I wrote four posts on what really goes on inside ad agencies, put them on Threads, then looked at the data

— Inside the 4A ad agencies (bonus post)

TL;DR

• How the four-part series performed on Threads:"Half of the big six's methodology is real knowledge, half is sales talk" — 17,527 views, 241 likes, 100 shares, 19 reposts; "This system is being eaten by AI and consulting firms, and it deserves it" — 2,635; "Clients ask for a Big Idea, but what they want is to not get burned" — 953; "Why are ad agencies getting into AI?" — 49. Same account, same series, same posting time, a 358× gap
• In the same period, the most-viewed post on my account was a geography trivia teaser, at 62,229 views — but only 3 shares and 0 reposts.Views are traffic, followers are assets: three weeks of trivia teasers brought 11 followers; the "methodology" post brought 30 in two days
• Why did the "methodology" post take off, and not the "deserves it" post? It's the same thing the "Big Idea" post talked about —people who repost need to feel safe too. "Half real knowledge, half sales talk" is a stance you can put your name on and pass along; "deserves it" isn't
• The people who showed up were peers, not clients. The commenters were ex-agency people with 4A backgrounds; 64% of followers are in Taipei and New Taipei. For a project-based shop, Threads is a credit account, not a sales funnel
• I nearly split the trivia teasers and the opinion posts into two accounts on gut feeling. The data said:no synergy, but no harm either— and the teasers are the only intake channel for another product. Don't split; set a trigger condition for splitting instead

This is a bonus post in the "Inside the 4A ad agencies" series · Series navigation ↓

In the previous four posts I covered a lot: half of the methodology is sales talk, what clients want is reassurance, the industry is being eaten by AI and consulting firms, and how Satsuma does things itself.

After that, I did something ad people usually don't do.

I treated myself as the client, the four posts as the material, put them on Threads, and thenignored my gut and looked at the data。

The reason is simple. In the first post, "Half of the big six's methodology is real knowledge, half is sales talk", I said the real knowledge underneath the methodology is "why people are moved". So with the things I post myself — are people actually moved? And what do they do after? All of that now comes with numbers. Not looking would be a waste.

Below is what came out of lining up four sources: post insights from the official Threads API, the follower curve in the Threads dashboard, GA4, and the database of my own product.

The report card for the four posts

First, the numbers. The four posts, in order:

• 〈Why are ad agencies getting into AI?" — "getting into AI" below
• 〈Half of the big six's methodology is real knowledge, half is sales talk" — "methodology" below
• 〈Clients ask for a Big Idea, but what they want is to not get burned" — "Big Idea" below
• 〈This system is being eaten by AI and consulting firms, and it deserves it" — "deserves it" below

All were posted at 8:30 am, with the argument in the body and the link in a comment.

Same series, same author, same account. A 358× gap between the best and the worst.

After thirty years in advertising, I know "one hit, three that go nowhere" is the norm. But this time there was something I never had before:every post came with complete numbers, all measured with the same ruler. They can be compared.

Views are traffic, followers are assets

First, a number that leads people astray.

This account also runs a second line: three trivia teasers a day — "What's the second-highest mountain in Taiwan?" and the like — to drive traffic to our quiz game. The view counts on that line are staggering. The top one in the same period got 62,229 views, 3.5 times the "methodology" post.

If you only look at views, the teasers win.

But convert it to actions per thousand views and the result flips completely.

A hundredfold difference in share rate.

Then look at followers. On the follower curve in the Threads dashboard, over the three weeks from 7/23 to 8/14 — three teasers a day, well over a hundred thousand views accumulated — followers went from 155 to 166,up 11. The "methodology" post went out on 8/16, and by 8/17,up 30. Then it slowly flattened out again.

A flat line, a step, another flat line. Sixty percent of a full 30 days of follower growth happened in those 48 hours.

Views are traffic, followers are assets. The traffic the algorithm gives you, it takes back the same day; teaser readers answer the question and leave — no follow, no share, no friends brought along. Only the things people are willing toput their own name on and pass alongbecome assets for the account.

From now on I'm not looking at views to judge performance. I'm looking at shares, reposts and follower growth.

Why "methodology" took off and "deserves it" didn't

This is the question most worth thinking about.

The two posts actually had similar share rates — "methodology" got 5.7 shares per thousand views, "deserves it" got 4.9. The difference is inreposts: 19 to 0.

A share is passing something privately to a friend. A repost is publicly hanging it on your own feed — it amounts to declaring "I think so too".

The "methodology" post talks about something industry people carry in their heads but can't say in meetings: half of the methodology exists to give the boss something to sign off on. And the framing is balanced — half sales talk, half real knowledge. Reposting it is safe for the reader: they haven't attacked anyone, they're just saying "I get this". The 23 replies underneath bear this out — ex-agency people adding their own versions, like "My boss said 360-degree marketing: 359 degrees are advertising, 1 degree is the internet."

And the "deserves it" post? It relays a foreign article, then delivers a one-way verdict: this industry deserves to be eaten. Readers might agree after reading, but they have noposition they can stand behind. Reposting means endorsing the words "deserves it", and in industry circles that carries a cost.

So 13 people shared it privately, and 0 reposted it publicly.

I wrote about this in the second post, "Clients ask for a Big Idea, but what they want is to not get burned".Readers reposting follow exactly the same logic. What they want isn't just a good article — it's one that won't get them in trouble when they pass it along, and that makes them look good for doing so.

The difference isn't in the writing. It's in whether you give the reader a safe position to take.

The people who showed up were peers, not clients

Third: look at who came.

The commenters: ex-agency people, 4A veterans, marketing consultants, and "well written, but it's missing the MVP logic" types. Not a single "our company is looking for someone to do our advertising".

Follower geography: New Taipei 36.5%, Taipei 28.4%, Taichung 4.8%. Sixty-four percent in Taipei and New Taipei.

Over in GA4: the "methodology" post brought 729 people to our website, 463 of themfirst-time visitors. After reading the full article, some of them kept going down the series — before the "deserves it" post had even gone out on Threads, its full-article page had already been seen by 46 people.

These are all good things. But you need to be clear about what kind of good they are.

CMOs don't comment on Threads. The people who gather on Threads are peers. Peers won't hand you a project directly, but when a client asks around, they're the ones who speak up; when a client searches, what they reposted is what ranks at the top.

So for a project-based shop, what Threads doesis serve as a credit account, not a sales funnel. Measure it with funnel metrics and it looks like wasted effort; measure it with credit metrics and 729 peers reading a 5,000-character article, 463 of them meeting you for the first time, is a very good deal.

A decision I nearly got wrong

At this point I was about to make a decision: split the trivia teasers and the opinion posts into two accounts.

The reasoning sounded perfectly sensible: an account posting three trivia teasers a day and the occasional marketing opinion piece gets typecast by the algorithm as a "trivia account" and pushed to people who like answering questions. Those people have no interest in articles about ad agencies, so every opinion post has to find its audience from scratch. The two lines dilute each other.

I believed this reasoning for about half a day. Then I went to check it.

Check one: did the 30 new followers from the opinion post make the teasers perform better? 29 teasers before 8/16, 21 after. Median views 1,117 → 437; likes per thousand views 3.4 → 2.4. No lift.

Check two: did the teasers' hundred-thousand-plus views turn into readers for the opinion posts? Three weeks, +11 followers. No.

So the synergy is zero. Two groups of people who never cross paths, living in the same account.

But looked at the other way,the dilution is zero too. The "methodology" post took off while this account was posting three teasers a day. The algorithm did not lock me in. The thing I was worried about is nowhere in the data.

Then I went to look at what the teaser line was actually doing. In our quiz game's database, every new player's source is recorded at sign-up. After we stopped Meta ads on 8/7,85% of new players came from the Threads teasers. One or two a day — not many, but half of them come back to play again after signing up, the same return rate as players brought in by ads.

In other words: splitting the accounts has no known benefit, but it would cut off the only intake channel for another product.

No split.

But I did set a trigger condition for splitting: over the next four weeks, one opinion post of the same type every week. If three in a row drop below 2,000 views while the teasers hold steady, then the algorithm really has typecast the account, and I'll split it then.

This reminded me of what I wrote in the fourth post, "How Satsuma Creative does advertising": real decisions aren't made in the meeting room. This time my "meeting room" was my own intuition, and it handed me a conclusion that sounded very right. Lay out the numbers, and the conclusion flipped.

How this post relates to the previous four

In the first post, "Half of the big six's methodology is real knowledge, half is sales talk", I said methodology is sales talk with real knowledge underneath, and the real knowledge is "why people are moved".

What this post does is use my own account to test that real knowledge. What came out isn't complicated:

• People being moved and people being willing to publicly vouch for you are two different things. The former is a like, the latter is a repost. A hundredfold difference
• What makes people willing to vouch for you is asafe position, not a harsher conclusion
• Who gathers around you decides which metrics you should use. Peers mean credit; clients mean funnel
• Intuition will hand you a very sensible decision. Check first, then act

All of this is in the big six's methodology decks, just under mystical-sounding names. Now every line has its own numbers to check against, and the names are no longer needed.

I'll come back in a while to see whether that trigger condition has been hit.

How it was measured(for anyone who wants to do this themselves): post-level data comes from the official Threads API's insights(views / likes / replies / reposts / quotes / shares); the follower API only gives the current total, so history has to come from the Threads dashboard; website traffic uses the GA4 Data API, with source under l.threads.com; on the product side, sign-up source is a utm suffix we planted on the landing page ourselves. All four are read-only, and a single script can run them every day.

"Inside the 4A ad agencies" series

1. Half of the big six's methodology is real knowledge, half is sales talk
2. Clients ask for a Big Idea, but what they want is to not get burned
3. This system is being eaten by AI and consulting firms, and it deserves it
4. How Satsuma Creative does advertising
5. [This post · bonus] I wrote four posts on what really goes on inside ad agencies, put them on Threads, then looked at the data