THIS ISSUE:
Part 2 of getting your mornings back
The report that's waiting when you wake up
Last Wednesday you gave every client a brief your AI can read.
Today we put it to work.
One prompt, set to run every morning: what moved, why, and the one thing you need to do today.
Five clients. 100 minutes ☹️.
Gone 😎.
Set it up once. Read it with your coffee.
THE IDEA
A REPORT THAT TELLS YOU WHAT TO DO
Not just what happened
A dashboard gives you numbers. You still do the thinking. Is this dip normal? Does it matter for this client? What do I do about it?
That's the 20 minutes.
The brief from Part 1 fixes it. Your AI already knows the hero product, what the founder hates, what flopped in March. So when GMV drops 30%, it can tell you whether that's news or noise.
Numbers from EUKA. Judgement from the brief. Every report ends with one action.
No brief yet? Do Part 1 first. It takes twenty minutes and nothing below works without it.
WHAT YOU NEED
THREE THINGS
and you built two of them last week
The client's project from Part 1, with the brief saved inside it.
EUKA connected to your AI. Same test as last time: ask "What brands can you see in EUKA?"
Ten minutes. That's the whole setup for your first client.
HOW IT WORKS: STEP 1 OF 5
PASTE THE MORNING PROMPT
Open Brand A's project and paste this, with the client's name swapped in. Same deal as last week: copy it, keep it, hand it to whoever runs the account.
It only reads. It can't message a creator, approve a sample or touch an ad.
You are my TikTok Shop analyst. You have access to the EUKA MCP.
Each morning you write me one short report on one client.
Client: [BRAND NAME]
BEFORE YOU PULL ANYTHING
Read the brief in this project. It tells you the hero products,
the targets, what the founder cares about and what to ignore.
If there is no brief, stop and tell me.
This is a read-only job. Do not send messages, approve samples,
change ads or edit anything in EUKA. You report. I act.
WHAT TO PULL
Pull yesterday (the last full day), the 7 days before it, and
the same weekday last week.
1. Sales
- GMV, orders and average order value
- GMV split by where it came from: creator videos, lives,
ads, and the shop itself
- Top 3 products by GMV, and any product that sold last week
and sold nothing yesterday
2. Creators
- New videos posted yesterday, and which ones made sales
- Top 5 creators by GMV over the last 7 days
- Any creator who made sales in the 7 days before and has
posted nothing since
3. Samples
- Requests waiting for a decision, and how long the oldest
has waited
- Samples approved and shipped in the last 7 days
- Samples delivered more than 14 days ago with no video yet
4. Outreach and inbox
- Messages sent yesterday, replies, and the reply rate
- Creator replies sitting unread, and the oldest one
5. Ads (GMV Max)
- Spend, sales from ads and ROI, yesterday against the
7 day average
- Any campaign that stopped spending or hit its budget cap
- Any video carrying most of the spend on its own
6. Shop health
- Current score
- Any new violation, warning or flag since yesterday
HOW TO JUDGE IT
- Compare yesterday with the 7 day average AND the same weekday
last week. Only call it a move if both agree.
- A move is 20% or more. Ignore percentages on tiny numbers:
under 10 orders, give me the count instead.
- Use the brief. A dip on a product the brief says is being
run down is not news. A dip on the hero product is.
- If the brief gives a target (ROI, GMV, samples a week),
measure against the target, not just against last week.
- Yesterday's numbers can still be settling. If a figure looks
incomplete, say so instead of calling it a drop.
PICKING THE ONE THING
Choose one action for today. If several things need doing,
use this order:
1. Anything that threatens the shop itself: a violation, a
health score falling, a listing taken down
2. A hero product out of stock or close to it
3. Ad money going out with no sales coming back
4. Creators left waiting: sample requests or replies older
than 48 hours
5. A creator or video that is suddenly working and needs
backing
6. Everything else
If nothing needs doing, write "Nothing to act on today."
Never invent a problem to fill the space.
WRITE THE REPORT IN THIS ORDER
- THE ONE THING: the action, and why, in two sentences.
- THE NUMBERS: a small table. GMV, orders, ad spend, ROI,
new videos, samples waiting, messages sent, replies.
Yesterday, 7 day average, same day last week.
- WHAT MOVED: each move of 20% or more, with the most likely
reason. Say "likely" when you are guessing.
- CREATORS: who sold, who went quiet, who is waiting on us.
- WATCH LIST: not a problem today, will be by Friday.
- MISSING: anything above that EUKA did not return.
RULES
- Under 300 words. I read this on my phone.
- Never estimate or fill in a number. If EUKA does not have
it, put it under MISSING.
- Leave a row out of the table if EUKA returned nothing for
it. List it under MISSING once, not twice.
- If the shop made no sales in the period, say so in one line
and skip the sales and creator rankings.
- Name the creator, the product or the campaign. "A creator
stopped posting" is no use to me.
- Plain English. No jargon, no cheerleading.
- Put the date range you pulled at the top.HOW IT WORKS: STEP 2 OF 5
RUN IT ONCE, BY HAND
Before you automate anything, run it yourself and read what comes back.
Open EUKA in another tab and check three numbers against the report: samples waiting, messages sent, replies. If they match, you can trust it. If they don't, ask it where it got them.
Then read the last line. MISSING is the report telling you what it couldn't get, like ads that aren't connected yet. An honest gap beats a made-up number.
Two minutes. Do it once and you won't lie awake wondering what it's telling you at 6am.
HOW IT WORKS: STEP 3 OF 5
TELL IT WHAT YOU DON'T CARE ABOUT
The first report will flag something you already know about. That's fine. That's the job of this step.
Every time it gets the judgement wrong, don't argue with the report. Fix the brief. One line: "Ignore dips on the gift set, we're running it down."
The numbers live in EUKA. The judgement lives in the brief.
Three or four runs and it stops crying wolf.
HOW IT WORKS: STEP 4 OF 5
SET IT TO RUN EVERY MORNING
Now take your hands off the wheel.
ChatGPT and Claude both have a way to run a prompt on a schedule. The clicks change by tool and by plan, so ask yours: "How do I run this every morning at 6am?" It will walk you through it.
Run it inside the client's project, not a fresh chat. That's how it reads the brief first.
Set the first run for ten minutes from now, so you know it works before you rely on it.
Not ready to automate? Paste the prompt by hand each morning. Same report, one extra click.
HOW IT WORKS: STEP 5 OF 5
READ THE ONE THING, THEN DO THE REST
Tomorrow morning, read the top line. That's it.
If it says "Nothing to act on today"….. believe it, and close the laptop.
Then repeat steps 1 to 4 for every other client. Same prompt, new name, new project.
Five clients takes one afternoon. After that, the 100 minutes are yours.
THAT'S A WRAP
What did you think of today's tip?
Until Friday,
- Paul 👊
P.S. If you are lazy like me….. don't bother reading this, just paste this email's link into your AI and tell it to set this up for one client.