Reference
What to check before relying on a number
A recent connection check is not the same as recent data. Start with the number's context.
- Compare the same reporting period.
Use completed periods when comparing performance. A month-to-date total and a completed month answer different questions.
What to inspect: Check the period on both readings. A cash balance is a point-in-time figure, so use its balance date instead of treating it as a monthly total.
- Read the unit with the value.
Sondola records the unit with the value. A runway shown in days and one shown in months need the same basis before you compare them.
What to inspect: Check currency, counts, percentages and time units. Revenue and cash can both be expressed in dollars while measuring different things.
- Check which direction is better.
The direction of improvement belongs to the metric. Lower churn is better. Lower revenue is not.
What to inspect: Read the direction label with the change, rather than assuming that every increase is good or relying on color alone.
- Keep missing values separate from zero.
A missing reading or a rate suppressed for an insufficient sample does not establish a zero. Unknown obligations are not treated as zero in the cash view.
What to inspect: Look for the reason a value is unavailable. Do not fill the gap with zero or use it in a comparison as if it were a measured result.
- Separate connection status from data coverage.
The source's last check, its latest data-through date and the metric's reporting period describe different things. A successful check can still return no usable readings.
What to inspect: Check whether a new value is due. A completed monthly reading does not become late just because the next day has arrived.
- Read the correction history.
When a reading has a recorded correction, its history identifies the previous value, replacement value and reason. Product-wide fixes and improvements appear on the corrections page.
What to inspect: If a historical figure changes, inspect the recorded reason before comparing it with an older export or report.
- Inspect the source and definition.
Open the metric to see its definition, source and reporting period. Movements beneath a change show how a total was formed. They do not establish what caused customer behavior.
What to inspect: When two tools disagree, compare their definitions and periods first. An event count, a user count and a subscriber count are different measures.
- Recognize demo and illustrative data.
The public demo is a read-only fictional workspace. Its figures come from a simulated company, and how the demo is built says what the model is calibrated to. Illustrative previews on the marketing site are labelled separately from live demo readings.
What to inspect: Use the demo to explore the workflow and source context. Its figures are not evidence of customer results or your own company's performance.
Which date are you looking at?
- Reporting period
- The interval the metric measures, such as a completed week or month.
- Balance date
- The point in time a balance represents. It is not a total accumulated over a month.
- Captured date
- When the reading was recorded. This does not change the period the reading covers.
- Last source check
- When the connection was checked. A successful check does not guarantee usable business data.
- Data-through date
- The latest date covered by the available source data. Compare it with the period you need.
If a number still looks wrong
Compare the same metric, unit and period in the source tool. Then inspect the source status and any correction history. The setup and troubleshooting guide explains the next checks to make. For a worked example of the date differences, see Data freshness: a worked example.
These checks make a number easier to inspect. They do not guarantee that the source data or the way you interpret it is correct.
The live demo runs on a fictional company. Every figure shows its source, period and read date.