Which of Your Substack Notes Actually Worked?

I have posted 933 notes. My total reaction count across all of them is 799, which works out to roughly 0.9 per note. My best single note got 16 reactions, 3 restacks and 5 comments.

On a good day that feels like a win. On a bad day the whole feed feels like shouting into a corridor.

The problem was never the numbers. It was how I was reading them.

One note is a coin flip, not a data point

For a long time I would post a note, check the likes an hour later, feel vaguely good or vaguely bad, and move on.

I was not learning anything. I was reacting to noise.

Your note might do well because you caught the right moment. It might flop because a big account posted at the same time and swallowed the attention. You cannot tell from one note, ever.

The only way to learn something real is to group your notes and look at averages across the groups.

My best category is the one I trust least

Here is what my 933 notes looked like broken down by topic:

• Culture: 3.1 average, across only 12 notes
• Philosophy: 1.7 average, across 19 notes
• Education: 1.7 average, across 80 notes
• Technology: 1.3 average, across 173 notes

At a glance that looks obvious. Culture is my best category by a mile. Write more Culture.

Wrong, and this is the most important thing in the article.

Culture has twelve notes. Twelve. That 3.1 could easily be one or two notes that caught a wave. If a single note gets 10 reactions and the other eleven get 1 each, the average jumps to 1.8 on the strength of that one note alone.

Now look at Technology. It averages 1.3, which reads as worse. But it is spread across 173 notes, and one lucky note cannot drag that number around.

The 1.3 is real. The 3.1 probably is not.

My rule now: if a group has fewer than about 30 notes, I note the number and do not act on it.

How to group your own

Substack shows you a note at a time. It does not group them, so this part is on you.

The spreadsheet version. Columns for the date, the time, the weekday, the rough topic, the format, and then reactions, restacks and comments separately. Then work through it in this order:

1. By format. Short and punchy, longer and thoughtful, quotes, questions, links with commentary. Does any format consistently pull higher across at least 30 examples?
2. By hour and day. Morning against evening, weekday against weekend. Same rule: enough notes in each bucket to mean anything.
3. By topic. As I did above, checking the count every single time.

The catch is that a spreadsheet only holds what you remember to put in it, and the useful version of this needs months of rows before it says anything.

The other version is to have something record it as you post. That is what StackBuddy does: it keeps every note with its timing and its engagement, and shows exactly these breakdowns with the note count printed beside every average, so a number built on twelve notes cannot quietly pass itself off as a finding. Every account starts on Pro for 14 days with no credit card, and analytics is part of Pro, so you can look at your own history before deciding whether it is worth paying for.

Whichever you pick: when a gap shows up across more than one grouping, that is worth acting on. One viral note is luck. A consistent gap across dozens is a lesson.

Reactions are the weakest signal

One more trap, and it is the one I fell into.

Reactions are the easiest thing to measure, so they become the thing you optimise for. They are also the weakest signal you have.

A like takes one tap and means almost nothing. A restack means someone thought it was worth sending to their own subscribers. A comment means someone stopped scrolling entirely and typed.

Out of my 933 notes I have 83 restacks. That number is small, and it tells me something: notes that get restacked are doing something different from notes that get liked. Optimising for reactions alone would hide that completely.

So track comments and restacks in their own columns. They often tell a different story than the like count.

What to actually do with this

The short version:

• You need volume before you can learn anything.
• You need grouping before individual numbers mean anything.
• You need the sample size before you believe an average.
• You need to weight restacks and comments above likes if you care about reach.

Build the grouping habit before you worry about tooling. That habit is what turns your notes feed from a slot machine into something you can learn from, and it is worth having whether you keep it in a spreadsheet or let something keep it for you.