Slide 1 of 4
Every fix has two addresses
You can correct this asset, or correct the thing that produced it. They take about the same time and only one of them scales.
Running a campaign with drafted content
DoLesson 11 of 12Review at volume
Editing the pattern instead of the instance
Turn what the sample found into upstream changes, so a single fix applies to the whole batch instead of one asset.
22 minIntermediate
Part 1 of 3
Diagnose before you edit
When the sample turns up something wrong, there is a fork that decides whether the next hour is spent well. You can fix the asset in front of you, which takes two minutes and resolves one page. Or you can fix what produced it, which takes about the same time and resolves every page carrying the same defect, including the ones you did not sample.
Choosing correctly needs a moment of diagnosis rather than a reflex. The question is not whether this is wrong; it is why generation produced it, and the answer routes the fix to a different place depending on which input was responsible.
- A wrong claim or missing proof point → fix the artifact library
- Tone that is off across several assets → fix the sequence style attributes
- The wrong argument being made at all → fix the campaign brief
- A problem specific to one account's situation → fix that asset
Part 2 of 3
Telling a one-off from a pattern
The genuinely hard judgment is that a pattern often shows up exactly once in a sample of twelve. Seeing something once is not evidence that it is isolated. It is evidence that it occurs at least once in twelve, which across a hundred-asset batch is a meaningful rate and probably eight more instances you have not looked at.
The cheap resolution is to check rather than infer. When something might be systematic, pull two or three more assets that would exhibit the same defect if it were, and look. That is a few minutes of work, and it converts a guess into a finding, which is the difference between fixing the campaign and fixing a page.
Part 3 of 3
Verify the upstream change
Changing an artifact or a style attribute changes how everything generates, which is exactly why it is worth doing and exactly why it needs verifying. Regenerate the affected assets and draw a fresh sample, including at least one asset that was fine before, to confirm the fix did not introduce a new problem somewhere else.
That loop is the whole skill of working at volume: sample, diagnose, fix upstream, regenerate, re-sample. Teams who learn it review large campaigns in a morning. Teams who do not spend a week hand-editing assets and ship a campaign whose actual defects were never addressed, because every fix was applied one page at a time to a problem that existed in a hundred.
Do this in the product
Work the sample into upstream fixes
- List every issue found in your sample, one line each.
- Against each, write the cause: artifact, style, brief, or genuinely account-specific.
- For anything that might be systematic, pull two more assets that would show the same defect and check whether they do.
- Make the upstream fixes, starting with artifact edits since those benefit future campaigns too.
- Regenerate the affected assets.
- Draw a fresh sample including one asset that was already fine, and confirm the fix worked without side effects.
Key questions
You should be able to answer each of these from memory before opening it. Recalling the answer is what makes it stick; recognizing it when you read it does not.
You find the same tone problem in three sampled assets. Where does the fix belong?
The sequence style attributes, not the assets. Style governs every future generation, so one change fixes the batch and everything after it.
Why is a defect seen once in a sample of twelve worth investigating?
Because it is evidence of a rate, not of an exception. Across a hundred assets that is roughly eight more instances you have not looked at. Pull two or three more assets that would show the same defect and check.
What should a re-sample after an upstream fix include?
At least one asset that was already fine, to confirm the change fixed the defect without introducing a new problem elsewhere.
What to take away
- Route every fix to its cause: artifact, style, brief, or the one asset. Only the first three scale.
- A defect seen once in twelve is a rate, not an exception. Pull more assets and check.
- Regenerate and re-sample after an upstream fix, including something that was already fine.
Teach this lesson
The argument in 4 slides, for presenting it to your team
Slide 2 of 4
Route the fix to its cause
- Wrong claim or missing proof → the artifact library
- Wrong tone across several assets → the sequence style attributes
- Wrong argument entirely → the campaign brief
- Wrong for this account only → edit the asset
Slide 3 of 4
Once in twelve is a rate, not an exception
Seeing something once in a sample of twelve is evidence it happens at least once in twelve. For a hundred assets that is meaningful.
Slide 4 of 4
An upstream fix is a hypothesis
Regenerate and re-sample, including one asset that was already fine, to confirm the fix worked and broke nothing.
Demo
Try the demo.
See agents carry the repeatable work of GTM across sales, marketing, customer success, and RevOps. Every action prepared, reviewed, and recorded. Fictional data, real product.
Explore the demo