Lesson 10 of 10 · Lab
Fix a data flow when next week's file changes
A flow runs the same steps on whatever file you give it. Some changes to the file stop it with an error. Others let it run and report the wrong number.
A loud failure turns something red. A silent one shows only against what you expect, such as a normal week's size.
It's Monday, September 28, at fictional Kramerica Bikes, and Tomas has sent Harbour's export to September 27. This live flow in Strata, Oplema's browser app for data prep, starts after Select: Manual Input holds Tomas's new file, already cut to your six columns, and Filter is set to September 21 to 27. His till was updated on September 23, and nobody said how.
Last week had 93 finished hires. If this week's flow reports 25, is it right?
Probably not. A week a quarter of the usual size calls for a look at what the filter rejected.
Repair the flow for next week's file
- Summarize opens first, with no result. Select Number: its Column reads
Unavailable: amount, because no column of that name reaches it any more. - Select Manual Input and read the new file's columns. The till now calls the column
price. - Select Number and set Column to
price. - Select Summarize. It reads 25 hires and a
netof 611.60. That's a quarter of a normal week, and nothing is red. - Now check what Filter rejects, without step-by-step help: add a Summarize from Transform, wire Filter's False output to it and count its rows by
status.If you're stuck
Two actions: Group by
status, then Count. - Read your table. Beside
completed,cancelledandno-showthere'sreturned: 83 rows the till began writing on September 23. No row readscompletedfrom that day on. - Select Filter. In Expression, change
status = 'completed'tostatus IN ('completed', 'returned'), then select the Filter block. If you break it, press Ctrl+Z or Cmd+Z to undo. - Select the first Summarize. It reads 108 hires, 155 bikes and a
netof 2876.40.
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).
What you should see
After the first fix, 25 hires and €611.60 net. Your count by status shows 83 returned rows on False, worth €2,831.00 at till prices. After the second fix, 108 hires, 155 bikes and €2,876.40 net.
A flow that runs can still be wrong
The rename stopped the flow. The new status didn't: the rule still ran but no longer meant every finished hire, and only the result's size gave it away.
So keep a check you'd notice, such as the False output counted by reason. Harbour's report week is 93 hires and 159 bikes, and its busiest-looking day rests on one 26-bike booking. That's Tomas's answer for now, not Marit's. Her reply: "I asked about all three, and it's month-end on Thursday."
For later: Checking a dashboard after a refresh explains what a successful run does and doesn't prove.
Common repair mistakes
- Fixing the error and stopping. The rename was one change of two.
- Renaming the column back in the file. Next week's export will say
priceagain. Fix the flow instead. - Reading only the kept rows. The new status sat in the False output the whole time.
Changed files in SQL and Excel
SELECT status, count(*) AS rows
FROM harbour_to_sep_27
WHERE NOT (status IN ('completed', 'returned')
AND hire_on BETWEEN DATE '2026-09-21' AND DATE '2026-09-27')
OR hire_on IS NULL
GROUP BY status;A query that names amount fails on the new file in the same way. In Excel, a SUMIFS on "completed" keeps working and quietly skips "returned", the same silent failure.
Course complete Now use your own file.
One shop's weekly report for fictional Kramerica Bikes, shipped as a file and a chart, and a flow you've repaired after the export changed.