Your reports can be 100% accurate and still lead you to the wrong decision.
That’s the uncomfortable truth waiting on the other side of Parts 1 and 2. You can know every metric. You can have flawless tracking. And you can still misread what’s happening in your business — because knowing the numbers and understanding the numbers are two different skills.
LTV — Customer Lifetime Value
LTV is the total revenue a customer generates over their entire relationship with your brand — not just their first purchase.
LTV = AOV × Purchase Frequency × Customer Lifespan
A customer with a $60 average order who buys four times a year for two years has an LTV of $480.
In Part 1, you learned CAC — what it costs to acquire a customer. But CAC on its own is meaningless. A $70 CAC is reckless if customers only ever spend $50. It’s a bargain if they spend $500.
The number that actually matters is the relationship between the two. Most healthy e-commerce brands aim for an LTV:CAC ratio of at least 3:1 — meaning every customer returns three times what you paid to get them.
LTV is also what gives you permission to spend. If you know a customer is worth $400 and your competitor thinks they’re worth $80, you can afford to outbid them on every impression and still profit. That’s not a bigger budget. That’s better information.
CAC tells you what you paid. LTV tells you whether it was worth it.
Cohort Analysis
Cohort analysis means grouping customers by something they share — usually the month they were acquired — and then tracking that specific group’s behavior over time.
Instead of asking “what’s our repeat purchase rate?” you ask “what’s the repeat purchase rate of customers we acquired in January, versus February, versus March?”
The difference is significant. Blended averages hide everything. A brand can look completely stable at the top line while its newer customers quietly get worse.
Say your January cohort repurchases at 40% by month three. Your April cohort repurchases at 12%. Total revenue looks flat, so nothing seems wrong — but something changed. Maybe you leaned harder on discounts and attracted deal-seekers. Maybe a new channel is bringing in the wrong audience. Maybe a product quality issue started three months ago and you’re only now able to see it.
Cohorts are how you catch that. They turn a snapshot into a trend line.
Notice that January’s curve flattens around 28% rather than falling to zero. That plateau is your real repeat-customer base — and it’s the number worth protecting.
Averages tell you what happened. Cohorts tell you what’s changing.
Attribution Modeling
Attribution modeling is the set of rules that decides which marketing touchpoint gets credit for a conversion.
Say a customer sees your Instagram ad on Monday, clicks a search ad on Wednesday, and buys after an email on Friday. Three touchpoints, one sale. Who gets the credit?
The answer depends entirely on the model you’re using.
Same customer. Same sale. Four completely different stories about which channel is working.
This is the part that catches new marketers off guard: attribution is a choice. Every report you read was built on a model someone selected — and that selection quietly determines which channels look like heroes and which look like waste.
One caveat worth knowing, because it will save you from hunting for settings that no longer exist. In 2023, Google removed first-click, linear, position-based, and time-decay from both Google Ads and GA4, and migrated everyone still using them to data-driven attribution. Only last-click and data-driven remain, and data-driven is the default. You will still find the classic rules-based models in most third-party attribution platforms, and time-decay logic survives inside data-driven attribution itself, which weights recent touchpoints more heavily than older ones.
But notice what that shift actually means. You are no longer choosing the model. An algorithm is choosing for you, and you cannot open it up and inspect how it decided.
That makes the habit more important, not less. Before you kill a campaign for underperforming, find out what is producing the number you are about to act on. That single question will make you sharper than most people in the room.
Attribution doesn’t tell you the truth. It tells you a version of it.
Incrementality Testing
An incrementality test is a controlled experiment that measures whether your advertising actually caused a sale — or whether it would have happened anyway.
The setup is simple: one group of people sees your ads (the test group), another group is deliberately excluded (the control or holdout group). Everything else stays the same. The difference in conversions between the two groups is your true incremental lift.
This answers the single most important question in marketing, and it’s one no dashboard can answer for you: would this revenue have existed without the ad spend?
The classic example is brand search. You’re running ads on your own brand name and they show a phenomenal ROAS. But the people searching your brand name already know who you are — many of them were going to buy regardless. Are those conversions incremental, or are you paying for customers you’d have gotten for free?
You cannot know without testing it. Plenty of brands have run holdout tests on brand search only to discover a meaningful chunk of that performance was revenue they already owned.
Attribution tells you who to credit. Incrementality tells you what actually caused it.
Audience Suppression
Audience suppression is the deliberate practice of excluding certain people from seeing your ads.
Recent purchasers. Existing subscribers. Anyone who converted in the last 30 days. People already deep in your email flow. All of them can be suppressed from a prospecting campaign.
This sounds like giving up reach. It’s the opposite. Every dollar spent showing an ad to someone who already bought is a dollar not spent on someone who hasn’t. Suppression doesn’t shrink your budget — it concentrates it.
It’s also one of the fastest efficiency wins available to a new marketer. You don’t need a bigger budget, new creative, or a better offer. You just need to stop paying to reach people who don’t need convincing.
The strategic lesson underneath it is bigger than the tactic: targeting isn’t only about who you want. It’s about who you’re willing to leave out.
Good marketers decide who to reach. Great marketers decide who to skip.
Where This Leaves You
Three posts ago, the goal was reading a report without getting lost.
Part 1 gave you the money metrics — AOV, CAC, ROAS, MER, CPM, CRM. The language of performance. Part 2 gave you the infrastructure — cookies, pixels, APIs, CAPI, UTMs. The machinery that makes those numbers possible in the first place.
This post gave you something different: permission to question all of it.
Because the marketers who grow fastest in this industry aren’t the ones who memorize the most terms. They’re the ones who look at a 6x ROAS and ask which attribution model produced it. Who see a strong month and check whether the newest cohort is holding up. Who notice a campaign winning and wonder whether it’s creating demand or just taking credit for it.
That instinct — the reflex to ask why before accepting what — is the whole job.
You now have the vocabulary. The rest is curiosity.
