What Is Cohort Analysis?
Cohort analysis groups customers by time.
Cohort analysis is a method of segmenting customers into groups (cohorts) based on a shared characteristic — usually the date of their first purchase — and tracking their behavior over subsequent time periods. Google's analytics documentation defines it as the study of user behavior over time for groups who share a common trait. Unlike aggregate metrics, cohort analysis reveals whether your business is actually improving at retaining and monetizing customers — or just growing through new acquisition.
A cohort is a group of customers who share a defining event within the same time window. The most common cohort type in ecommerce is the acquisition cohort: all customers whose first purchase fell within a specific week or month. You then track what percentage of each cohort returns to purchase in month 2, month 3, month 6, and beyond.
The reason cohort analysis matters more than aggregate retention rates is survivorship bias. If your store-wide repeat purchase rate is 28%, that number blends together customers acquired last month (who have barely had time to repurchase) with customers acquired two years ago (who have had ample opportunity). The aggregate number tells you nothing about whether your retention is improving or deteriorating. Cohort analysis separates those groups so you can compare them on equal terms.
Consider two scenarios. In Scenario A, your January cohort has a 22% month-3 retention rate, your February cohort has 24%, and your March cohort has 27%. Retention is improving. In Scenario B, those numbers reverse: 27%, 24%, 22%. Retention is eroding. Both scenarios might produce the same aggregate retention figure in a given quarter. Only cohort analysis makes the trend visible.
Why Does Cohort Analysis Matter for Ecommerce Brands?
Cohort analysis is the only reliable method for measuring whether changes to your product, onboarding, or marketing are actually improving customer retention over time. According to Bain & Company research, a 5% increase in retention produces a 25-95% increase in profit. Without cohort-level tracking, you cannot attribute retention improvements to specific initiatives or time periods.
Three problems make cohort analysis non-negotiable for ecommerce:
1. Acquisition volume masks retention decay. A brand acquiring 5,000 new customers per month can show growing total revenue even while retention rates collapse. Cohort analysis isolates retention from acquisition, exposing the real health of each customer group.
2. Marketing changes need isolated measurement. You launched a post-purchase email sequence in March. Did it work? Comparing the March cohort's 90-day retention to the February cohort's 90-day retention gives you a direct answer — something no aggregate metric can provide.
3. Channel quality varies dramatically. Customers acquired through Facebook ads during a 40%-off promotion behave differently than customers acquired through organic search at full price. Cohort analysis by acquisition channel reveals which channels produce durable customers and which produce one-time buyers. This directly informs how you allocate your ad budget and set ROAS targets.
Without cohort analysis, you are making retention decisions based on averages that obscure every meaningful pattern. Your ecommerce KPIs become more actionable the moment you start viewing them through a cohort lens.
How Do You Build a Cohort Retention Table?
A cohort retention table is a matrix where rows represent customer groups (typically by month of first purchase) and columns represent time periods since that first purchase. Each cell shows the percentage of the original cohort that made a purchase in that period. Building one requires only three data points per customer: customer ID, first purchase date, and subsequent purchase dates.
Here is an example retention cohort table for a DTC skincare brand:
| Cohort (Month) | Customers | Month 1 | Month 2 | Month 3 | Month 4 | Month 5 | Month 6 |
|---|
| January | 1,200 | 100% | 32% | 24% | 19% | 16% | 14% |
| February | 1,350 | 100% | 34% | 26% | 21% | 18% | 15% |
| March | 1,500 | 100% | 36% | 28% | 23% | 19% | — |
| April | 1,100 | 100% | 30% | 22% | 17% | — | — |
| May | 1,450 | 100% | 38% | 29% | — | — | — |
| June | 1,600 | 100% | 35% | — | — | — | — |
Reading this table: of the 1,200 customers who made their first purchase in January, 32% returned for a second purchase in month 2, 24% purchased in month 3, and so on. The staircase pattern of missing data in the bottom-right is inherent to cohort tables — newer cohorts have not had enough time to reach later periods.
What the table reveals
The March and May cohorts show stronger early retention (36% and 38% in month 2) compared to January and April (32% and 30%). If you launched a new post-purchase email sequence in March, this is evidence it worked. If April's numbers dipped, check what changed — did you run a deep discount campaign that attracted lower-quality buyers?
How to build one
Step 1. Export your order data with customer ID, order date, and order total.
Step 2. For each customer, identify their first order date. This is their cohort assignment.
Step 3. For each subsequent order, calculate the number of months since their first order. That determines which column the purchase falls into.
Step 4. For each cohort row, count unique customers who purchased in each subsequent month. Divide by the total cohort size to get the retention percentage.
A spreadsheet handles this for stores with under 10,000 customers. Beyond that, use your analytics platform or a SQL query against your order database.
What Is a Revenue Cohort Table and How Does It Differ?
A revenue cohort table tracks cumulative revenue per cohort rather than retention percentage. This reveals how much each cohort is worth over time — and is the foundation for calculating customer lifetime value at the cohort level. Revenue cohorts expose whether your high-retention cohorts are also high-spending cohorts, which is not always the case.
Retention tells you whether customers come back. Revenue tells you how much they spend when they do. The two do not always move together. A cohort might retain well but downtrade to cheaper products over time, reducing per-customer revenue despite strong repeat rates.
Here is a cumulative revenue cohort table for the same skincare brand:
| Cohort (Month) | Customers | Month 1 | Month 2 | Month 3 | Month 4 | Month 5 | Month 6 |
|---|
| January | 1,200 | $78,000 | $106,200 | $124,800 | $138,600 | $149,400 | $157,800 |
| February | 1,350 | $91,100 | $125,700 | $149,000 | $168,200 | $183,100 | $194,400 |
| March | 1,500 | $105,000 | $147,600 | $177,000 | $202,800 | $221,400 | — |
| April | 1,100 | $68,200 | $90,600 | $105,400 | $117,200 | — | — |
| May | 1,450 | $101,500 | $144,700 | $175,300 | — | — | — |
| June | 1,600 | $115,200 | $158,400 | — | — | — | — |
To calculate cohort-level LTV, divide cumulative revenue by cohort size. The January cohort's 6-month LTV is $157,800 / 1,200 = $131.50 per customer. The March cohort is on pace to exceed that: $202,800 / 1,500 = $135.20 per customer at month 4, which already surpasses January's month-4 figure of $138,600 / 1,200 = $115.50.
This is the data that informs whether you should scale the acquisition strategies used in March and dial back whatever drove April's weaker cohort.
How Do You Run Cohort Analysis in GA4?
Google Analytics 4 has a built-in cohort exploration report under the Explore tab. It supports retention, engagement, and transaction-based cohorts with configurable time granularity. For ecommerce cohort analysis specifically, you need enhanced ecommerce tracking configured to capture purchase events with revenue data.
GA4's cohort exploration is found under Explore > Cohort exploration. Here is how to set it up:
Step 1: Select the cohort inclusion criteria. For ecommerce, choose "first_visit" or a custom event like "purchase" as the inclusion condition. Using "purchase" as the inclusion event creates acquisition-based purchase cohorts.
Step 2: Set the return criteria. This defines what counts as a "return" in subsequent periods. For retention analysis, use the "purchase" event. For engagement analysis, you might use "session_start."
Step 3: Choose granularity. Daily cohorts are useful for flash sale analysis. Weekly cohorts work for subscription products with short replenishment cycles. Monthly cohorts are the standard for most ecommerce retention analysis.
Step 4: Set the date range. Use at least 6 months to see meaningful retention trends. Twelve months is better for products with longer repurchase cycles.
Step 5: Add breakdowns. Segment by traffic source, device category, or user campaign to compare cohort performance across acquisition channels.
GA4's built-in cohort report has limitations. It caps at 60 cohorts and does not support revenue-based cohort tables natively. For revenue cohorts, export event-level data to BigQuery (GA4's free BigQuery export) and build the cohort table using SQL. Alternatively, platforms like Mixpanel, Amplitude, or Shopify's built-in cohort reports provide more flexible cohort analysis for ecommerce.
Want to see which signals drive your best-performing cohorts? ConversionStudio scans your brand's market signals and surfaces the patterns that matter — so your acquisition attracts customers who stick around. Try ConversionStudio free. 3 minutes. No pitch.
What Retention Benchmarks Should You Target by Cohort?
Ecommerce retention benchmarks vary significantly by product category and business model. Subscription brands retain 35-50% of customers at month 3, while non-subscription DTC brands average 18-28%. These benchmarks from aggregated Shopify and industry data provide context — but the trend across your own cohorts matters more than any single number.
Absolute retention numbers mean less than the direction of your cohort curves. A 22% month-3 retention rate is healthy for a furniture brand but concerning for a consumable supplement brand. The table below provides category-level context.
| Category | Month 2 Retention | Month 3 Retention | Month 6 Retention | Month 12 Retention |
|---|
| Subscription (supplements, coffee) | 55-65% | 40-50% | 30-38% | 22-30% |
| Consumable (skincare, food) | 30-40% | 22-32% | 15-22% | 10-18% |
| Apparel / Fashion | 20-30% | 15-22% | 10-16% | 7-12% |
| Home & Garden | 12-20% | 8-14% | 5-10% | 3-7% |
| Electronics | 8-15% | 5-10% | 3-7% | 2-5% |
The most important benchmark is not an industry number — it is the month-2 retention rate of your best-performing cohort versus your worst. If your best cohort retains at 38% in month 2 and your worst retains at 22%, the gap represents the improvement ceiling for your current product and experience. Your goal is to make every future cohort look more like your best one.
Use customer segmentation to drill into what separates high-retention cohorts from low-retention ones. Common differentiators: acquisition channel, first product purchased, whether a discount was used, and whether the customer engaged with post-purchase email within the first 7 days.
What Are the Most Common Cohort Analysis Mistakes?
Three errors undermine most ecommerce cohort analyses: using time windows that are too short, ignoring cohort size differences, and conflating retention with engagement. Each produces misleading conclusions that can send acquisition and retention strategies in the wrong direction.
Mistake 1: Measurement window too short
A 30-day cohort analysis tells you almost nothing about retention. Many ecommerce products have natural repurchase cycles of 45-90 days. If you measure retention at 30 days, your cohort table will show 5-10% retention and you will conclude your retention is terrible — when in reality, customers simply have not had time to repurchase. Match your measurement window to your product's natural repurchase cycle. Consumables need at least 90-day windows. Apparel needs 6 months. Furniture needs 12-18 months.
Mistake 2: Ignoring cohort size
A cohort of 50 customers with 40% month-2 retention is not statistically meaningful. Small cohorts produce noisy data that looks like signal. Before drawing conclusions from any cohort's performance, check the absolute numbers. A 2% difference in retention between a 2,000-customer cohort and a 200-customer cohort is likely noise, not insight.
Mistake 3: Conflating retention with engagement
A customer who visits your site in month 3 but does not purchase is engaged but not retained (in the revenue sense). Be precise about what your cohort table measures. If it tracks site visits, call it an engagement cohort. If it tracks purchases, call it a retention cohort. Mixing the two leads to inflated retention figures that do not translate to revenue.
Mistake 4: Ignoring seasonality
A cohort acquired in November (Black Friday) will always look different from a cohort acquired in February. Before concluding that your November cohort has lower retention, account for the fact that discount-driven buyers systematically retain at lower rates. Compare November-to-November year-over-year rather than November-to-February within the same year.
How Do You Use Cohort Data to Improve Retention?
Cohort analysis is diagnostic, not prescriptive. It tells you where retention is strong or weak, but you must investigate why. The three highest-leverage interventions informed by cohort data are post-purchase experience optimization, first-product-purchased analysis, and channel-specific retention tracking.
Optimize the post-purchase experience
Your month-2 retention rate is the most actionable number in your cohort table. It represents the percentage of first-time buyers who come back. If month-2 retention is below your category benchmark, the post-purchase experience is the first place to look. Are you sending a post-purchase email sequence? Does it educate, or just sell? Is the delivery experience meeting expectations set during purchase?
Analyze first product purchased
Segment your cohorts by the first product each customer bought. You will almost certainly find that some products produce dramatically higher retention than others. A skincare brand might discover that customers who start with a cleanser retain at 35% in month 3, while customers who start with a single lipstick retain at 12%. This insight should reshape your acquisition creative — lead with the products that produce long-term customers, even if they have lower margins on the first transaction.
Track retention by acquisition channel
Build separate cohort tables for each major acquisition channel: Facebook, Google, TikTok, organic, email, and referral. This reveals which channels produce customers who stick around and which produce one-time buyers. A channel with a $45 CAC but 35% month-3 retention may deliver better customer lifetime value than a channel with a $25 CAC but 15% month-3 retention.
Frequently Asked Questions
What is the difference between cohort analysis and segmentation?
Customer segmentation divides your customer base by shared characteristics at a single point in time — demographics, purchase behavior, value tier. Cohort analysis tracks a specific group over time to observe how their behavior evolves. Segmentation answers "who are my customers?" Cohort analysis answers "how do they behave over time?" The two are complementary: segment your cohorts to understand which customer types retain best.
How much data do I need to start cohort analysis?
You need a minimum of 3 months of order history and at least 200 customers per cohort to produce statistically meaningful results. If you have fewer than 200 customers per monthly cohort, use quarterly cohorts instead. The quality of insight improves significantly at 6-12 months of data and 500+ customers per cohort.
Can I do cohort analysis without GA4?
Yes. Any tool that captures customer ID and purchase date can support cohort analysis. Shopify has a built-in customer cohort report. Klaviyo, Mixpanel, and Amplitude all offer cohort analysis features. You can also build cohort tables manually in a spreadsheet using exported order data — the process described in this article works with any data source.
What is the difference between a time-based cohort and a behavior-based cohort?
Time-based cohorts group customers by when they took an action (first purchase in January, first purchase in February). Behavior-based cohorts group customers by what they did (customers who purchased product A, customers acquired through Facebook, customers who used a discount code). Both are valuable. Time-based cohorts reveal trends over time. Behavior-based cohorts reveal which actions or channels predict long-term retention.
How often should I review cohort data?
Monthly review is sufficient for most ecommerce brands. Update your cohort table at the start of each month, adding the newest cohort and extending the columns for existing cohorts. If you are actively running retention experiments (new email sequences, loyalty programs, onboarding changes), review weekly for the first 90 days of the experiment to catch early signals.
Keep Reading
