Subscription billing is the first and it needs to be native rather than improvised. What you need is recurring charges handled automatically, plans and pauses, and above all a process for failed payments, because cards expire and fail continuously and a meaningful share of subscriber loss is involuntary rather than a decision to cancel. Automatic retry logic and a sequence of messages recovering failed payments is not a refinement, it is a substantial share of your retention.
Churn is the metric that determines whether the business works, and it needs to be visible monthly from the start. A subscription business with strong acquisition and high churn is filling a bucket with a hole, and the arithmetic gets worse as it grows because retaining a larger base requires more effort. Knowing the rate, and knowing whether it concentrates in the first month or later, tells you whether the problem is the promise or the product.
Forecasting demand is harder here than in ordinary retail because you must buy inventory for a box before knowing the final subscriber count, and lead times on custom items can be months. The cutoff date, when your subscriber number is fixed for that month, is the operational hinge of the whole business, and everything upstream needs to work backward from it. Over ordering ties up cash in items you cannot sell separately, and under ordering means failing subscribers who paid.
Fulfilment at volume needs deciding before you reach it. Packing thirty boxes yourself is an evening. Packing four hundred is not, and the transition to a fulfilment partner takes time to arrange, which means the decision should be made before the month it becomes necessary. Shipping cost per box is also a margin item worth modelling carefully, since it frequently exceeds the cost of the contents.
Communication is the retention lever that costs least. Subscribers who know what is coming, when it shipped, and what is inside stay longer than those who receive an unexplained parcel, and the messages around each shipment are the main relationship you have with somebody you never meet.
Give subscribers a way to pause rather than only to cancel, because a substantial share of cancellations are about timing rather than dissatisfaction. Somebody travelling for two months, or managing a tight month financially, will cancel if that is the only option and frequently will not return. A pause converts a permanent loss into a temporary one, and it costs nothing beyond making the option available and visible.