The losing ad you're about to pay for again
Last year's losers are still in your account. Find them before Q4 prices hit.
Right now, brands are planning November.
The best input for that plan is whatever last November taught you, and in a lot of accounts nobody can find it.
The lessons are buried in an old Slack thread, or it walked out the door with a former media buyer, or it's sitting in a dashboard that shows the results but you can’t see why it worked.
Lessons compound like interest, but only if they're kept.
Skip it and you end up paying twice just to rediscover the same lessons.
What Happens to Knowledge In An Ad Account?
You've watched this happen.
A test wraps up. Someone posts the result in Slack, a couple of people react, and the thread scrolls out of view by Friday.
The winning ad gets archived under a name like "BF_new_v3_final."
Six months later, a new buyer pitches the exact same test. Nobody can say for sure it already ran, so it runs again.
You pay for the spend, plus another trip through the learning phase, to find out something the account already knew.
Now picture the same thing with a test that lost… A winner at least leaves an ad behind that someone might stumble across.
A loser leaves a paused ad set and nothing else, so the same bad idea looks brand new to whoever finds it next.
Ads Manager keeps a perfect record of what every test costs. It has no field for WHY the loser lost.
Owning your affiliate program data is one of the easiest ways to scale TikTok Shop.
Your outreach volume is vastly less important than knowing what's actually working (mass DMs vs curated lists - truly doesn't matter)... find a way to see which creators, products, and messages actually produce sales.
> Know which messaging copy converts and which gets ignored
> Know which of your creators and products actually drive sales
> Know where people drop off between message, sample, video, and sale
> Fix the biggest leak first, then repeat
The problem is almost no program can answer these, so the whole channel runs on gut feel.
That's what makes Reacher's API + MCP connection worth knowing about.
Your program data plugs straight into Claude and you build whatever you need on top of it.
And it's being given away free - just set up your Reacher API key and it runs, their team helps with setup.
It's extremely simple - yet most programs never look. Go here to get started:
One Row For Every Test
The fix is one shared doc with one row per test.
Each row has six fields:
- Hypothesis
- Single variable you changed
- Dates
- Spend
- Verdict
- One sentence on why
The why is the field doing the work. "Lost" tells the next buyer nothing.
"Lost because a promo email hit in week two and muddied the read" tells them the test never got a fair shot.
The doc only stays useful if nobody treats it as optional.
A test counts as finished when its row is filled in, and a Slack post about it doesn't get it there.
Write the verdict the same week the test ends, while people still remember the stockout or the CPM spike that skewed it.
Wait a month and that context is gone, and the row turns into a number with no story behind it.
That goes for losers too. Log them with the same care as winners, because a dead end you wrote down is the cheapest lesson in the account.
You only pay for it once.
Even an inconclusive result earns a row, with a note on what got in the way, so the rerun happens in a clean window.
Open a doc today and set up the six columns before your next test ends.
The Angle Map Sits On Top
Give the log a few months and patterns start showing up across rows.
The same angle keeps winning with one audience and dying with another.
Those patterns deserve a page you'll actually open when you plan creative, and that's where the angle map comes in.
- Angles that are proven
- Angles that are disproven
- Angles nobody has tested yet
Angles mean the reason this audience should buy.
A holiday shopper might respond to gifting, while a skeptic needs to see the price per use. Hooks and formats belong in the log.
The map only tracks which reasons to buy are working with each audience.
Every new row in the log updates the map.
In November, before brainstorming new ads, open the map for each audience.
Anything on the untested list becomes a test. Anything on the disproven list stays off the slate, because you already paid to learn that.
The catch: to fill the log with past tests, you have to find them in Ads Manager, and that's hard when ads are named by date or "final."
Name each ad after its angle and what you changed, like "Gifting_HookQuestion_UGC_v2" (angle, hook, format).
Then you can search one angle and see every test on it.
Rename the ads in one live campaign to the angle-and-variable format this week, then name everything new that way as it launches.
Start Backfilling Before You Forget
Backfill every test from the last 90 days while people still remember them. Rough rows are fine. A best guess at the “why” beats a blank cell.
If you only get through part of it, start with anything you tested last October and November. That's the stretch of history your Q4 plan leans on hardest.
Start a row for every test that's live right now.
Add ten minutes of "what did we learn" to your weekly creative review, in the same slot where you read the comments.
Pre-committing your BFCM scorecard was the once-a-year version of this. The log is the weekly version that will save you for months and years to come.
Every lesson you write down is one you never pay for again.
At Nord Media, we focus on profitable growth at scale, we help forecast revenue, and contribution margin goals, then build the marketing plans to execute those goals.
This means full clarity on testing, scaling and improving profitability at every aspect of your funnel.
If you want a growth partner aligned with your real business goals, let’s chat.
-
-
-
