You acquired one hundred customers in January. You acquired one hundred customers in February. Both cohorts are now in June. January cohort: fifty still customers. February cohort: thirty still customers. Same acquisition, different retention. Why. Cohort analysis reveals this. It compares groups of customers acquired at the same time. Shows which cohorts are healthy and which are leaking. This article explains cohort analysis and how to use it to improve retention.
Understanding cohorts and why they matter
What a cohort is and why time-based grouping matters
A cohort is a group of customers acquired in the same time period. January customers are one cohort. February customers are another. Cohort analysis tracks each group separately. Shows which cohort is healthy. Which is leaking. January cohort might retain fifty percent. February might retain thirty percent. Something changed between January and February. Cohort analysis reveals it.
Cohort analysis vs aggregate customer analysis
Aggregate view: currently have two thousand customers. Lost one hundred last month. Retention ninety-five percent. But this hides cohort differences. Some cohorts retain eighty percent. Some retain ninety-nine percent. Cohort view reveals the truth. Aggregate hides patterns.
Building a cohort retention table
Tracking customers by acquisition month
January cohort: one hundred customers acquired. Track them. February cohort: one hundred customers acquired. Track them separately. March cohort: one hundred customers acquired. Do the same. Each cohort is a column. Each month is a row.
Month zero, month one, month two retention
January cohort month zero: one hundred customers. January cohort month one: seventy customers still. Retention seventy percent. January cohort month two: fifty-five customers. Retention fifty-five percent. Track each cohort separately. Month-by-month retention.
Calculating retention rate by cohort
Percent of customers returning each month
Month zero: one hundred customers. Month one: seventy return. Seventy percent retention. Month two: fifty-five return. Fifty-five percent retention. Track percentage not absolute count. Percent shows the real retention rate.
Retention curves and decline patterns
Good cohort: one hundred in month zero. Eighty in month one. Seventy in month two. Sixty in month three. Gradual decline. Healthy pattern. Bad cohort: one hundred in month zero. Fifty in month one. Twenty in month two. Ten in month three. Steep cliff. Leaking fast. Curves reveal health.
Identifying high-retention vs low-retention cohorts
What changed between good and bad cohorts
January cohort retained well. February cohort retained poorly. What changed. Product launch in February. Customer service change. Marketing changed. Acquisition source changed. Find the difference. The change caused the problem.
Seasonal patterns in cohort retention
Summer cohorts might retain poorly. Holiday cohorts might retain better. Customer type varies seasonally. Find patterns. Summer might bring bargain hunters. Holiday might bring gift givers. Different customer types have different retention.
Diagnosing retention problems
Is it the acquisition source or the product
High retention from organic traffic. Low retention from paid ads. Problem is acquisition source. Paid ads bring wrong customers. Fix targeting. Low retention from all sources. Problem is product or onboarding. Fix product.
Is it a seasonal effect or a systemic issue
January cohort: fifty percent retention. February cohort: forty-five percent. March cohort: fifty-two percent. Variation is normal. Systemic issue: January cohort: fifty percent. February cohort: twenty percent. March cohort: fifteen percent. Declining. Something broke. Find it.
Improving retention across cohorts
Changes that improve all cohorts
Product improvements help all cohorts. Better customer service helps all. Better onboarding helps all. Changes that help all cohorts are systemic improvements. Changes that help only some cohorts are targeted.
Testing retention improvements
Test new onboarding with February cohort. See if retention improves. Do not test with all cohorts. A/B test. See what works. Implement what works. Test is the only way to know.
Forecasting revenue based on cohort retention
Predicting future revenue from retention curves
Cohort acquired one thousand customers. Month one retention fifty percent. Five hundred remaining. Month two retention seventy percent. Three hundred fifty remaining. Forecast future months. Retention curves predict cash.
Impact of retention improvements on revenue
Improve month one retention from fifty percent to sixty percent. One hundred more customers stick. Multiply by customer lifetime value. Calculate impact. Five percent retention improvement might increase annual revenue fifty thousand dollars. Retention is high leverage.
Frequently asked questions
Should I measure retention monthly, weekly, or daily?
Depends on purchase frequency. If customers buy monthly, measure monthly. If they buy weekly, measure weekly. If they buy daily, measure daily. Match retention measurement to purchase cycle. Otherwise noise hides signal. With WEMASY's website builder, you can set this up directly on your website.
What is a good retention rate?
Depends on industry and customer type. SaaS might aim for ninety percent monthly retention. E-commerce might aim for twenty percent repeat rate. Luxury might have low frequency but high value. Know your baseline. Improve from there.
Can I segment cohorts by acquisition source?
Yes. January organic cohort. January paid ads cohort. January referral cohort. Compare retention by source. Organic might retain eighty percent. Paid might retain forty percent. Find what works. Invest in high-retention sources.
How do I know if a cohort retention problem is fixable?
If problem is acquisition source, fixable. Change source, retention improves. If problem is product, fixable. Improve product, retention improves. If problem is market, unfixable. Market does not want product. Move on.
Should I focus on improving retention or acquisition?
Both but prioritize retention. Acquiring and losing is expensive. Retaining is cheap. Improve retention first. Then grow acquisition. Ten percent retention improvement saves more than acquiring ten percent more customers.
How far back should cohorts go in my analysis?
Go back as far as you have data. One year is good. Two years is better. Longer history shows more patterns. But watch for market changes. Three-year-old cohorts might not predict today's cohorts.