Cohort Analysis for Ad Campaigns: Tracking Long-Term Customer Value

I'm going to let you in on something that changed how I think about ad campaigns completely.

Three years ago, I was running ads for an e-commerce brand. We had two Facebook campaigns running side by side—let's call them Campaign A and Campaign B.

Campaign A: CPA of ₹450, first-purchase ROAS of 3.2x. Everyone loved it. Campaign B: CPA of ₹680, first-purchase ROAS of 2.1x. Everyone wanted to kill it.

Guess which campaign was actually making us more money?

Campaign B. By a lot.

How did I figure this out? Cohort analysis. And once I started tracking customers by acquisition cohort instead of just looking at first-purchase metrics, everything changed.

Let me show you how this works.

What Actually Is Cohort Analysis?

Okay, forget the textbook definition. Here's what cohort analysis means in practice:

Instead of looking at "all customers" or "this month's revenue," you group customers by when they were acquired, then track how each group behaves over time.

Example:

• Customers acquired in January = January cohort
• Customers acquired from Facebook ads = Facebook cohort
• Customers acquired from that specific Valentine's Day campaign = V-Day cohort

Then you watch what happens to each cohort over the next 3 months, 6 months, 12 months.

Do they come back? Do they buy more? Do they refer friends? Or do they buy once and disappear?

This is how you figure out which campaigns are bringing you customers who actually stick around—and which ones are bringing you one-and-done bargain hunters.

Why First-Purchase ROAS Is Lying to You

Here's the thing everyone gets wrong about ROAS.

When you calculate ROAS on first purchase only, you're measuring a campaign's ability to generate immediate revenue. That's it.

You're not measuring:

• Whether those customers come back
• Whether they have higher lifetime value
• Whether they're more profitable after accounting for returns and support costs
• Whether they refer other customers

I've seen this play out dozens of times. A campaign looks amazing on day 30, then falls apart when you track it to day 180.

Here's a real comparison from one of my accounts:

See what happened? The discount campaign looked like the winner at 30 days. By 180 days, it was the worst performer.

The brand storytelling campaign looked like a disaster initially. Turned out it was acquiring the highest-quality customers we'd ever seen.

If I'd killed that campaign at day 30 (which everyone wanted to do), we would've lost our best acquisition channel.

How to Actually Set Up Cohort Tracking

Alright, let's get practical. Here's how I set this up for every client.

Step 1: Define Your Cohorts

You need to decide how you're grouping customers. I typically track by:

Acquisition month – Broad view of how customer quality changes over time Acquisition channel – Which channels bring better long-term customers Acquisition campaign – Specific campaign comparison Promotion type – Discount vs full-price customers

The key is making sure every customer gets tagged with their cohort info at the point of acquisition. This data has to live in your system permanently.

Step 2: Choose Your Tracking Window

How long are you going to follow each cohort?

For e-commerce, I typically use:

• 30/60/90 days – Standard tracking
• 180 days – Better picture of repeat behavior
• 12 months – Full annual cycle including seasonal behavior

For subscription businesses, I go longer—often 18-24 months to see full churn patterns.

Step 3: Decide What Metrics Matter

Don't just track revenue. That's incomplete. Here's what I actually monitor:

Purchase behavior:

• Repeat purchase rate
• Average order value (1st vs 2nd vs 3rd purchase)
• Purchase frequency
• Revenue per customer

Profitability:

• Contribution margin per customer
• Return rate
• Support ticket rate
• Payment failure rate

Engagement:

• Email open rates by cohort
• SMS engagement by cohort
• Referral rate

You want to build a complete picture of what makes a "good" customer for your specific business.

Step 4: Build Your Cohort Dashboard

This is where most people give up because it seems complicated.

It's not. You need three views:

1. Cohort retention table – Shows what percentage of each cohort is still active over time 2. Revenue by cohort over time – Tracks cumulative revenue for each cohort as they age 3. Cohort comparison table – Lets you compare key metrics across different cohorts side by side

I use AdsMAA for this because it has cohort analysis built in and automatically pulls data from ad platforms plus your CRM. The alternative is building this yourself in Google Sheets or a BI tool, which honestly takes forever.

What the Data Actually Tells You

Okay, you've got your cohort dashboard set up. Now what?

Here's how I read the data and what I look for.

Red Flag #1: Steep Drop-Off After First Purchase

If 85% of a cohort never buys again, that campaign is bringing you one-time customers. Could be:

• You're over-promising in ads and under-delivering on product
• You're attracting deal-seekers with heavy discounts
• Your product isn't actually solving the problem you're advertising
• Post-purchase experience is terrible

What I do: Audit the full customer journey for that campaign. Look at the ad creative, the landing page, the product, the follow-up emails. Find the disconnect.

Red Flag #2: High Return/Refund Rates

If a cohort has 2x the return rate of others, something's wrong with acquisition quality.

I had a client where one campaign had a 34% return rate while others averaged 12%. Turned out the ad creative was showing the product in unrealistic lighting that made colors look different. Customers got the actual product and immediately returned it.

What I do: Segment by campaign, look at return reasons, fix the creative or targeting issue.

Green Flag #1: Improving Metrics Over Time

If repeat purchase rate grows from 15% at day 30 to 35% at day 90, you've found a cohort that's engaged and discovering more value.

These are your best customers. Figure out what makes them different and acquire more like them.

Green Flag #2: Higher AOV on Subsequent Purchases

If first purchase is ₹800 but second purchase is ₹1,400, those customers are becoming more valuable over time.

This usually means:

• They trust you enough to buy higher-ticket items
• They're discovering your full product line
• They're moving from "testing" to "committed"

What I do: Scale the campaign, and create lookalike audiences based on this cohort.

Real Example: How Cohort Analysis Changed Our Entire Strategy

Let me walk you through a real situation where cohort analysis completely flipped our strategy.

Client was a skincare brand. We were running four main campaigns:

Discount campaign – 30% off first order

Before/after social proof – Real customer results

Ingredient education – Science behind the formulation

Influencer partnerships – Collab with beauty creators

Everyone was pushing hard on the discount campaign because it had the lowest CPA and highest first-purchase ROAS.

I ran a 6-month cohort analysis and found something wild:

"The discount campaign had a 9% repeat purchase rate. The ingredient education campaign had a 42% repeat purchase rate. And customers from the education campaign spent 2.3x more over 6 months despite having a higher initial CPA."

Here's what the full picture looked like:

The education campaign looked like the worst performer on a 7-day window. It was actually the best by a mile.

What we changed:

• Cut discount campaign budget by 60%
• Tripled education campaign budget
• Created more educational content for social proof and influencer campaigns
• Stopped using discount as primary hook

Results:

• Overall CAC went up 18% (everyone panicked)
• 6-month LTV went up 73% (everyone stopped panicking)
• Net profit up 94% despite spending the same total amount

This is what happens when you optimize for the right metric.

The Cohort Metrics That Actually Matter

Don't track everything. You'll drown in data. Here are the metrics I actually care about:

Customer Lifetime Value (LTV)

Total revenue per customer over your tracking window. Broken down by cohort.

This is your north star. If LTV is going up, you're acquiring better customers.

Repeat Purchase Rate

Percentage of cohort that makes a second purchase. Then third. Then fourth.

I track this at 30, 60, 90, and 180 days. Shows you engagement trajectory.

Time to Second Purchase

How long between first and second purchase?

Shorter is usually better—means they're engaged and ready to buy again. But not always. Some products have natural repurchase cycles.

Cohort Contribution Margin

Total profit per customer after all costs, by cohort.

This accounts for returns, support costs, fulfillment—everything. Most honest metric you've got.

Month 3 Retention Rate

What percentage of the cohort is still active (made a purchase) in month 3?

This is my quick litmus test. If month 3 retention is below 20%, something's wrong with acquisition quality.

How to Use This to Make Better Campaign Decisions

Alright, you've got cohort data. Here's how I actually use it.

Decision 1: Budget Allocation

Simple rule: move budget toward campaigns with highest 90-day or 180-day LTV, even if their initial ROAS is lower.

You're optimizing for long-term value, not short-term revenue.

Decision 2: Creative Testing

Look at which creative themes produce cohorts with better retention and LTV. Double down on those themes across all campaigns.

I had a fitness brand where "transformation journey" creative consistently produced cohorts with 2x better retention than "quick results" creative, even though quick results got more clicks.

Decision 3: Audience Targeting

Create lookalike audiences based on your highest-LTV cohorts, not your largest cohorts.

Facebook/Google will optimize for conversion. You need to point them toward high-quality conversions.

Decision 4: Pricing and Promotion Strategy

If discount-acquired cohorts have terrible LTV, stop leading with discounts. Test value props that attract quality customers instead.

If free-shipping cohorts perform just as well as full-price cohorts, you can run that promo without hurting customer quality.

Common Mistakes I See All the Time

Mistake 1: Not waiting long enough

You can't evaluate a cohort after 14 days. Give it at least 90 days, ideally 180. Yes, that means slower decision-making. That's the point.

Mistake 2: Comparing cohorts from different seasons

Don't compare January cohorts to December cohorts. Seasonal behavior will skew everything. Compare year-over-year or against rolling averages.

Mistake 3: Not segmenting by offer type

A "Facebook campaign" cohort isn't specific enough. You need to separate discount customers from full-price customers from free-trial customers. They behave completely differently.

Mistake 4: Ignoring small cohorts

Sometimes your best-performing cohort is small. That's fine. Figure out why they're good and how to acquire more like them. Don't dismiss it because the sample size is low.

Your Implementation Checklist

Don't overthink this. Here's exactly what to do:

Week 1: Set up cohort tagging

• Tag every new customer with acquisition date, channel, campaign, and offer type
• Make sure this data flows into your CRM or data warehouse

Week 2: Build your first cohort report

• Start with simple stuff: repeat purchase rate and revenue per customer
• Track at 30, 60, 90 days
• Compare at least 3 different cohorts (channel or campaign level)

Week 3: Add profitability metrics

• Layer in return rates, support costs, contribution margin
• This is where the real insights come

Week 4: Make your first decision

• Find your highest LTV cohort
• Find your lowest LTV cohort
• Move 15% of budget from worst to best

Then repeat monthly. Your cohort data will get richer and your decisions will get better.

If you need a tool that handles this automatically, check out AdsMAA's cohort analysis features. Saves you weeks of setup and connects everything in one dashboard.

FAQ

What if my product has a long purchase cycle?

Extend your tracking window. If customers naturally buy every 6 months, you need to track cohorts for 12-18 months minimum to see true patterns.

How do I handle customers acquired from multiple channels?

Use multi-touch attribution. Assign partial credit to each touchpoint, or use a model like time-decay that weights later touches more heavily. AdsMAA handles this automatically.

Should I optimize campaigns based on cohort data even if it hurts short-term numbers?

Yes. If you're building a real business, not just chasing this month's numbers, optimize for LTV. Your short-term metrics will catch up once you're acquiring better customers.

How many cohorts do I need before this data is reliable?

Minimum 3-4 cohorts with at least 100 customers each. Fewer than that and you're dealing with too much noise. More is better—I typically want 6+ months of cohort data before making major strategic changes.

Look, here's the bottom line.

If you're still optimizing campaigns based on first-purchase ROAS, you're flying blind. You have no idea which campaigns are actually building long-term value and which ones are just generating one-time transactions.

Cohort analysis tells you the truth. It shows you which customers stick around, which ones become advocates, and which ones cost more to acquire than they'll ever return.

It takes longer to see results. You need patience. But once you start making decisions based on actual customer lifetime value instead of day-7 ROAS, everything changes.

Your campaigns get better. Your customers get better. Your business gets better.

Stop optimizing for first purchase. Start optimizing for lifetime value.

That's how you actually build something that lasts.