Lesson 6 of 7

Lesson 2.1.6

Not Found is not empty: the filter mistake that costs the most

6 min readIntermediate

What this costs

Free. This lesson saves credits rather than spending them.

Before you start

A column with some blank cells in it.

Three different things look identical in the table and mean completely opposite things. A cell the tool never ran on. A cell the tool ran on and found nothing. A cell the tool ran on and returned an empty value for. Filtering on "is empty" treats all three as the same, and that is where the money goes.

The three states

StateWhat happenedWhat it costsWhat to do
SkippedThe tool never ran. A filter or condition excluded the row, or a required input was missingNothingFix the input or widen the condition, then run it
Not foundThe tool ran and genuinely found nothingNothing on a charged on success toolAccept it, or try a different tool. Re running the same tool changes nothing
Returned emptyThe tool ran, concluded, and the answer is legitimately blankYou were chargedTreat it as an answer, not a failure

How to tell them apart

A Kuration people column sitting at 36 percent, with some rows returning a profile and others reading Skipped
1
2
3
  1. 1Thirty six percent, in green, red and amber. Completion, rejections and skips, in one bar.
  2. 2Skipped. The tool never ran here, so nothing was charged and nothing was missed.
  3. 3The reason: this row is B2C, and the condition asked for B2B.
A column at 36 percent that is working exactly as instructed. The Skipped rows are the B2C companies a condition excluded, and they cost nothing.
  1. 1

    Open the cell rather than reading the table

    The detail view distinguishes a row the tool never touched from one it processed. This is the whole diagnosis, and it takes ten seconds.

  2. 2

    Check the condition on the column

    If the column has a condition, work out how many rows it excluded. Those are your skips, and they are almost always the biggest group.

  3. 3

    Check the required input

    Most skips are a missing dependency rather than a deliberate condition. A people tool with no website. A domain tool with no company name. Fix the input column and the skips resolve themselves.

  4. 4

    Filter on the tool status field, not on the value

    Where a tool exposes a status or match type, filter on that. It is the only filter that distinguishes the three states reliably.

What to do with each group

The Kuration project analytics panel showing per column completion, with completed, running and skipped counts
1
2
3
4
  1. 1Completed. The tool ran and returned a value. 299 of 300.
  2. 2Running. Still working. A dependent column reading this now is the race from lesson 2.1.7.
  3. 3Skipped. The tool never ran on these 15 rows, so they cost nothing and re running the column will not fix them.
  4. 4Total Failed is a fourth thing again: the tool ran and errored. That is the number worth a targeted re run.
Project Analytics counts the three states for you, per column, so you do not have to diagnose them cell by cell. Open it before you re run anything.

Worked example · A people column with 400 blank cells, diagnosed

GroupRowsCauseActionCost to fix
Skipped265No website on the rowRun Website Finder, then re run people on those rows only530 plus the people run
Not found118Tool ran, no person matchedLoosen the role description or accept0, misses are free
Returned empty17Matched a company with no public contactAccept. This is an answer0, do not re run

Without this diagnosis you would have re run 400 rows. With it, you fix 265 properly, adjust the approach on 118, and leave 17 alone. Same outcome, and you know which of your three problems is actually a problem.

Filtered on "is empty" and re ran everything

What you see
Repeat spend on rows that will always be blank
The fix
Diagnose first. The three states need three different responses.

Concluded the tool has poor coverage

What you see
A support ticket about data quality
The fix
Check how many of the blanks are skips. On a list with missing websites this is usually most of them.

Kept re running to fix a not found

What you see
Same blank result, every time
The fix
A not found is a real answer. Change the approach or the tool, not the number of attempts.

Check yourself

0 of 3 answered

1.Which of the three blank states costs you money if you re run it?

2.Your people column shows 400 blanks. What is usually the biggest group?

3.Where can you see completed, running and skipped counted per column?

Answer all 3, then open Let a column finish before the next one reads it to tick this lesson off.

Try it yourself

Take your worst performing column. Count the blanks, then sort them into skipped, not found and returned empty. Write the three numbers down.

You have done it when
The three numbers add up to your blank count, and you have a different planned action for each group.
Credit budget
Free.

Related and next