Look at Cohorts Before You Scale Spend

Why First-Order ROAS Lies

A campaign delivering 3.0x ROAS on day 1 looks profitable. But if 80% of those customers never return, the blended metric obscures a unit economics problem. First-order ROAS is a snapshot of acquisition efficiency, not customer value.

Blended ROAS averages new and repeat revenue together. A brand acquiring 100 customers at $50 AOV (3.0x ROAS) and retaining 5 of them for a second $30 purchase reports strong metrics while the retention cliff is invisible in the top-line number.

Scaling spend on a 3.0x first-order ROAS campaign assumes repeat revenue will follow. It won't, if the cohort isn't built to retain. The result: higher absolute spend, same or worse lifetime value per customer.

Cohort View: What to Track

Organize customers into weekly or monthly acquisition buckets. For each cohort, calculate repeat purchase rate (RPR), average order value on repeat purchases, and cumulative revenue per customer by week 1, 4, 12, and 52.

The standard cohort table rows are time periods (weeks or months since acquisition); columns are cohorts (acquisition dates). Each cell shows repeat revenue, repeat rate, or customer count. A healthy D2C brand shows RPR stabilizing around 15 - 25% by week 4 and repeat AOV holding within 10% of first-order AOV.

Week 1 repeat rate: Captures immediate re-engagement (gift-giving, bundle stacking, or product fit). Target: 2 - 5%.
Week 4 repeat rate: First natural replenishment cycle. Target: 8 - 15% for consumables, 3 - 8% for durables.
Week 12 repeat rate: Seasonal and subscription signal. Target: 12 - 25% for consumables.
Repeat AOV vs. first AOV: If repeat AOV drops >15%, product fit or positioning may be weak.
Cohort decay slope: If RPR drops >50% from week 4 to week 12, retention mechanics are broken.

The Scaling Decision Rule

Do not scale acquisition spend until the cohort is 4 weeks old and shows RPR ≥ 8% and repeat AOV ≥ 85% of first AOV. Scaling before this gate risks acquiring customers with poor lifetime value at higher volume.

If a cohort hits 4-week RPR of 5% or lower, pause the campaign and audit product fit, email sequence, and post-purchase experience before increasing spend. Throwing more budget at a broken cohort compounds the loss.

If a cohort shows strong 4-week metrics (RPR ≥ 12%, repeat AOV ≥ 90% of first AOV), scale spend by 20 - 30% weekly until CAC rises or RPR declines. Monitor weekly to catch decay early.

Cohort Decay Patterns and Fixes

Cliff decay (RPR drops 60%+ from week 4 to week 8) signals a one-time purchase trigger (discount, novelty, urgency) without product stickiness. Fix: Strengthen product education, add subscription or loyalty mechanics, or narrow targeting to repeat-prone segments.

Gradual decay (RPR drops 20 - 30% from week 4 to week 12) is normal. Accelerate it only if the slope steepens unexpectedly, which may indicate seasonal demand or competitive pressure.

Flat or rising RPR after week 4 is rare and valuable. It signals strong product-market fit or a subscription model working. Allocate more budget to these cohorts and analyze their characteristics for targeting clues.

Cohort Analysis in Practice: Checklist

Set up a weekly cohort table in a spreadsheet or BI tool. Rows = acquisition week, columns = metrics (customer count, repeat rate, repeat revenue, cumulative LTV). Automate the pull from your data warehouse or Shopify API.

Review cohorts every Monday. Flag any cohort with 4-week RPR below threshold or repeat AOV below 85% of first AOV. Assign an owner to investigate and propose a fix within 48 hours.

Compare cohorts across channels (paid search, social, email, organic). If one channel's cohorts consistently underperform, reduce spend and reallocate to higher-retention channels.

Test retention levers on new cohorts: email timing, product bundling, loyalty incentives, post-purchase messaging. Measure impact on week 4 and week 12 RPR before scaling.

Cohort Metrics vs. Blended Metrics

Blended ROAS = total revenue / total ad spend. It averages all customers and all time periods. Useful for board reporting, not for operational decisions.

Cohort ROAS = repeat revenue (weeks 1 - 12) / acquisition spend for that cohort. It isolates the true return of a single acquisition effort. This is the metric that predicts scalability.

Example: Campaign A shows 3.0x blended ROAS but 1.8x cohort ROAS (first-order revenue only). Campaign B shows 2.2x blended ROAS but 2.8x cohort ROAS (strong repeat). Scale Campaign B, pause Campaign A.

Common Cohort Mistakes

Waiting too long to analyze. If you wait 12 weeks to review a cohort, you've already scaled spend on a broken acquisition engine. Review at week 4.

Ignoring seasonality. A cohort acquired in November will have different repeat patterns than one acquired in June. Compare cohorts within season, not across.

Confusing correlation with causation. A cohort with high RPR may have high AOV because of product mix, not because of the campaign. Segment by product to isolate the signal.

Not accounting for attribution windows. If your email platform has a 30-day attribution window, repeat purchases may be credited to email, not the original acquisition channel. Align your cohort window to your attribution model.

Questions

FAQ

What's a good repeat purchase rate for a DTC brand?

For consumables (food, beauty, wellness): 15 - 25% by week 12. For durables (apparel, home goods): 5 - 12% by week 12. For subscriptions: 60%+ by week 4. Benchmark against your category and competitor cohorts, not absolute numbers.

How do I know if a cohort is too young to scale on?

Never scale on first-order ROAS alone. Wait until the cohort is 4 weeks old and has at least 50 repeat purchases. If the cohort is smaller, wait longer. A cohort with 10 repeat purchases and 20% RPR is noise; one with 200 repeat purchases and 15% RPR is signal.

Should I segment cohorts by traffic source or product?

Both. Start with traffic source (paid search, social, email, organic) to identify which channels drive retention. Then segment by product within each channel to isolate which products drive repeat. A cohort acquired via TikTok may have 8% RPR overall but 25% RPR for a specific SKU.

What if my repeat AOV is higher than my first AOV?

This is often a good sign - customers are buying more on repeat, either because they trust the brand or because your post-purchase upsell is working. But verify it's not driven by a small number of high-value repeat buyers. Check median repeat AOV, not just mean, to catch outliers.

Want this on your account?

Thirty minutes. Bring the number that keeps you up.

Book a demo

Cohort Analysis: The Gate Before Scaling Spend