Why this matters
An objective with no tree beneath it produces a dashboard of disconnected numbers. The KPI tree framework fixes this by stating the arithmetic between levels, so a change at the top has to be explained by a change lower down. The objectives and KPIs guide supplies the other half: which of those numbers marketing should be judged on, and which are diagnostics. A subscription tree adds one feature an online shop's lacks: customers leave as well as arrive.
The objective
Everything about Boxwell is invented for the exercise.
Boxwell: the objective and last month's figures
Boxwell sends a monthly box of loose-leaf teas to subscribers in the UK. The founders' objective for the year is to grow monthly recurring revenue (MRR) from £72,000 to £95,000 by December, without discounting the core plan.
Last month:
| Measure | Value |
|---|---|
| Active subscribers at the start of the month | 4,000 |
| Average revenue per subscriber per month | £18.00 |
| Site sessions | 60,000 |
| Subscribe rate (new subscribers ÷ sessions) | 0.75% |
| New subscribers | 450 |
| Monthly cancellation rate (of start-of-month subscribers) | 8% |
| Cancellations | 320, of which 150 were in their first two months |
| Sessions by channel | Organic search 24,000; paid social 18,000; paid search 9,000; email and referral 6,000; direct 3,000 |
| Plans | Monthly plan £20; annual plan paid upfront, equivalent to £15 a month |
| Contribution margin per box | 40% |
| Blended cost to acquire a subscriber | £38 |
| Finance's ceiling on cost to acquire a subscriber | £45 |
Product, pricing and delivery are run by the founders and an operations lead; marketing is one marketer and a freelance designer.
Building the tree
Start from the outcome and split it by arithmetic, not by what seems related. MRR divides multiplicatively into subscribers and revenue per subscriber. Subscribers at the end of the month divide additively: those at the start, plus new, minus cancelled. Keep splitting each branch until you reach inputs a named person can change this week, and check that every multiplicative split is measured on the same population and period.
A subscription tree has two kinds of input that marketing touches. Acquisition inputs, sessions by channel and the subscribe rate, are mostly marketing's. Retention inputs are shared: onboarding emails are marketing's, but the tea, the packaging and the delivery are not. Mark ownership honestly, because a tree that gives marketing every leaf will be used to blame marketing for every miss.
Two economic measures sit beside the tree rather than in it: lifetime value (LTVTotal value a customer brings over time.) and the cost to acquire a subscriber (CACCost to acquire one new customer.). They are guardrails, telling you whether growth in the acquisition branch is worth having.
Task
- Write the tree from MRR down to channel sessions, with the formula beside each split.
- Check that the arithmetic balances on last month's figures and produces the closing MRR.
- Split cancellations into early and established subscribers and explain why.
- Mark each input as owned by marketing, shared, or owned elsewhere.
- Add at least two guardrails with thresholds.
- Estimate what a realistic change in one marketing input and one shared input would do to MRR.
- Name the one KPI you would report weekly and justify it.
Worked answer
Boxwell's KPI tree and the arithmetic check
| Level | Node | Formula | Owner |
|---|---|---|---|
| Outcome | MRR | Active subscribers × average revenue per subscriber | The business |
| Driver | Active subscribers (end of month) | Start + new − cancelled | Shared |
| Driver | Average revenue per subscriber | Weighted by plan mix and add-ons | Founders (pricing); marketing (plan promotion) |
| Driver | New subscribers | Sessions × subscribe rate | Marketing |
| Input | Sessions by channel | Organic + paid social + paid search + email and referral + direct | Marketing |
| Input | Subscribe rate | By landing page and channel | Marketing |
| Driver | Cancellations | Early cancellations + established cancellations | Shared |
| Input | Early cancellation rate | Onboarding emails, first-box experience | Shared: marketing and operations |
| Input | Established cancellation rate | Product range, delivery, price | Operations and founders |
Check: 60,000 sessions × 0.75% = 450 new subscribers. 4,000 × 8% = 320 cancellations. End of month: 4,000 + 450 − 320 = 4,130 subscribers. MRR: 4,130 × £18.00 = £74,340. The tree balances.
Sensitivity, one input at a time:
- Subscribe rate from 0.75% to 0.85% through a landing page test: 60,000 × 0.85% = 510 new, 60 more subscribers, about £1,080 more MRR.
- Early cancellations down by a third through a new onboarding sequence: 150 to 100, 50 more subscribers, about £900 more MRR, and the gain compounds because those subscribers keep paying.
Guardrail economics: lifetime value ≈ £18.00 × 40% ÷ 8% = £90 per subscriber, against a £38 acquisition cost and a £45 ceiling.
Why split cancellations. Early leavers are often people the acquisition message mismatched or the first box disappointed, which marketing can act on. Established leavers are judging the product. One combined rate hides which problem you have.
Ownership. Acquisition inputs are marketing's; early cancellation is shared with operations; price, plans and product sit elsewhere. Notice that promoting the annual plan lowers revenue per subscriber in the tree while probably lowering cancellations; the tree makes that trade-off visible rather than settling it.
Guardrails. Cost to acquire a subscriber stays at or below finance's ceiling; refunds and complaints stay at or below last quarter's level; the share of new subscribers arriving through discount codes is capped, because the objective rules out discounting the core plan.
The weekly KPI. Report new subscribers who are still active after their first box, against the acquisition-cost ceiling. It moves weekly, marketing can act on it, and it joins the acquisition branch to the early-cancellation branch, so it cannot be hit by buying sign-ups who leave. MRR itself is reported monthly.
Common mistakes
- A tree that does not balance. If the figures beneath a node do not reproduce it, the tree is a diagram, not a tool.
- Giving marketing every leaf. Owners for price, product and delivery belong in the tree too.
- A weekly KPI nobody can move in a week. MRR and lifetime value belong in the monthly view.
The measurement plan guide turns this tree into events and reports, and the budget allocation guide uses the acquisition branch to decide where the next pound goes. Both feed the quarterly marketing plan capstone; the intermediate level overview shows where this sits.
