lands differently when it is written “we earned 38,000 less gross profit than last month”. The amount decides how much time the explanation deserves.
Step 3: Split it into drivers
Most KPIs are made of simpler numbers multiplied or divided together. Split the change into those parts:
- Revenue = number of orders × average order value. Did fewer people buy, or did they spend less?
- Gross margin moves with price, cost and mix — selling more of a low-margin line lowers the average even if no price changed.
- Conversion = orders ÷ visitors. Did traffic change, or did the same traffic buy less?
- Labour cost % = labour cost ÷ sales. Did hours go up, or did sales go down with the same rota?
A worked example: revenue is down 9% on last month. Orders are down 2%, average order value is down 7%. So the story is mostly “customers spent less per order”, not “we lost customers” — a very different conversation.
Step 4: Break it down until one cause stands out
Now take the driver that moved most and break it down by each dimension in turn — region, store, product category, customer type, sales channel, salesperson. Sort each breakdown by the size of the change, not by the value. You are looking for the one or two rows that account for most of the movement.
Continuing the example: average order value is down 7%. By channel, online is flat and stores are down 11%. By store, three of twenty stores account for 80% of the fall. In those three, the bundle promotion ended on the 1st. That is the explanation: “Revenue fell 9%, almost all from lower basket size in three stores after the bundle promotion ended.”
If no single row stands out — every region down a little — the cause is usually company-wide: a price change, a seasonal pattern or a market shift.
Step 5: Write the one-line answer
An explanation people will repeat has four parts: the change, the size, the cause and the action. “Margin fell 1.5 points (38k) because freight costs doubled on imported lines; we are re-quoting two carriers this week.” Put that sentence at the top of the dashboard or the report, above the chart that proves it.
Resist listing every factor you found. If the top cause explains 80% of the change, lead with it and mention the rest as “smaller effects”. Readers remember one reason.
Making the next explanation faster
- Define the KPI once, so nobody spends the first hour arguing about what it includes.
- Build the driver KPIs next to it — orders and order value beside revenue — so step 3 is already on the page.
- Keep the breakdowns one click away by drilling from the tile into the rows behind it.
- Use automated insights that flag the biggest change and top movers, so step 4 starts with a shortlist.
In Klayara, every dashboard shows free automatic insights — the biggest change, top movers and unusual values — computed within each reader’s permissions. With Klayara AI, click “Why did this change?” on a figure for a plain-language explanation of what moved, or “Summarize this dashboard” before a meeting. It runs only when you click, usually for 1 credit. See AI dashboard insights.