Loading tutorials…
Loading tutorials…
Blended ROAS tells you which channels drove revenue last month. Cohort analysis tells you which channels drove customers who are still spending today. That's the difference between scaling profitably and scaling broke.
Who this is forDTC owners on Triple Whale Scale tier or above (cohort analysis is gated by tier). Especially valuable if you have 12+ months of order history and 1,000+ customers — without that, cohort data is statistically thin.
What you'll need
Step 1
Cohort analysis below a data threshold is noise, not signal. You need 6+ months of orders + 1,000+ customers per cohort month for the curves to be readable.
Open Shopify Admin → Analytics → Customers. Count total customers and check earliest order date.
Minimum thresholds: 6 months of order history, 200+ orders/month average, 1,000+ total customers.
Below these thresholds: cohort curves are statistically noisy and you'll make wrong decisions based on small sample sizes.
If you don't have enough data yet: wait. Focus on basic blended ROAS and channel attribution until you cross the threshold. Coming back to cohort analysis in 6 months will be more useful than struggling with thin data now.
Step 2
Cohort LTV without COGS is just cohort revenue. Set "cost per item" in Shopify for at least 80% of revenue SKUs before building cohort dashboards.
Shopify Admin → Products → bulk action → Edit "Cost per item" for top revenue SKUs.
If you have 200+ SKUs, export the product CSV, fill in cost per item in spreadsheet, re-import.
COGS should include: product cost + inbound shipping (per unit) + packaging. Exclude: fulfillment-side shipping (Triple Whale tracks separately) and ad spend.
Verify in Shopify Admin → Reports → Margins. Top 20 products should all have a margin number, not 100%.
Triple Whale syncs COGS overnight. By tomorrow, gross profit numbers should populate.
Step 3
Group customers by the month they first ordered. Watch them spend over time. This is the foundation of LTV analysis.
Triple Whale → Cohorts → + New Cohort.
Cohort type: "Acquisition Month."
Metric: "Cumulative Revenue Per Customer" (most useful) or "Cumulative Gross Profit Per Customer" (if COGS is clean).
Time range: last 12 months. Cohort months: monthly.
Save as "Acquisition Cohort — LTV."
The chart shows: each month's acquisition cohort as a line. X-axis = months since acquisition. Y-axis = cumulative revenue per customer.
Read it: do cohorts curve up steeply (high repeat) or flatten quickly (one-and-done)? Do recent cohorts curve faster than older ones (LTV improving) or slower (LTV declining)?
Step 4
Same cohort dashboard, broken down by the channel that acquired the customer. This is where you discover which channels drive lifetime value vs one-and-done purchases.
Open the Acquisition Cohort dashboard → "Break down by" → "First Order Channel."
You'll see one cohort line per channel: Meta, Google, TikTok, Email, Direct, etc.
Compare: which channel's curve goes highest after 6/12 months?
Common finding: Meta acquires customers with higher repeat rates than TikTok (TikTok tends toward impulse). Google Brand search customers have the highest LTV (already brand-loyal). Email-acquired customers spend least (often already on email list).
Action: shift acquisition budget toward channels with highest 90-day or 180-day LTV — not the channel with the best week-1 ROAS.
Step 5
Break cohorts down by the first product. Some products are "gateway" products that drive repeat purchases. Others are one-time gifts.
Cohort dashboard → "Break down by" → "First Order Product."
You'll see one cohort line per top-N first product.
Identify "gateway products": products where first-time buyers come back and spend more.
Identify "one-and-done products": products with high first-order revenue but low repeat. These are usually gifts, impulse purchases, or seasonal items.
Action: feature gateway products in acquisition ads even if they have lower AOV. Total LTV per acquired customer > Day 1 AOV.
Step 6
Different from LTV — retention shows what % of each cohort is still ordering N months in. Critical for understanding churn.
Cohorts → + New → "Retention Curve."
Metric: "% of cohort that ordered in month N."
You'll see: a heatmap or curve showing what % of customers came back in months 1, 2, 3, etc.
Healthy DTC benchmarks: 20-30% retention at month 1, 10-15% at month 3, 5-10% at month 6.
Below benchmark: your product or post-purchase experience has a churn problem. Above benchmark: scale acquisition aggressively because every customer you acquire is worth more than the industry average.
Step 7
Cohort data is slow-moving. Don't make weekly decisions on it. Monthly review, quarterly action.
Monthly: open the Acquisition Cohort dashboard. Look at the most recent complete cohort month vs the cohort 3 months prior. Is LTV improving or declining?
Quarterly: deeper review. Look at channel breakdown. Is the channel mix shifting LTV up or down? Action: shift budget allocation if a channel shows persistently lower LTV.
Annually: full strategic review. Are gateway products still driving repeat? Is overall LTV trending up year-over-year? Use cohort data to justify pricing changes, subscription launches, or expansion into new product categories.
Common mistakes
Running cohort analysis without enough data
What goes wrong: You have 3 months of orders and 200 customers. Cohort curves look erratic. You make budget decisions on what's actually noise. Six months later, you realize the 'pattern' you optimized for was random variation.
How to avoid: Wait until you have 6+ months of history and 1,000+ customers minimum. Below that, focus on weekly attribution dashboards, not cohorts.
Cohort LTV without COGS = cohort revenue
What goes wrong: You read "Cumulative Revenue Per Customer" as if it's profit. High-revenue, low-margin products look great. You scale them, lose money, blame the model.
How to avoid: Set COGS in Shopify before reading any LTV chart. Switch the cohort metric to "Cumulative Gross Profit Per Customer" once COGS is in.
Comparing cohorts of different sizes
What goes wrong: September cohort = 800 customers. October cohort = 200 customers (you ran a small test). October curve looks erratic — small samples are noisy. You misread it as a strategic shift.
How to avoid: Filter cohorts by minimum size (e.g., 500+ customers). Drop cohorts below threshold from comparison.
Optimizing acquisition for Day 1 ROAS while LTV declines
What goes wrong: TikTok shows 4x Day 1 ROAS, Meta shows 2.5x. You shift budget to TikTok. 90 days later, TikTok customers haven't repurchased. Total LTV is half of Meta's. You scaled the worse channel.
How to avoid: Track 90-day LTV by acquisition channel. Decisions on channel mix should weight 90-day LTV heavily, not just Day 1 ROAS.
Reading cohort data weekly
What goes wrong: Cohort numbers move slowly — a 1% change week to week is noise. You react to weekly noise, switch budget allocations, never let any strategy run long enough to validate.
How to avoid: Monthly review minimum. Quarterly action. Cohorts don't earn weekly attention.
Ignoring the gateway product insight
What goes wrong: You feature your highest-margin product in ads. It's a one-time purchase. New customers buy once and never return. Repeat rate stays low. You blame retention emails when the issue is product mix at acquisition.
How to avoid: Break cohort by first product purchased. Identify your gateway products (high repeat rate, even if lower margin). Feature gateways in acquisition ads.
Recap
Done — what's next
How to set up Triple Whale for your Shopify store
Read the next tutorial
Hand it off
Cohort analysis is the most strategic feature in Triple Whale and the one most DTC owners never use well. The gap between 'can read a curve' and 'can translate cohort data into channel + product strategy' is where specialists earn their fee. A vetted DTC analytics specialist runs monthly cohort reviews + quarterly strategy updates — typically $300-600/mo for ongoing cohort governance plus the initial $500-800 setup.
See specialist rates
Hugely category-dependent. Beauty/supplements: $80-150 (high repeat). Apparel: $60-100. Furniture: $200-400 (high AOV, low repeat). Compare your cohorts year-over-year — improving is what matters more than hitting an absolute number.
Triple Whale cohorts are sufficient for 90% of DTC use cases. Build in a data warehouse (BigQuery + Looker, etc.) only if you have a data team and need custom cohort definitions Triple Whale can't express.
Three usual causes: (1) newer cohorts haven't had time to mature — month-1 vs month-12 retention are different stages; (2) you shifted acquisition mix toward channels with worse LTV; (3) product quality or post-purchase experience declined. Diagnose by breaking cohorts by channel.
Four levers: (1) acquire from higher-LTV channels; (2) feature gateway products at acquisition; (3) improve post-purchase email/SMS for retention; (4) launch subscriptions for high-affinity products. Cohort breakdowns tell you which lever has the most slack.
Not directly — you don't have their data. Industry reports (Recharge, Klaviyo, Triple Whale's own benchmarks) publish category averages. Use them as rough benchmarks, not targets.
90 days minimum to see month-3 LTV. 6 months to see month-6 LTV. Don't kill or scale a channel based on month-1 cohort behavior — too noisy.
Triple Whale
Most founders sign up for Triple Whale, click 'Connect Shopify,' and assume the rest will figure itself out. Three weeks later they're paying $200/mo for a dashboard nobody opens because attribution looks weird. Here's the setup that actually delivers value from day one.
Triple Whale
Triple Whale ships with 40+ default widgets and most owners build dashboards with all of them — then never open the dashboard again. Here's how specialists build dashboards that get read every morning.
Triple Whale
Triple Whale's attribution model picker is buried in Settings and most users never touch it. The default is fine for some stores and dangerously wrong for others. Here's the decision framework that drives 80% of the value you get out of the platform.
Triple Whale
Moby is Triple Whale's pitch for the AI era — natural-language queries against your DTC data. It's genuinely useful for specific tasks and overhyped for others. Here's the setup + the honest framework for when Moby beats opening the dashboard.
Klaviyo
Welcome flows are the highest-revenue flow in most Klaviyo accounts — typically 30-45% of automated email revenue. A bad welcome flow leaves 60% of that on the table. This is the build that captures it.