I built a synthetic 12-month dataset modelled on a DTC subscription supplement brand and ran the kind of segmentation and retention analysis I would want in place for a brand at this stage: where renewal is dropping off, which channels bring subscribers who stick around, and what a CRM programme built around those segments could look like.
For subscription supplement brands, the headline growth number is usually new subscribers acquired, but the number that actually determines whether the business compounds is the renewal rate: how many of those subscribers are still getting a delivery in month two, month three, month six. I built a synthetic dataset modelled on a subscription vitamin brand: 1,170 subscribers, 1,552 orders across 12 months, across four acquisition channels, to dig into three questions: is the renewal rate healthy, which channels bring subscribers who actually stick around, and which subscriber segments need a different kind of CRM treatment.
Benchmark the overall renewal rate against what's typical for a subscription supplement brand, and find out whether it's a red flag or just room to improve.
Acquisition cost is only half the story for a subscription brand: a cheap subscriber who cancels after one delivery is more expensive than an expensive one who renews for a year.
Not every subscriber should get the same lifecycle emails. Segment subscribers by behaviour and turn that into a CRM and content plan a team could actually run.
As a self-directed concept project, I played every role: I built the synthetic 12-month order dataset in Python, ran the renewal-rate and channel analysis, segmented subscribers using an RFM-style (recency, frequency, monetary) approach, and translated the four resulting segments into a CRM lifecycle and content plan, the kind of handoff I'd want to deliver to a marketing team alongside the analysis itself.
The segmentation pointed to two connected moves: where to put acquisition budget, and how to treat subscribers differently once they're in.
Renewal rate varied hugely by channel, from 14.2% on Paid Social to 37.6% on micro creator partnerships. Modelled recommendation: shift budget away from the channel bringing in the most one-and-done subscribers, and restructure the creator programme around the partners whose audiences actually renew.
Four segments, four different jobs: deepen the relationship with Founding Subscribers, nudge Building the Habit subscribers toward a second and third renewal, give Tried It Once subscribers a reason to come back, and build a structured win-back flow for subscribers who've Paused & Lapsed.
Modelled 1,170 subscribers and 1,552 orders across 12 months and four acquisition channels: Creator Partnerships (Macro & Micro), Paid Social, and Organic & Referral, with realistic variation in order frequency, timing and value.
Calculated second-month renewal rate overall and by channel, alongside CAC and ROAS trends over the 12-month window, to find where acquisition cost was rising fastest relative to subscriber quality.
Segmented subscribers using a recency/frequency/monetary approach, producing four groups with very different renewal behaviour and lifetime value, from Founding Subscribers averaging 3.3 orders to Tried It Once subscribers who never came back.
For each segment, defined what a CRM team should actually do differently: which flow to put them in, what the message should be, and what success looks like for that group.
A Chart.js dashboard built from the synthetic dataset, the kind of view I'd build in Looker Studio or Power BI for a real subscription brand.
These are modelled, illustrative figures from the synthetic dataset, not Ritual's real results, intended to show the kind of impact a segmentation-led CRM programme could plausibly have.
Via RFM-style analysis of 1,170 modelled subscribers
37.6% (Creator Micro) vs. 14.2% (Paid Social)
From rebalancing acquisition toward higher-renewal channels
Loyalty, habit-building, renewal education & win-back
The channel bringing in the most subscribers (Paid Social, 460 of 1,170) had the worst renewal rate. A volume-first view of acquisition would have missed this entirely. The segmentation only became useful once renewal was the headline metric.
The segmentation itself is a starting point. The value came from forcing a concrete answer to "what does the CRM team do differently for this group", for each of the four segments.
Every figure in this project is modelled. The goal wasn't to claim a specific revenue outcome. It was to show the analytical approach (renewal-rate benchmarking, channel-level CAC/ROAS, RFM segmentation) I'd bring to a real dataset.
This dataset treats "renewal" as binary. A real subscription business has pauses, skips and plan changes. Modelling those states would make the segmentation (especially "Paused & Lapsed") more realistic.