Lesson 6: Analytics for People Who Hate Marketing Math
you've been posting for a few weeks. you're getting some replies, some likes, some profile views. but there's a nagging question: is this actually working? or am i just shouting into the void with slightly better formatting?
most people answer this question with vibes. "i feel like i'm getting more engagement." "i think the last post did well." vibes are not a measurement system. vibes are how you convince yourself everything is fine while nothing is working.
The insight: you already instrument your code. instrument your distribution the same way. you don't need a 12-tab spreadsheet or a $300 analytics tool. you need 4-6 numbers tracked consistently over time. that's it.
Why Engineers Are Bad at Marketing Analytics (Despite Being Good at Analytics)
the irony is thick here. you measure CPU usage, p99 latency, error rates, and deployment frequency. you run A/B tests on button colors. but when it comes to your content, you post something, glance at the like count, and move on.
why this happens:
- marketing metrics feel fuzzy compared to system metrics (impressions vs. requests served — one feels soft, one feels hard)
- the feedback loop is slower (a post takes hours or days to play out, not milliseconds)
- you don't have a dashboard (your code has grafana, your content has nothing)
Reframe: every post is an experiment. every metric is a log line. your content strategy is a system under active development. you wouldn't deploy to production without observability. don't deploy content without it either.
The Only 5 Numbers That Matter
forget "engagement rate," "reach," "impressions per follower," and the other 40 metrics platforms throw at you. 90% of them are vanity numbers designed to keep you checking the app. you need 5:
1. reply conversion rate
what: number of people who reply to your content / number of impressions (or views) why: replies are the highest-signal engagement. they require thought, effort, and emotional investment. a like is a nod. a reply is a conversation. this metric tells you whether your content is provocative enough to make people speak.
reply rate = (replies per post) / (impressions per post) × 100
track this per post, then average it weekly. a healthy reply rate on most platforms is 0.5-2%. if you're below 0.3%, your content isn't sparking conversation — you're probably summarizing instead of provoking.
2. profile visit rate
what: profile visits / impressions why: people don't follow from a single post. they follow from a single post that made them check your profile. this metric is the bridge between "interesting post" and "interesting person." if your post gets impressions but no profile visits, the content is fine but your authority or framing isn't compelling enough to make people want more.
3. follow conversion rate
what: new follows / profile visits why: this is your profile's conversion rate. someone visited — did they stay? low follow rate means your bio, pinned post, or content history isn't closing the deal. high follow rate means your profile matches the expectation your post set.
4. reply-game hit rate
what: replies you wrote that got engagement (likes, author reply, nested replies) / total replies you wrote why: the reply game is your highest-volume activity. you need to know if you're getting better at it or just doing busywork. track this weekly. aim for 20-40% of your replies generating some form of engagement.
5. content velocity (the meta-metric)
what: number of posts, replies, and cross-platform repurposes per week why: this isn't an outcome metric — it's an input metric. but it's the most predictive number you'll track. consistency is the leading indicator of every other metric. if content velocity drops, everything else drops with a lag. treat it like your deployment frequency.
The minimum viable dashboard: these 5 numbers. updated once a week. takes 10 minutes. if you skip a week, you lose trend visibility. if you skip a month, you're back to vibes.
Building the Tracking System (10 Minutes, No Tools)
you don't need a SaaS product for this. here's the system:
weekly review, every sunday (10 min)
open a plain text file, a notion page, or a notes app. record:
# Week [N], [date range]
posts published: [N]
replies published: [N]
cross-posts: [N]
top post by replies: [post summary, reply count, reply rate]
top post by profile visits: [post summary, visit count]
top reply: [who you replied to, what engagement it got]
followers gained: [N]
profile visits: [N]
content velocity: [total posts + replies + cross-posts]
avg reply rate: [total replies received / total impressions across all posts]
what worked this week:
- [pattern 1]
- [pattern 2]
what flopped this week:
- [pattern 1]
- [pattern 2]
one change to test next week:
- [hypothesis]
This is not optional. the weekly review is the difference between "doing marketing" and "getting better at marketing." without it, you'll repeat the same mistakes indefinitely and never know why. with it, every week is a learning cycle.
the monthly deep-dive (20 min)
once a month, pull back and look at the 4 weekly reviews. ask:
- which post format is working best? look at your top 4 posts. what do they share? (format, topic, hook style, length, time of day)
- which reply style gets the most engagement? review your best-performing replies from the month. is it contrarian takes? personal stories? adding nuance? asking questions?
- is content velocity trending up, down, or flat? if it's trending down, that's an early warning — fix it before the outcome metrics follow.
- what's the 1 thing to double down on? not 3 things. not a pivot. one thing. the format or topic that consistently outperforms. do more of it.
The doubling-down rule: when you find something that works, you don't "diversify." you do more of it until the returns diminish. most creators spread themselves thin trying to cover every topic and format. the ones who grow fastest find one thing that works and hammer it relentlessly.
Leading vs. Lagging Indicators
this distinction will save you months of confusion:
| leading (predictive) | lagging (confirming) |
|---|---|
| content velocity (posts/week) | follower count |
| reply game hit rate | inbound DMs |
| profile visit rate | newsletter subscribers |
| reply rate per post | collaboration offers |
The trap: obsessing over lagging indicators while neglecting leading indicators. follower count is a lagging indicator — it tells you what already happened. content velocity and reply rate are leading indicators — they tell you what's about to happen. if your follower count is flat but your reply rate has been climbing for 3 weeks, you're about to grow. if your follower count is growing but your reply rate has been dropping, you're about to stall. bet on the leading indicators.
When to Kill Something vs. When to Be Patient
this is the hardest judgment call in distribution. here's a decision framework:
kill it when:
- you've published the same format 10+ times and reply rate is consistently below 0.2%
- you've been replying to a specific account's content for 2 weeks and never gotten a reply or engagement
- a platform shows zero growth (not slow growth — zero) after 60 days of consistent posting
- a format takes 2x the effort of your others but performs worse
keep going when:
- engagement is low but growing week over week (even if slowly)
- a format has worked for others in your niche but hasn't clicked for you yet (you may be executing poorly, not choosing poorly)
- you've only tried something 2-3 times (insufficient data)
- you personally enjoy the format even if it's not top performer (burnout prevention is a valid metric)
The commitment test: before you kill anything, ask: "have i done this consistently enough to have a real sample size?" most people kill things after 2 attempts. 2 attempts is noise. 10 attempts is a signal. 20 attempts is data you can act on.
What to Ignore Completely
platforms will show you dozens of numbers. here's what to aggressively ignore:
- total impressions (without engagement context). 10,000 impressions with 0 replies = ghosted. 1,000 impressions with 20 replies = signal. impressions alone are meaningless.
- follower count (in isolation). follower count without engagement rate is a vanity number. a 500-follower account with 10% engagement rate is worth more than a 5,000-follower account with 0.5%.
- individual post like counts. likes are low-signal noise. a post with 200 likes and 3 replies is a post people nodded at. a post with 40 likes and 15 replies is a post that started conversations. the second one builds your reputation.
- day-to-day fluctuations. metrics are noisy day to day. only weekly and monthly trends matter. if you check your numbers every 3 hours you will make bad decisions based on random variance.
Quick Reference: Healthy Benchmarks
these are rough. your niche may vary. but if you're far below these after 3 months, something is off:
| metric | healthy (month 3+) | warning | dead |
|---|---|---|---|
| reply rate per post | > 0.5% | 0.2-0.5% | < 0.2% |
| profile visit rate | > 2% of impressions | 1-2% | < 1% |
| follow conversion | > 5% of profile visits | 2-5% | < 2% |
| reply game hit rate | > 25% | 15-25% | < 15% |
| content velocity | > 15 posts+replies/week | 8-15 | < 8 |
Check Your Understanding
Practice: Build Your Dashboard
- Create a new document (plain text, notion, notes app — whatever you'll actually use weekly).
- Copy the weekly review template from this lesson.
- Fill it in for the current week (or last week if it's monday).
- Set a recurring calendar event for every sunday: "10 min — content review."
That's it. The system is built. Now use it every week. If you skip two weeks in a row, the system is dead and you need to either simplify it or address why you're avoiding it.
The review habit is harder than the posting habit. posting is action. reviewing is reflection. action feels productive. reflection feels like you're not doing anything. but reflection is where the learning lives. without it, you're just running on a treadmill — lots of motion, zero progress.
1. 1. Why do engineers, who are good at system analytics, often fail at marketing analytics?
2. 2. Which is a leading indicator (predicts future growth) vs. a lagging indicator (confirms past growth)?
3. 3. When should you kill a content format?
4. 4. What's the minimum viable weekly review?