The challenge
Clean Skin Club sells skincare on TikTok Shop and runs paid ads on its creators’ videos with spark codes. Every brand running creator ads at volume has to answer the same four questions:
- Spark codes: what is each boosted video’s angle, content type, offer and funnel stage?
- Angles: which angles are working, and do creators follow this month’s brief?
- Reporting: how does the director see the week without someone pulling the numbers?
- Operations: how do the VA reports get built without days of copy and paste?
Every spark code was logged by a person watching the video. With thousands of videos, nobody could say which angles worked, and the VA reports were filled in by hand every two weeks. I built each piece against the team’s own sheets and reports, so the output landed where they already worked.
What I built
The spark-code sheet fills itself
The gap: every spark code needed its angle, content type and offer logged by hand.
- Each week the spark codes creators shared are tied to their videos and read from the transcript and the video itself.
- Each one is tagged for funnel stage, content type and offer using a written rule set the team can read.
- Rows go into the team’s existing spark-code sheet in the exact columns they already use, and a summary posts to the team’s Slack channel.
Now: 749 spark-code videos have been labeled so far, an estimated 61 hours of watching and typing the team no longer does.
Every video graded against the brief
The gap: there was no way to see which angles were working across thousands of creator videos.
- An angle tracker reads every video posted since October 1 for its hook, selling points and content angle.
- Each video is matched by meaning to the month’s creative brief, and a second AI pass re-checks every on-brief flag.
- A live dashboard, with a tab per angle, shows the team which angles are being made and which ones sell.
Now: 4,371 videos graded against the brief, and the team sees which angles convert.
The Monday exec report, with nobody preparing it
The gap: the director’s weekly read meant pulling numbers from the shop, the ads and the samples by hand.
Every Monday a slide deck lands in the team’s Slack channel.
| The report covers | So the director sees |
|---|---|
| Program performance against last week | Whether affiliate sales moved, and by how much |
| Top creators and creator retention | Who is driving sales, and who keeps posting |
| Ads and top creatives | Which boosted videos are earning their spend |
| Marketplace trends and what is working | Where to point briefs and samples next |
Now: four straight weeks of Monday reports at the time of writing, with nobody on the team preparing them.
VA reports rebuilt on live data
The gap: the VAs rebuilt the same two reports by hand every two weeks.
- Both reports are rebuilt on live shop data with the same layout the team already reads, plus PDFs for Slack.
- Anything the data source cannot supply shows as not available, never as a zero.
- The first automated report covered 2,279 samples.
Now: the VAs review the report instead of building it.
An A/B test workflow in Cruva
The gap: there was no side-by-side way to compare two versions of creator outreach.
- Set up an outreach workflow in Cruva as an A/B test, so two versions run side by side and the team compares the results.
Now: the team can test an outreach change before rolling it out.
What this means for your brand
If your team spends its week tagging videos, checking briefs and rebuilding reports, I can put the same pieces in place for you:
- Spark codes tagged by angle, offer and funnel stage as they come in
- Every creator video graded against your brief, with a live dashboard
- A Monday exec report delivered to Slack
- Recurring team reports rebuilt on live data
I build it with your team so you own and run it after I leave.