For two and a half years I ran operations for a wellness subscription box: a monthly mix of supplements, a tea or functional beverage, and one rotating self-care item. We grew from 180 subscribers to just under 4,000. Somewhere around subscriber 900, I learned the lesson this whole post is about: a subscription business is not an ecommerce business that repeats. It is a manufacturing schedule wearing an ecommerce costume, and the schedule is the product.
A one-time shop can have a slow week and nobody notices. A subscription box that ships late, ships wrong, or double-bills a paused subscriber does not just lose an order. It loses the subscriber, the lifetime value, and often a public review on the way out. Here is the operating system that got us to 4,000 subscribers without a dropped cycle, and the three places we nearly died learning it.
The monthly cycle is a train timetable
Everything in box operations hangs off one repeating skeleton. Ours looked like this for a box shipping in the third week of the month:
- Day 1: Billing runs. Failed payments enter dunning.
- Day 3: Skip and pause window closes. This is the moment your unit count becomes real.
- Day 4: Final unit count locked. Purchase orders for any shortfall are already too late at this point, which is why forecasting happens the month before.
- Days 5 to 10: Kitting. Boxes are assembled, inserts added, lot numbers recorded per batch of boxes.
- Days 11 to 13: Label runs, manifest to the carrier, pallets staged.
- Day 14: Carrier pickup. Tracking emails fire.
- Days 15 to 20: Delivery window, and customer service surge.
Write your version of this down with real dates, every month, and treat each date as a cutoff, not a target. The word “cutoff” matters. A target can slip a day. A cutoff cannot, because six other dates are chained to it. When our kitting slipped two days one August (a late tea shipment), the carrier pickup missed the Friday slot, sat over a weekend, and 4,000 boxes landed four days late. Support tickets tripled. Cancellations that month ran double our average. One slipped cutoff, measurable churn.
Skips, pauses, and address changes: the silent order-droppers
New operators obsess over acquisition and treat subscription state changes as an afterthought. That is exactly backwards. The orders you drop are almost never new orders. They are existing subscribers whose state changed and whose change did not propagate.
The three killers:
Skips after the lock. A subscriber skips on day 5, after the count locked. If your skip flow does not check the cutoff, you now either ship a box they will refund or eat a box you kitted. We put the cutoff date directly in the skip confirmation screen and cut these complaints by more than half.
Address changes mid-cycle. The subscriber moves, updates their address on day 8, but labels were generated from a day 4 export. The box goes to the old address. The fix is boring and absolute: labels generate from a fresh export on label day, never from the count-lock export.
Dunning ghosts. A card fails on day 1, recovers on day 9 after the count locked. Do they get this month’s box or next month’s? Either answer is fine. Having no answer is what generates the angry email. Decide once, write it into your policy, and encode it in your flow.
Lot tracking inside the box
If your box contains supplements, every box batch needs to trace back to the lots inside it. When 4,000 boxes each contain three ingestible products, “which subscribers received lot 24117” must be answerable in minutes. We recorded lot numbers per kitting batch, and each kitting batch mapped to a range of order numbers. Twice in two years a supplier flagged a lot for early expiration reprints, and both times we knew exactly which few hundred subscribers were affected instead of blasting all 4,000 with an alarming email.
This costs you one extra column and thirty seconds per kitting batch. The first time you need it, it repays years of that effort in a single afternoon.
Forecasting: order for the subscribers you will have
Your unit order this month serves next month’s subscriber count, so you are always buying for a number that does not exist yet. Our forecast was unglamorous and worked: trailing three-month growth rate, minus seasonal churn adjustment (January boom, summer slump, the wellness calendar is real), plus a 4 percent overage buffer for damage, reships, and influencer requests.
The overage buffer deserves a note. You will accumulate extra units. Budget for it, and have a planned outlet: flash sales to your list, a one-off “vault box,” bundling into win-back offers. Some box operators quietly move sealed overage through resale marketplaces, and if you go that route it is worth reading how the folks at Market 2 Hands think about cross-marketplace inventory, because listing the same fifty units in three places without tracking is how you oversell stock you no longer have.
The master ops sheet
Every subscription operation I have seen runs, at its core, on one grid: the cycle calendar crossed with the product plan, the unit counts, the lot map, and the vendor deadlines. Ours started as a spreadsheet, and the spreadsheet was genuinely fine until roughly subscriber 2,000. Then it wasn’t. Twelve months of cycles, per-batch lot rows, per-vendor PO tabs, and a couple of years of history put us against the practical ceiling of a consumer spreadsheet, and the person editing the count while I edited the lot map created a real dropped-orders scare.
We eventually moved the whole thing to wisegrid.co, which took an afternoon precisely because it is the same grid and sheet interface the team already knew, just with roughly double the cell capacity we kept bumping into and proper multi-user editing. Viewers are free, so the co-packer and our 3PL contact watch the cycle sheet live without costing a seat, and only the two of us who edit pay for one. I mention it because the migration afternoon was the highest-leverage four hours of that year, not because the tool matters more than the discipline. It does not. A well-run cycle in a spreadsheet beats a sloppy one in anything.
The weekly ops review
Last piece. Once a week, thirty minutes, same three questions: What is the next cutoff and are we ahead of it? What state changes came in this week (skips, pauses, failed payments, address changes) and have they all propagated? What does the forecast say about the PO we need to place this week for next cycle?
That meeting sounds too simple to matter. It is the entire reason we never dropped a cycle. Subscription operations is not about heroics in shipping week. It is about never needing them, because the timetable was honest and somebody looked at it every single week.
Start with the timetable. Write real dates on it today, for the next three cycles. Everything else in this post is just defending those dates.