What 3 Months of X Data Actually Says (and Doesn't) About Engagement
I’m running a faceless Twitter account to sell an ebook. I built a script with Claude Code to analyze X Analytics exports. Took about 30 minutes.
I ran it on my personal account first since it has more data (392 tweets from the last 3 months).
Findings
Correlations with impressions:
Original tweets do better than replies (-0.265 for replies)
$tickers help (+0.226)
Character count: weak positive, 150-200 chars best but noisy
Correlations with engagement rate:
t.co present (+0.427) - see caveat below
Character count: nothing
Day of week:
Thursday best (4.01% engagement)
Saturday worst (2.35%)
Questions: No effect (-0.02). Could be sample size.
What the Data Can’t Tell Me
The numbers need interpretation.
$tickers: are those impressions valuable?
Ticker tweets averaged 1,628 impressions vs 978 without — a 66% uplift. But how many of those are fundamental investors intentionally searching a ticker vs. ticker-monitoring bots and algo distribution? A tweet with $APO might get 1,600 impressions, but are those the eyeballs I actually want?
I don’t know yet.
“t.co present” is a messy proxy.
My script flags tweets containing t.co/ URLs, but that catches images, videos, quote tweets, and external links in one bucket. The +0.427 correlation with engagement is real, but I suspect images and videos are carrying it. External links to articles probably hurt. X Analytics doesn’t break these out in the export, so I can’t isolate the signal.
Questions don’t matter — waiting for more data.
The “end with a question to boost engagement” advice is everywhere. In my data: -0.02 correlation. Basically zero.
Could be my audience. Could be sample size. Could be the questions I ask. Could be that questions only work in certain formats. Flagging it but not acting on it yet.
The Faceless Account
With my personal account, I post what I want. There is no single goal.
The faceless account is trying to sell an ebook. We’re at 7 followers. We reply to podcast hosts with notes on their episodes. Do we get more replies when we ask the host a question? Or does that look spammy? We’re going to track it.
What I’m Actually Changing
Personal account: Probably a few more pictures and quote tweets. That’s about it.
Faceless account: We’re going to track it.
If you want to chat or learn more about using social media, whether you’re just getting started or more advanced, book a free 1:1 office-hours session with me here.
If you want to keep following along with this newsletter: