Task 1 — Exploratory Data Analysis
Task 1 · Exploratory Data Analysis
Dataset Structure
| Table | Rows | Grain |
|---|---|---|
club_dim | 34,953 | One row per customer (clubhouse) |
order_dim | 44,080 | One row per order |
order_fct | 86,303 | One row per product line item |
order_fct is at line-item grain — a single order creates 2–3 rows (Camera + Base + optionally Add-on).
The New Customer / Legacy Account Split
Critical context: You cannot buy a subscription without a camera. Therefore, any
CreateSubscriptionorder that contains no camera line item belongs to an existing club whose original camera purchase predates this dataset.
Loading...
No Results
Applying the new-customer lens:
| Bundle | Customer type | % |
|---|---|---|
| Camera + Sub + Add-on | ✅ New customer, full bundle | 19.8% |
| Camera + Sub | ✅ New customer, no add-on | 19.0% |
| Camera only | ✅ New customer (or replacement camera) | 2.8% |
| Sub only | 🔄 Legacy account — camera predates dataset | 33.5% |
| Sub + Add-on | 🔄 Legacy account expanding features | 13.7% |
| Add-on only | 🔄 Legacy account adding a feature | 10.7% |
~41.6% of all orders relate to new customer acquisition. ~58.4% relate to managing the existing base.
Product Category Mix
No Results
Camera rows are always null MRR — hardware doesn't generate recurring revenue. Camera revenue is tracked separately via
unit_price × unit_quantity.
Data Quality
No Results
Revised interpretation of data quality issues
| Issue | Revised understanding |
|---|---|
| Camera MRR = 100% null | Correct by design — hardware, not subscription |
| 75% null subscription length | Month-to-month subs likely not assigned a term. The most committed cohort (annual) is visible (19%); the largest cohort (monthly) is invisible |
| 11.4% unknown channel | Can't attribute these to ecommerce or direct — ecommerce share is understated |
| 33.5% sub-only CreateSub orders | NOT a data quality issue — these are real legacy customers whose camera predates the dataset |
| 0 orphaned order_dim rows | All order_fct orders have a matching order_dim record — join is clean |
| 321 orphaned club rows (0.37%) | clubhouse_id in order_fct with no club_dim record — minor pipeline issue |
What this data can and cannot answer
| Question | Feasible? | Gap |
|---|---|---|
| How many true new customers were acquired per month? | ✅ ~674–1,760/mo (camera orders) | — |
| What % of new customers buy an add-on at checkout? | ✅ ~48–50% | No plan tier to exclude Starter |
| Which markets acquire the most new customers? | ✅ | No sport type or season date |
| How many legacy accounts are active (vs churned)? | ❌ | No subscription state / churn events |
| What is customer lifetime value? | ❌ | No renewal or cancellation data |
| What is the Starter → Team upgrade rate? | ❌ | No plan tier in data |
| When did legacy accounts originally purchase? | ❌ | Camera purchase predates dataset |