ytpartners case study.
Data system and KPI engine
The scale plan rested on numbers that only existed in raw transaction ledgers. I turned them into a KPI dashboard with LTV, cohort reporting that went on to shape product defaults and offer design, and a churn model that flags customers before they leave rather than after.
Executive summary
Before changing strategy, distribution, or digital systems, the business needed clear answers: which customers repeat, when churn happens, what drives reorder behavior, and how those realities translate into product defaults and marketing motions. I built an analytics engine that produced consistent outputs leadership could use weekly.
How the system works
A live engine: ledger in, decision outputs out, run weekly.
Starting point and diagnosis
The constraint was visibility: retention and churn dynamics were discussed, not staged and operationalized.
- Transaction history existed, but reporting was not systematized into weekly outputs
- No stable cohort views to guide offer design and defaults
- Churn was not staged by behavioral signals
- No repeatable insight extraction method tied to owned actions
KPI dashboard sample
Illustrative weekly view. Percentages shown, no customer counts.
| Area | KPI | This week | WoW | Target band | Owner |
|---|---|---|---|---|---|
| Revenue | Total revenue (index) | 100 | +6% | 95-115 | GM |
| Customers | Returning customers | 68% | +2 pts | 65-75% | Growth |
| Retention | Repeat rate | 44% | +3 pts | 40-50% | Lifecycle |
| Unit economics | LTV (index) | 112 | +4% | 100-120 | Finance |
| Quality | Support tickets per 100 orders | 9 | -2 | 6-10 | Ops |
Values are illustrative to show structure and readability.
Primary plan cohorts
I focused the strategy on the cohorts that actually repeat. Names are anonymized.
| Cohort | Description | % of customers | Repeat signal | Default implication |
|---|---|---|---|---|
| Plan A: Custom | High personalization, stable repeat | 28% | Strong | Default for high-intent buyers |
| Plan B: Pro | Performance baseline plan | 22% | Strong | Default recommendation for new buyers |
| Plan C: Lean | Lightweight, price-sensitive | 18% | Mixed | Offer as secondary option |
| Plan D: Signature | Chef's selection, variety-led | 20% | Mixed | Use as upgrade path |
| Program E: Reset | Short-duration, high drop-off risk | 12% | Weak | Gate behind education + onboarding |
Churn risk and mitigation
Elastic, per-customer churn staging driven by order-gap signals.
- Per-customer monitoring of order gaps relative to historical cadence
- Stage classification driven by observable signals: healthy, drifting, at-risk, churned
- Mitigation motions mapped to stage with owners and success criteria
What changed
- Retention and churn became measurable and staged
- Product defaults were set by cohorts that actually repeat
- Lifecycle motions became trigger-based instead of generic outreach
- Weekly reporting converted into owned actions
Assets delivered
- KPI dashboard pack and KPI definitions
- Cohort framework and repeat behavior summaries
- Churn risk bands and mitigation prompts by stage
- Automated refresh workflow from raw ledger exports
Outcomes
- A single source of truth for weekly performance
- Faster diagnosis of churn and repeat behavior
- Clearer underwritable story for strategy and capital readiness
- Decision system the team can run consistently
Applied AI in execution systems
- Automated KPI pack generation from ledger exports
- Automated churn monitoring report and mitigation prompts
- Automated cohort refresh and switching summaries
- Structured insight extraction into weekly action items