Lesson 13 of 15 · Lab
How to check a dashboard after a data refresh
Refreshing a dashboard means running its data through the same steps again with newer files. A successful run tells you those steps finished, and that's all it tells you. So check the number after each refresh, and show readers the date the data runs through.
| Result | Tells you | Doesn't tell you |
|---|---|---|
| Run succeeded | Every step finished | That the number still means the same |
| Rows changed | New data arrived | Which rows dropped out, or why |
| Run failed | The page is out of date | How old its numbers are, unless the page says so |
At fictional Vandelay Coffee Supply, Q3 net value stands at €11,478, 95.7% of target. Invoice INV-1014 is unpriced, and any price of €522 or more reaches €12,000. Sydney Swell in Finance sends the October export, with a price for it at last.
INV-1014 comes in at €620. What will the Q3 total be?
You'd expect €12,098, and you'd be wrong. The file holds another change, and the flow has no reason to mention it.
Connect the October file
- Select Summarize, the last block of this live flow in Strata, Oplema's browser app for data prep. Note
net. - Drag from the right dot of the lower Manual Input to the left dot of Union. Drop on the dot, not the block.
- Select Summarize again. Nothing is red.
- Select Filter, and open OUTPUT FALSE. Find INV-1013.
Swipe sideways to see the canvas, or open this page on a larger screen to work with it.
In an expression, press Esc to apply it and move on (press it twice if a suggestion list is open).
The total falls, and here's why
net falls to 11,198, or 93.3%, which is €280 lower. South rises €620, to €6,520. North falls €900, to €4,678. INV-1013 is now part-paid, and the rule keeps only booked.
The run succeeded and the number fell
Nothing failed. The rule did what it said, about a status nobody had considered. So put the rejected rows where someone will read them, and name that person. After each refresh, check row counts, required fields, customer matches and any new status.
Readers need a "data through" date, separate from the time they opened the page. A pipeline that ran this morning can't make a late export any more current. When a refresh fails, that label should stay on the last good date, so nobody takes old numbers for today's.