Back to site
>>> article

Cohort Analysis for Ecommerce: Retention and LTV Guide

Discover how cohort analysis in ecommerce boosts customer retention and explains lifetime value, shaping your growth strategies effectively.

Decorative title card illustration

Cohort analysis groups your customers by the month they first purchased and tracks whether they come back, how often, and how much they spend over time. Run a first-purchase month cohort right now, check your Month 2 repeat rate, and you will immediately see whether your acquisition spend is buying loyal customers or one-time buyers. That single number shapes every retention and LTV-to-CAC decision you make.

Key Takeaways

Cohort analysis is the most direct method for separating true retention improvement from the noise of changing acquisition volume, and the Month 1 to Month 2 repeat rate is the single number that tells you whether your acquisition spend is building a durable business.

Point Details
Start with acquisition-date cohorts Group customers by first-purchase month and track repeat purchases; this is the baseline before any channel or product segmentation.
Retention rate formula Divide repeat buyers in Month N by cohort starting size and multiply by 100; the Month 1-to-Month 2 drop is your most urgent signal.
Join ad spend to cohort data Without CAC joined to cohort rows, you cannot calculate payback period; this is the gap most Shopify-native reports leave open.
Run on a monthly cadence Full cohort matrix monthly, channel and product deep-dives quarterly; weekly smoke-checks catch anomalies before they compound.
Commerce Catalyst for complex data For brands with fragmented data or multi-channel ad spend, the DTC Financial Health Assessment connects cohort revenue to margin and CAC for clear payback decisions.

Table of Contents

What is cohort analysis in ecommerce, and which cohort types matter most?

In ecommerce, the cohort event is always the first purchase. The cohort clock starts ticking from that order date, and every subsequent purchase by that customer is measured as time elapsed since that first transaction. This framing is deliberate: it isolates true retention behavior from the noise of new customer volume, which aggregate metrics like monthly revenue or average order value can never do cleanly.

Shopify’s cohort retention analysis guide describes this as grouping customers by first-purchase date and tracking their revenue and repeat behavior over time, which is the baseline structure every ecommerce team should build before anything else.

The five cohort types worth building, roughly in order of priority:

  • Acquisition date (first-purchase month): The foundational cohort. Groups every customer who placed their first order in a given calendar month. Use this as your baseline for all retention benchmarking.
  • Acquisition channel or campaign: Segments cohorts by the traffic source that drove the first purchase (paid social, paid search, organic, email, influencer). Use this when you need to evaluate whether a channel is buying repeat customers or one-time traffic.
  • First product bought (gateway product): Groups customers by the SKU or category of their first order. Use this to identify which products act as entry points that lead to long-term purchasing versus dead ends.
  • Promotional vs. full-price cohort: Separates customers who first purchased on a discount from those who paid full price. These two groups behave very differently in subsequent months, and mixing them produces misleading averages.
  • Behavioral cohorts (loyalty signup, subscription conversion): Groups customers by a post-purchase action, such as joining a loyalty program or converting to a subscription. Use these to measure whether a specific program is actually changing retention curves.

The acquisition-date cohort is the baseline because it requires no additional data joins, works in every analytics tool, and gives you a clean time-series view of retention decay across every customer group you have ever acquired.

Why cohort analysis reveals what aggregate metrics hide

Monthly revenue is a blended number. It mixes the behavior of customers acquired six months ago with customers who ordered for the first time last week, and the two groups behave nothing alike. When a store grows quickly, new-customer volume can mask the fact that older cohorts are churning faster than they used to. Cohort curves expose that pattern immediately.

Hands packaging ecommerce orders

The business decisions that cohort data should directly inform:

Channel budget allocation. A paid-media channel might show a strong day-0 ROAS, but if Month 3 repeat rate for that cohort is half what organic search delivers, the true payback period is far longer than the platform dashboard suggests. Polar Analytics warns explicitly about this omnichannel-CAC trap: misjoined or missing ad spend data makes poor-quality acquisition channels look profitable in blended metrics.

Promotion design. Customers who first purchase on a steep discount tend to anchor to that price point. Tracking a promotional cohort separately from a full-price cohort shows whether a sale event is building a loyal base or filling the funnel with buyers who will only return when there is another discount. Given how widespread coupon use is in the United States, per Statista’s coupon-use data, this distinction carries real financial weight for most stores.

Win-back timing. Cohort data shows you the exact month when repeat purchase probability drops most sharply. That cliff is your win-back trigger point, not an arbitrary 30-day email rule.

Payback period forecasting. If you know the cumulative revenue per customer at Month 3, 6, and 12 for each acquisition cohort, you can calculate when a given channel’s CAC is recovered, which is the number that determines whether you can afford to scale that channel.

Core cohort metrics and the exact formulas you need

KISSmetrics recommends tracking repeat purchase rate at 30, 60, 90, 180, and 365-day intervals to compare cohort performance over time. That cadence maps cleanly to the core metrics below.

The metrics every ecommerce cohort analysis should track:

  • Retention rate: The percentage of customers from a cohort who made at least one additional purchase in a given period.
  • Repeat purchase rate: The share of customers who purchased more than once within a defined window (30/60/90/180/365 days).
  • Cumulative revenue per customer: Total revenue generated by a cohort divided by the number of customers in that cohort, accumulated month by month.
  • Average order value (AOV) by cohort: Tracks whether repeat buyers spend more or less per order than they did on their first purchase.
  • LTV per cohort: Projected or observed lifetime value for a specific acquisition group, often calculated as cumulative revenue per customer at a defined horizon (12 or 24 months).
  • Payback period: The number of months until cumulative revenue per customer equals the CAC for that cohort. For a deeper look at how this connects to customer profitability analysis, the mechanics are the same.
  • LTV:CAC ratio: Cumulative revenue per customer at a given horizon divided by the acquisition cost for that cohort.

Retention rate formula and sample calculation:

Retention Rate (Month N) = (Customers who purchased again in Month N ÷ Total customers in cohort) × 100

Example: A January cohort has 200 customers. In February (Month 1), 60 of them place a second order.

Month 1 Retention Rate = (60 ÷ 200) × 100 = 30%

In March (Month 2), 38 of the original 200 purchase again.

Month 2 Retention Rate = (38 ÷ 200) × 100 = 19%

The drop from 30% to 19% between Month 1 and Month 2 is the cliff. That is where your post-purchase sequence needs to work hardest.

On time intervals: Use weekly cohorts for high-frequency consumables (supplements, pet food, coffee). Monthly cohorts work for most ecommerce categories. Quarterly cohorts suit long-buy-cycle categories like furniture or outdoor equipment, where monthly data is too sparse to be meaningful.

How to run a cohort retention analysis step by step

The workflow below works whether you are pulling data from Shopify, querying a warehouse, or working from a CSV export in Excel.

  1. Define the retained event. A repeat purchase is the standard retained event for ecommerce. Avoid using site visits or add-to-cart events as your retention signal; they do not represent revenue.
  2. Choose cohort granularity. Monthly cohorts are the right default for most stores. Weekly cohorts are useful for high-frequency categories or during a specific campaign window.
  3. Export or query first-purchase data. You need two fields at minimum: customer identifier and first-order date. From Shopify, this is available in the customer export. In SQL, a simple MIN(order_date) GROUP BY customer_id query gives you the cohort assignment.
  4. Build the cohort matrix. Assign each customer to their first-purchase month (cohort row). For each subsequent month, count how many customers from that cohort placed any order. The result is a matrix of counts.
  5. Normalize to percentages. Divide each month’s count by the cohort’s starting size and multiply by 100. This converts raw counts into retention rates, making cohorts of different sizes directly comparable. ChurnDefense’s methodology guide covers this normalization step clearly, including the distinction between logo cohorts (customer count) and revenue cohorts (dollar value retained).
  6. Visualize as a heatmap or retention curve. Color-code the percentage matrix from dark (high retention) to light (low retention). Reading across a row shows how a single cohort decays over time. Reading down a column shows whether newer cohorts are retaining better or worse than older ones at the same age.

Example cohort matrix (counts and percentages):

Reading this table: the January and February cohorts show nearly identical Month 1 and Month 2 retention, which suggests the drop from Month 1 to Month 2 is structural, not a one-month anomaly. The March and April cohorts show slightly better Month 1 retention, worth watching as they age.

Visualization tips: The Month 1 to Month 2 drop is almost always the steepest. If that cliff is larger than 10 percentage points, your post-purchase sequence is underperforming. Reading down a column (same cohort age, different acquisition months) tells you whether your retention is improving over time as you make changes. A flattening curve, where retention stabilizes after Month 3 or 4, signals a loyal core worth investing in.

Statistical significance note: Cohorts smaller than 50 customers produce noisy percentages. A single large order in a small cohort can swing cumulative revenue per customer dramatically. Set a minimum cohort size of 50 customers before drawing conclusions, and flag any cohort below that threshold in your reporting.

Operational checklist for reproducible reporting:

  • Document the exact SQL query or export filter used to define first-purchase date
  • Version-control cohort definitions so a definition change does not corrupt historical comparisons
  • Assign one owner to run and distribute the cohort report on a fixed cadence
  • Store raw counts alongside percentages so the matrix can be rebuilt if the definition changes

Which tools help you run cohort analysis where you already work?

Shopify built-in cohort report

Shopify’s native customer cohort report (available on Shopify and Advanced plans) groups customers by first-purchase month and shows retention, revenue per customer, and orders per customer over time. You can slice by sales channel, product, or promotion without any SQL. The first chart to pull is the retention heatmap filtered by your top acquisition channel. Shopify’s guide walks through the five-step retention workflow in detail. Best for small to mid-size merchants who want a no-code starting point; the limitation is that you cannot join external ad spend data directly inside the report.

Google Analytics 4 (GA4)

GA4’s Retention report under the Life cycle section shows user retention by week, but it tracks sessions and engaged users, not purchase events. For ecommerce cohort analysis, the more reliable path is to export raw event data to BigQuery and run purchase-event cohorts in SQL. GA4 is best for stores that already have BigQuery connected and want to combine on-site behavior with purchase data. Setup complexity is moderate to high, and the native retention report is a rough proxy rather than a true purchase cohort.

Peel

Peel is a Shopify-native analytics tool built specifically for ecommerce cohort work. It pulls order data directly from Shopify and surfaces acquisition cohorts, product cohorts, and channel cohorts with minimal configuration. The standout feature is the ability to push cohort-defined segments back into Klaviyo or other marketing tools as audiences. Best for mid-market Shopify brands that want cohort analysis without a data team. Pricing is a paid SaaS subscription.

Klaviyo

Klaviyo is primarily an email and SMS platform, but its predictive analytics and segment builder let you approximate cohort logic. You can create segments based on first-purchase date, purchase count, and days since last order, then track how those segments perform over time. The limitation is that Klaviyo’s native reporting is not a true cohort matrix; it is more useful as the activation layer once you have identified a cohort signal elsewhere. Best for teams that want to act on cohort insights through email flows without leaving their ESP.

MySQL and Excel (manual workflow)

For teams with access to a database, a two-query approach works well. First, create a customer table with customer_id and first_purchase_month. Second, join that to your orders table and count orders per customer per month. Export to CSV and build the cohort matrix in Excel using a pivot table. This approach gives you full control over cohort definitions and lets you join ad spend data from any source. Setup complexity is high, but the output is the most flexible.

Pro Tip: The most common integration gap is the failure to join ad spend data to cohort outcomes. If your cohort tool cannot connect a customer’s acquisition cost to their cohort row, you cannot calculate true payback period. Peel and Polar Analytics both address this; GA4 and native Shopify do not without additional data work.

Tool Best for Data source Setup complexity Cost signal Actionability Visualization
Shopify report Small to mid merchants Native Shopify No-code Free (plan-gated) Export only Heatmap, retention curve
GA4 + BigQuery Enterprise, multi-channel Universal events High (SQL required) Free + BigQuery costs Audience export Retention curve
Peel Mid-market Shopify brands Native Shopify Low (no-code) Paid SaaS Klaviyo segment push Heatmap, LTV curve
Klaviyo Email-first teams Shopify/API Low Paid (per contact) Direct flow trigger Segment trends
MySQL + Excel Analysts, custom stacks Any database High Free (labor cost) Manual export Custom pivot/heatmap

Mixpanel’s cohort analysis primer adds a useful framework for evaluating any platform: look for multi-criteria cohort support, “did not do” logic (customers who did not repurchase), and identity resolution across devices. These features determine whether a tool can produce cohorts that are actually actionable as audiences.

Two worked examples: numeric cohort table and a paid-media channel comparison

Example A: reading a numeric cohort table

Using the January cohort from the matrix above (200 customers, 30% Month 1 retention), here is what cumulative revenue per customer looks like if average order value holds at $65:

If CAC for this cohort was $45, payback occurs before Month 1 on a revenue basis. But if contribution margin is 40%, the margin-adjusted payback is $45 ÷ ($65 × 0.40) = 1.73 orders, meaning you need the customer to place roughly two orders before you recover acquisition cost on a margin basis. That changes the urgency of Month 1 retention considerably. For a detailed breakdown of how to separate fixed and variable costs in this calculation, the mechanics matter when you are converting revenue-per-cohort into a true profitability signal.

It only becomes meaningful when you set it against your margin structure and CAC.

Example B: acquisition-channel cohort comparison

Suppose you run three channels: Meta paid social, Google paid search, and organic search. Here is a simplified channel cohort comparison at Month 3:

  • Meta paid social: 300 customers acquired, Month 1 retention 28%, Month 3 retention 9%, average CAC $52
  • Google paid search: 180 customers acquired, Month 1 retention 34%, Month 3 retention 18%, average CAC $48
  • Organic search: 120 customers acquired, Month 1 retention 41%, Month 3 retention 26%, average CAC $0 (content investment amortized)

Meta looks competitive on day-0 ROAS. But by Month 3, its retention is half of Google’s and one-third of organic. The payback math changes entirely once you factor in that Meta customers are generating far less cumulative revenue per customer at the same cohort age.

Recommended actions from this analysis:

  • Pause or reduce Meta spend on the specific campaign or audience driving the lowest-retention cohort. Test a different creative angle or audience targeting before scaling back up.
  • Add a post-purchase email sequence specifically for Meta-acquired customers, triggered at day 14 (before the Month 1 cliff). The goal is to close the retention gap, not just accept it.
  • Increase investment in organic content that is driving the highest-retention cohort, even if volume is lower. The LTV math supports it.
  • Measure impact by checking whether the next Meta cohort’s Month 2 retention rate improves after the creative change. One cohort cycle is enough to see a directional signal.

Common pitfalls that lead to wrong conclusions

Cohort analysis is only as reliable as the decisions made before the first query runs. These are the mistakes that produce misleading results:

  • Cohorts too small to be meaningful. A cohort of 15 customers will show 0% or 100% retention on any given month depending on whether one person orders. Set a minimum of 50 customers per cohort before drawing conclusions.
  • Blended averages masking channel differences. Reporting a single store-wide retention rate hides the fact that one channel might retain at 35% while another retains at 12%. Always segment by acquisition source before presenting retention numbers to leadership.
  • Attribution noise. Last-click attribution assigns the first purchase to one channel, but the customer may have touched four touchpoints. Cohort analysis inherits whatever attribution model your order data uses. Be explicit about this limitation when presenting channel cohorts.
  • Promotional cohort distortion. A Black Friday cohort will almost always show lower long-term retention than a non-promotional cohort. Mixing them into the same baseline cohort makes your “normal” retention look worse than it is. Label and isolate promotional cohorts. Given how prevalent discount-driven purchasing is in the US market, per Statista’s coupon-use research, this is not a minor edge case.
  • Inconsistent identity resolution. If a customer checks out as a guest, then creates an account, then uses a different email, they may appear as three separate customers in your cohort. This inflates cohort sizes and deflates retention rates. Resolve identity before building cohorts.
  • Forgetting to join ad spend. A cohort that shows strong revenue per customer but was acquired at 3× the normal CAC is not a success story. Without ad spend joined to cohort data, you cannot calculate payback period.
  • Using the wrong retained event. Tracking site visits as “retention” instead of purchases overstates engagement and understates churn. Always use a revenue-generating event as the retained event for ecommerce cohort work.

Best-practice checklist:

  • Minimum cohort size: 50 customers before drawing directional conclusions
  • Isolate and label promotional cohorts separately (naming convention: YYYY-MM_PROMO vs YYYY-MM_FULL)
  • Document cohort definitions in a shared location and version-control any changes
  • Run a baseline cohort (all customers, all channels) alongside segmented cohorts so you always have a reference point
  • For causal inference on a retention intervention (new email sequence, loyalty program), use an A/B holdout group rather than comparing pre/post cohorts, which conflates the intervention with seasonal or mix changes

How to translate cohort findings into concrete retention actions

Cohort data tells you where customers are leaving. The action menu below maps each signal to a specific response.

Build a 3-email sequence: a product education email at day 7, a replenishment or cross-sell prompt at day 21, and a loyalty or subscription offer at day 28. Measure impact on the next cohort’s Month 1 rate. Owner: CRM or email marketing.

Hand placing calendar sheet on table

Win-back at the identified drop point. Calculate the average time between first and second purchase for your retained customers. Trigger a win-back campaign at two-thirds of that interval for customers who have not yet repurchased. If your average time-to-second-order is 45 days, trigger win-back at day 30. KPI: Month 2 repeat rate for the cohort that received the win-back.

Product-led merchandising. If your gateway product cohort (customers whose first order was Product A) shows 2× the Month 3 retention of customers who first bought Product B, that is a merchandising signal. Feature Product A more prominently in acquisition campaigns and on the homepage. Owner: merchandising and paid media.

Reduce discount frequency and test a value-add offer (free gift, early access) instead of a price cut.

Channel budget reallocation. Once you have three months of data on a new channel cohort, you have enough signal to compare its payback trajectory against your best-performing channel. Shift budget toward channels where the cohort’s cumulative revenue per customer is on track to exceed CAC within your target payback window.

Example email trigger rule: Segment customers from the January cohort who did not purchase in Month 1. At day 25, send: “Your [Product Name] is running low: here’s what pairs well with it.” Cohort trigger: first_purchase_month = January AND order_count = 1 AND days_since_first_order >= 25.

What to run weekly, monthly, and quarterly

Cohort analysis only compounds in value when it runs on a fixed cadence. Here is a practical operating rhythm:

Weekly (smoke-check):

  • Review the most recent two cohorts for any anomalous Month 1 retention drop
  • Flag any cohort below the minimum size threshold
  • Check whether any recent campaign or promotion created a new cohort that needs to be labeled separately

Monthly (full review):

  • Run the complete cohort matrix and update cumulative revenue per customer for all active cohorts
  • Calculate payback period for each acquisition channel cohort
  • Compare Month 1 and Month 2 retention for the current month’s cohort against the prior three months
  • Distribute the report to marketing, finance, and product owners

Quarterly (strategic deep-dive):

  • Run channel cohorts and product cohorts side by side
  • Evaluate whether any retention intervention (new email sequence, loyalty program, subscription offer) has moved the curve for the cohorts it targeted
  • Update LTV projections for each channel based on observed cumulative revenue per customer at the 6-month mark
  • Review cohort naming conventions and definition governance; update documentation if anything changed

Operational checklist for reproducible exports:

  • Required fields: customer ID, first-order date, order date, order revenue, acquisition channel (UTM source/medium), and promotion flag
  • Minimum cohort size: 50 customers
  • Naming standard: YYYY-MM_[CHANNEL]_[PROMO/FULL]
  • Sign-off owner: one named analyst or operator per report cycle
  • Version-control: store the SQL query or export filter alongside the output file, not just the spreadsheet

Once the monthly cadence is stable in Excel or CSV, moving to a SQL-based or product analytics workflow (Peel, a warehouse query, or a BI tool like Looker) removes the manual rebuild step and makes the report reproducible without human intervention each cycle.

When should you DIY cohort analysis, and when does it make sense to hire help?

The honest answer is that most ecommerce teams can build a basic acquisition-date cohort in Shopify or Excel within a day. The methodology is not complicated. What gets expensive is the work that sits around the methodology: joining ad spend from Meta, Google, and TikTok to cohort rows, resolving customer identity across guest checkouts and multiple email addresses, and translating cohort findings into a prioritized action plan that finance and marketing will actually act on.

The DIY path works well when your data lives in one place (Shopify, one ad channel), your team has a basic analyst, and you are willing to accept some attribution imprecision. The signal is still directional and useful.

The hire-help threshold tends to appear when three or more of the following are true: your ad spend is spread across four or more channels with no unified attribution layer; your customer data is fragmented across Shopify, a 3PL, a subscription platform, and a loyalty tool; you need cohort findings presented to investors or a board in a format that connects directly to LTV-to-CAC and payback period; or your team has been looking at blended metrics for so long that there is organizational resistance to what the cohort data will show.

A short engagement in that scenario looks like a data audit, a cohort build across your top three acquisition channels, and a prioritized playbook that tells you which channel to cut, which retention sequence to build first, and what payback period improvement is realistic in the next two cohort cycles. The output is not a dashboard. It is a decision.

What Commerce Catalyst offers founders who need cohort-driven clarity fast

Running cohort analysis is one thing. Knowing what to do with the numbers when your data is fragmented, your ad spend is spread across five platforms, and your team is already stretched is a different problem entirely.

Commercecatalyst

Commerce Catalyst works with consumer brand founders at the $5M–$75M revenue stage who need cohort-driven retention and payback analysis translated into decisions, not just charts. The DTC Financial Health Assessment connects your cohort revenue data to your actual margin structure and CAC, so you can see which channels are genuinely profitable at Month 3 and which are quietly destroying cash flow. For brands with fragmented data or complex ad-spend attribution, the DTC Operator Diagnostic identifies the specific data gaps that are preventing clean cohort work and delivers a prioritized fix list. If you need ongoing support to act on cohort signals at scale, the fractional COO engagement provides hands-on execution alongside the analysis. Book a diagnostic call to find out where your retention curve is breaking and what it will take to move it.

Sources

The sources below were used in building this guide and are worth reading directly for deeper methodology or tool-specific detail:

Related tools

>>> next step

Want to see where your business actually stands?

Run the numbers through the diagnostic, or talk it through with someone who has been in your seat.

Get the Diagnostic Book a Founder Hour