What you'll take away
- A forecast must answer six questions before it drives production or cash decisions.
- Forecast at SKU and channel level, not blended. Total-company forecasts hide every stockout and every over-order.
- Safety stock formula: SS = (Max daily sales × Max lead time) − (Avg daily sales × Avg lead time). Not sophisticated, but reliable.
- Track forecast accuracy monthly. MAPE (Mean Absolute Percent Error) 15–25% is normal for growing brands; below 10% is unusual; above 40% means the forecast is not being used.
- S&OP is a rhythm, not software. Monthly meeting, shared file, decisions logged, actions closed. Enterprise systems come later.
Evidence note: Linked external claims point to the named primary source where available. Unattributed cost, margin, timing, and operating ranges are Logic Agency planning benchmarks drawn from client work, supplier quotes, and operating models. They are directional, not guaranteed market averages; validate them against current contracts, quotes, regulations, and channel requirements.
Who this is for: Founders, ops leads, finance leads, and heads of supply chain at CPG brands from $2M to $50M who need to move from spreadsheet-and-argument forecasting to a repeatable operating rhythm.
What's Inside
Why forecasting breaks as brands scale
Early-stage brands run on intuition because the founder is close to every order. That stops working when sales split across DTC, wholesale, marketplaces, distributors, and retail. Each channel has different timing, different order behavior, different data quality, and different lead-time consequences.
The common failure is a single annual number treated as a plan. "We're going to do $8M next year" is a revenue target, not a forecast. Operations needs a rolling view by SKU, channel, customer, and month, with a clear distinction between what has been ordered, what is expected, and what is merely hoped for.
The three signals your forecast has already broken. (1) You had a stockout on a top-10 SKU in the last 60 days. (2) You wrote off inventory in the last 12 months that was not intentional promotional over-supply. (3) Your co-manufacturer or purchasing team is placing orders based on their guess of your demand, not your plan.
The six questions a forecast must answer
Before you build the forecast, agree what it has to answer. Below is the shortest useful list.
These six questions are the entire job. Every planning meeting reduces to one of them. Open each card to see how the answer should be structured.
Question 01How much do we expect to sell?
By SKU, by channel, by month, for the next 12 months. Rolling. Updated monthly at minimum.
Split the number into three lanes:
- Committed — POs in hand or shipping in the next 60 days.
- Expected — run-rate reorders + confirmed promotions with a signed calendar.
- Aspirational — pipeline, unconfirmed distributor conversations, launches without a signed retailer PO.
Never blend the three lanes into a single number. That is how a "$12M year" becomes a $6M year with $2M of dead stock in a warehouse.
Question 02What is our target inventory position?
By SKU, by DC, at end of each month. Sufficient to cover expected sales plus safety stock through the next replenishment cycle.
The forecast tells you what target to hit; the safety-stock formula (next section) tells you how deep the buffer needs to be.
Question 03What production orders do we need to place, and when?
Back-scheduled from expected demand, minus current inventory, minus in-transit, plus safety stock, respecting production lead time and MOQ.
This is the number that changes the co-manufacturer's calendar. Communicate it before you need it — reactive PO placement burns capacity that could have been reserved earlier.
Question 04What working capital do we need, and when?
Cash out for production and freight vs cash in from receivables. The demand forecast is the primary input to the cash forecast.
A common mistake: forecast production against demand, then discover cash constraints at the deposit-invoice stage. Cash-align the plan at the same table you build the demand plan — not in a separate finance meeting a month later.
Question 05What are the constraints and risks?
Supplier capacity, ingredient availability, freight windows, retailer PO timing, seasonal peaks. Named, dated, ranked.
A risk without a date is a worry. A risk with a date and a threshold is an operating trigger.
Question 06What changed since last forecast, and why?
Every update tracks what moved. The learning is in the delta.
The five most useful questions in a monthly S&OP review are: what did we assume, what did actuals do, why the delta, what does that tell us about next month, and what do we change?
Safety stock: the formula and worked examples
Safety stock is the buffer inventory you carry between planned replenishments to cover demand variability and lead-time variability. Too little and you stock out; too much and you tie up cash.
Safety stock formula (max-max minus avg-avg)
This is the operating standard for growing CPG brands. More sophisticated statistical models exist (service-level based, using standard deviation), but the max-max minus avg-avg formula is reliable, defensible, and matches how retail buyers actually think about coverage.
Reorder Point = (Avg Daily Sales × Avg Lead Time) + Safety Stock
Days of Supply = Current On-Hand ÷ Avg Daily Sales
Worked example: 8oz supplement bottle
Safety stock calculation — worked example
This brand carries 70 days of safety stock — expensive, but justified by the 58-day worst-case lead time and the 80% variability between average and peak daily sales. Cut lead time to 45 days or reduce peak variability and safety stock drops meaningfully.
Building a forecast people can operate
The forecast people will actually use has a specific shape: SKU-level, channel-level, monthly-cadence, versioned, with base/upside/downside cases.
Start with the demand signal
Separate shipped orders, retailer POs in hand, subscriptions, expected promotions, pipeline, and one-time events. Do not blend committed demand with a sales wish list — they compound errors differently.
Forecast at SKU and channel level
A total-company forecast hides the stockout. Plan the items that actually drive service levels, margin, and production constraints — usually the top 20% of SKUs producing 70–80% of revenue.
Add the timing reality
Include production lead time, component lead time, ocean or domestic freight, receiving, QC, and retailer appointment windows. Forecasts denominated in "months" get translated back into "when do I place the PO" — build both.
Show a range
Base, upside, downside. The point is not false precision. It is making the cash and capacity implications of uncertainty visible. Board and investor discussions get more honest when the range is visible.
Log the assumptions
Every forecast update has a change note: "Moved SKU-7 Q4 up 20% based on Target reset confirmation." Assumptions logged make the forecast auditable and the lessons compoundable.
The monthly S&OP rhythm
S&OP (Sales and Operations Planning) does not require enterprise software or a full-time planner. It requires a recurring conversation, a shared file, a discipline of decisions logged and actions closed.
| Cadence | Meeting | Attendees | Output |
|---|---|---|---|
| Weekly | Ops standup | Ops lead, purchasing, warehouse/3PL rep | Inventory positions, in-transit, exceptions, immediate actions |
| Monthly | Demand review | Sales lead, marketing lead, ops lead | Updated 12-month forecast, promotional calendar, retailer PO changes |
| Monthly | Supply review | Ops lead, purchasing, quality | Production plan vs demand, supplier constraints, capacity issues, risks |
| Monthly | Executive S&OP | Founder / CEO, CFO, ops lead, sales lead | Decisions on trade-offs, cash implications, capacity investments, plan approval |
| Quarterly | Forecast accuracy review | Ops lead, sales lead, finance | MAPE by SKU, root cause of largest misses, forecast process changes |
The output of the executive S&OP
- One approved 12-month forecast (base case), with logged upside/downside
- One approved production plan matching the forecast
- One approved cash plan matching the production plan
- A list of open decisions and owners
- A short change note: what moved and why
The forecast has one owner, one number, one file
The most common S&OP failure is having three versions of the forecast in circulation: the sales forecast, the finance forecast, and the ops forecast. Different assumptions, different numbers, different actions. Pick one version, one owner, and one shared location. Everyone edits from that.
Running S&OP with three versions of the forecast? The rhythm is what turns the argument into a decision. We build and run it as part of Active Management retainers.
Forecast accuracy: what to track and why
A forecast you do not measure is a forecast you cannot improve. The industry-standard metric is MAPE (Mean Absolute Percent Error). Below is how to use it.
MAPE (Mean Absolute Percent Error) — SKU-level
Calculate per SKU per month, then average across SKUs weighted by revenue for a portfolio view.
Bias metric: (Actual − Forecast) ÷ Actual, signed. Consistent negative bias means chronic over-forecast; consistent positive bias means chronic under-forecast.
| MAPE range | Interpretation | Typical action |
|---|---|---|
| Under 10% | Excellent — usually only achievable on high-velocity mature SKUs | Sustain; use as benchmark for other SKUs |
| 10–20% | Good — expected for scaling CPG brands with 12+ months history | Sustain; iterate on largest misses |
| 20–35% | Acceptable — typical for new SKUs or seasonal categories | Track root causes; expected to improve with data |
| 35–50% | Poor — forecast is losing predictive power | Root-cause session; likely process or data-source problem |
| Over 50% | Broken — forecast is not being used or is being systematically overridden | Investigate whether S&OP exists in practice or only on paper |
The bias-and-variance diagnostic. MAPE alone can hide two very different problems: high variance (wide swings around the actual, both positive and negative) and high bias (chronic over- or under-forecast). Track both. A brand with 25% MAPE and no bias has a variance problem — usually solved by better data. A brand with 25% MAPE and negative bias (over-forecasting consistently) has a discipline problem — usually solved by removing sales-driven optimism from the forecast.
When to move off spreadsheets
Every brand starts on spreadsheets. Most stay on spreadsheets longer than they should. Below is the honest read on when to move.
Spreadsheet forecasting (SKU count up to 30, channels up to 3)
Google Sheet or Excel with the SKU-level forecast, S&OP rhythm, and version control. Works well if the discipline is real. Fails when the file becomes the constraint.
Purpose-built demand planning tool
Cogsy, Genlogs, Streamline, or Netstock. Purpose-built for CPG. Handles multi-channel forecasting, safety stock, reorder points, and integrates with Shopify/QuickBooks/NetSuite. Monthly cost $200–2,000.
ERP-integrated planning
NetSuite Advanced Inventory, Cin7 Omni, Acumatica, or SAP Business One. Higher cost, more integration, required at scale where multi-facility, multi-currency, and multi-entity complexity exceeds spreadsheet capability.
When Logic gets involved
A 90-day inventory plan that shortened order-to-cash
On Epicutis, we built a 90-day rolling SKU-level inventory plan alongside a monthly S&OP rhythm as the brand scaled from 3 to 21+ SKUs. The forecast discipline shortened order-to-cash time meaningfully and gave the CFO a cash forecast tied directly to production and shipment schedule.
The mechanism was not sophisticated software. It was a shared file, a monthly meeting, safety stock and reorder points per SKU, and change notes on every update.
Demand-forecasting engagements typically start when a brand has had a recent stockout, a recent inventory write-off, an investor asking why cash is tight, or a co-manufacturer asking for a rolling forecast.
What we build inside the engagement:
- SKU-level 12-month rolling forecast by channel, versioned and shared
- Safety stock and reorder-point calculations per SKU
- Production plan back-scheduled from demand, respecting lead times and MOQs
- S&OP rhythm — monthly executive review with logged decisions
- MAPE tracking dashboard with quarterly root-cause reviews
- Cash-forecast integration with the demand plan
- Tool recommendation if you're ready to move off spreadsheets
The operator's summary
- A forecast is an operating assumption, not a prediction. Its value is in the decisions it enables.
- Forecast at SKU and channel level, monthly, with base/upside/downside cases and logged assumptions.
- Safety stock = (Max daily sales × Max lead time) − (Avg daily sales × Avg lead time). Reorder point = (Avg × Avg) + Safety stock.
- S&OP is a rhythm, not software. Monthly meeting, one forecast, one owner, one file, decisions logged.
- Track MAPE monthly. 15–25% is typical for scaling brands. Above 35% means the process needs work.