Cohort Analysis Infrastructure for DTC

The problem cohort analysis solves

Your CAC is up twenty-two percent year over year. Your thirty day repeat rate is steady. Your AOV is steady. Your LTV number, if you have one, shows flat or slightly up. Leadership asks whether the business is healthy and the answer, honestly, is you do not know, because averages hide everything. The January cohort might be thriving while the June cohort might be collapsing, and the blended number tells you nothing because the two partially cancel.

This is what happens when a brand reports on aggregates without cohort structure underneath. New customer behavior blends with returning customer behavior. Good acquisition channels blend with bad. Subscription customers blend with one-time buyers. Discount-driven cohorts blend with full-price cohorts. Every meaningful decision depends on separating these groups, and without cohort infrastructure the separation is impossible.

The specific decisions that depend on cohort analysis are the ones that matter most. Whether to raise or lower discount depth depends on the repeat behavior of discount-driven cohorts versus full-price cohorts. Whether to expand a channel depends on the retention curve of customers acquired from that channel, not just its CAC. Whether to launch a subscription program depends on the repeat behavior of customers who bought the products that would convert best to subscription. Without cohort data these are all gut calls, and gut calls compound into eighteen-month mistakes.

Our approach

We run a four week cohort analysis infrastructure engagement.

Step one is the cohort definition. We agree on cohort dimensions: first order month is always included, and we typically add first product, first channel, first discount tier, and subscription status. We agree on retention metrics: order count at thirty, sixty, ninety, one hundred eighty, three hundred sixty, and seven hundred twenty days since first order, revenue cumulative at the same windows, and active customer rate at each window.

Step two is the warehouse model. We build a cohort table in the warehouse, one row per customer, with first order attributes plus a series of cumulative metrics by window. This table is the single source of truth for every cohort report and every downstream LTV model. We build it in dbt so transformations are version controlled and any stakeholder can read the SQL.

Step three is the dashboard layer. We build cohort dashboards tuned to decision cadence. Monthly cohort review for leadership, with revenue per cohort and retention curves. Weekly channel cohort view for the paid team, with CAC divided by twelve month revenue per cohort to produce a CAC payback curve by channel. Monthly product cohort view for merchandising, showing which first products produce which retention outcomes.

Step four is the readout and handoff. We write a twenty page cohort health report covering the last eighteen to thirty-six months of cohorts, identifying trends, outliers, and decisions the data supports. We train the leadership team, the paid team, and the merchandising team on the dashboards. We hand over the SQL so the client team can extend the cohort table without us.

What you get

▸ Cohort definition document specifying every cohort dimension and retention window. ▸ Cohort table in the warehouse, one row per customer, refreshed daily. ▸ dbt transformation layer producing the cohort table from Shopify and GA4 sources. ▸ Monthly cohort review dashboard for leadership. ▸ Weekly channel cohort dashboard for the paid team with CAC payback by channel. ▸ Monthly product cohort dashboard for merchandising. ▸ Subscription versus one-time cohort comparison if subscription is in the mix. ▸ Discount tier cohort comparison for promo planning. ▸ Eighteen to thirty-six month cohort health report with identified trends and decisions. ▸ Training sessions for each stakeholder group. ▸ Handoff documentation and SQL the client team can extend.

Timeline

Four weeks in three phases.

Week one is definition and discovery. We agree on cohort dimensions and retention metrics, audit the existing data, and identify any gaps in historical Shopify or GA4 data.

Weeks two and three are warehouse and dashboard build. We build the cohort table in dbt, verify it against Shopify order history, and build the dashboards.

Week four is readout and training. We deliver the cohort health report, train each stakeholder group, and hand over documentation.

Mini case anatomy

A beauty brand in the twenty to thirty million revenue range had a ninety day repeat rate that had been steady at twenty-eight percent for two years. Leadership assumed retention was healthy. When we built the cohort infrastructure we found that the twenty-eight percent blended rate was hiding a collapse from thirty-four percent in H1 2023 cohorts to twenty-two percent in H2 2024 cohorts. The rise in new customer volume from aggressive paid social had masked the retention decline in the blended number.

The cohort channel view showed that the H2 2024 decline was concentrated in TikTok-acquired customers, who were repeating at fourteen percent versus thirty-one percent for Meta-acquired and thirty-eight percent for email-acquired. The cohort product view showed that TikTok-acquired customers were disproportionately buying a hero SKU that did not naturally lead to repeat purchase. Meta-acquired customers were buying the starter set that did.

The brand did not cut TikTok. Instead they changed the TikTok landing page to feature the starter set rather than the hero SKU. Six months later the TikTok cohort repeat rate had moved from fourteen percent to twenty-four percent, still below Meta but closing. Blended ninety day repeat moved from twenty-two percent back to twenty-seven percent. For the underlying logic see our guide on ecommerce customer lifetime value.

FAQs

See frequently asked questions below. Cohort analysis is the foundation for LTV modeling, which we usually recommend as a follow-on engagement. It also depends on clean GA4 implementation and feeds into ecommerce dashboards. For the broader picture see our analytics and reporting hub and our attribution for DTC using MER guide.