Buyer’s guide · 2026
Quick answer: Subscription-box demand forecasting is really two forecasts multiplied together: a subscriber curve (renewals + new sign-ups − churn) and a box composition (which component SKUs go in next month’s box). No single mainstream tool models both natively — so the stack that works is your subscription platform’s analytics for the subscriber curve, a composition sheet that explodes boxes into component quantities, and a per-SKU forecasting tool to time the purchase orders against each component’s lead time. The subscriber curve is the glamorous part; the late-arriving component is what actually ships boxes late.
Subscription boxes look like the easiest forecasting problem in ecommerce — you literally know your customers in advance. Then month three arrives, one of the eight items in the box is stuck at a supplier, and you learn the real shape of the problem: a box is a bill of materials, and the box ships at the speed of its slowest component.
The two-layer forecast
Component demand = projected subscribers × units of that SKU per box(+ your one-off shop & marketplace sales of the same SKU)
Layer 1: The subscriber curve
Next cycle’s box count = current subscribers + expected new sign-ups − expected churn, with pauses and skips as their own line (a paused subscriber isn’t churned, but they don’t need a box this month either). Your subscription platform’s analytics already track these rates — that’s the right source for this layer. The discipline is forecasting the curve per cohort, because month-two churn and month-twelve churn are different animals.
Layer 2: The composition explosion
Once the curve says “4,800 boxes in September,” the demand for components is nearly deterministic: 4,800 of each single-unit item, 9,600 of anything that goes in twice. This layer is a spreadsheet, and honestly a spreadsheet is fine for it — the value isn’t sophistication, it’s doing the explosion early enough that layer 3 can act on it.
Layer 3: Per-component purchase timing
Each component has its own supplier, lead time, and MOQ — and this is where box businesses actually miss ship dates. A 90-day-lead-time item for the November box must be ordered in July; the artisan snack with a 2-week lead can wait until October. This layer is standard per-SKU replenishment planning: reorder points from lead times, safety stock sized to supplier variability, inbound tracked against the box deadline.
Subscribers are a curve. Components are a deadline. Most stockouts live in the gap between the two.
The tool landscape, by job
| Job | Tool class | Notes |
|---|---|---|
| Subscriber curve (churn, cohorts, pauses) | Your subscription platform’s analytics (Recharge, Loop, Skio on Shopify; or your box platform’s reporting) | Closest to the billing truth; use its numbers, don’t re-derive them |
| Box composition → component explosion | Spreadsheet (or your platform’s kit/bundle report) | Simple multiplication; keep it one sheet per upcoming box |
| Component replenishment & PO timing | Per-SKU forecasting tool (SKU Compass, Inventory Planner, SoStocked) | Lead times, reorder points, safety stock, PO tracking per component |
| One-off shop + marketplace sales of the same SKUs | Same forecasting tool, per channel | Past-box sales, add-ons, and Amazon listings pull from the same stock pool — forecast them per channel or the box math silently breaks |
The honest caveat: we’re in that third row, so weigh the source — and know what we’re not claiming. SKU Compass doesn’t model churn cohorts, and it won’t explode a box recipe into components for you; the subscription platform and the composition sheet own those. What it does is the part spreadsheets do worst: per-SKU, per-channel velocity, days of supply, and reorder points off real lead times, so the slow component surfaces in July instead of the week the November box packs out. If your whole catalog is the box and every component shares one supplier and one lead time, a good spreadsheet may genuinely be enough for a while.
Three failure modes to design against
1. The curve eats the calendar. Teams argue about whether September is 4,600 or 5,000 boxes while the 90-day component for either number goes unordered. Order the long-lead items to the range’s floor plus safety stock; refine the curve later.
2. The one-off shop raids the box stock. Selling past boxes or individual items from the same pool that next month’s boxes need is a silent double-count — per-channel visibility is what catches it.
3. Growth spikes get treated like baseline. A viral month bends the subscriber curve the way Prime Day bends velocity — verify the retention of the spike cohort before you re-rate every component buy upward.
Put the component layer on rails
Connect your Shopify and marketplace data and SKU Compass tracks per-SKU, per-channel velocity, days of supply, and reorder points off your real lead times — so the slow component shows up months before it’s a late box. Free for 30 days, no credit card.
Frequently asked questions
How do you forecast demand for a subscription box?
In two layers: forecast the subscriber count (current subscribers + sign-ups − churn, with pauses tracked separately), then multiply by the box composition to get component-SKU demand. Add one-off and marketplace sales of the same SKUs, then time each component’s purchase order against its own lead time.
What software do subscription boxes use for forecasting?
Typically a stack rather than one tool: the subscription platform’s analytics (Recharge, Loop, Skio, or the box platform’s own reporting) for subscriber and churn curves, a spreadsheet for the box-to-component explosion, and a per-SKU forecasting tool such as SKU Compass, Inventory Planner, or SoStocked for component replenishment and PO timing.
Why do subscription boxes stock out if demand is predictable?
Because the box ships at the speed of its slowest component. Subscriber counts are predictable weeks ahead, but each component has its own supplier lead time and MOQ — the 90-day item for November has to be ordered in July. Most misses are purchase-timing failures, not demand-forecast failures.
How should I handle paused subscribers in the forecast?
As their own line, not as churn. Paused subscribers don’t need this month’s box but return at a measurable rate — folding them into churn understates future demand, while counting them as active overstates this month’s. Your platform’s pause-rate report is the input to use.
Does SKU Compass work for subscription-box businesses?
For the component layer, yes: per-SKU, per-channel velocity, days of supply, reorder points from your own lead times, and purchase-order tracking. It doesn’t model churn cohorts or explode box recipes — pair it with your subscription platform’s analytics and a composition sheet.
How far ahead should a box business order inventory?
Per component, one full lead time plus safety margin before the box’s pack-out date — which commonly means ordering long-lead items two to three boxes ahead. Order long-lead components to the low end of your subscriber range plus safety stock, and refine the count closer in.
Related reading:
Supplier lead time: the input most sellers never tune ·
How to forecast inventory demand ·
What is a reorder point? ·
Best inventory forecasting software (2026)