S&OP Best Practices for Mid-Market Manufacturers

Most mid-market S&OP processes are theater. A demand meeting nobody trusts, a supply review that's really a capacity argument, and an executive meeting where the CFO overrides the consensus number with a gut call. The S&OP best practices that actually move forecast accuracy and free up stranded cash aren't about buying software first. They're about running a disciplined monthly cadence, owning a single set of numbers, and forcing the trade-offs out into the open where leadership can decide them. I ran this at a $250M industrial manufacturer. Here's what worked and what was noise.

What good S&OP actually produces

S&OP exists to do one thing: align demand, supply, and finance on a single 18-month plan, then update it every month with discipline. If your process doesn't end with a number that production, procurement, and the CFO all commit to, you don't have S&OP. You have meetings.

Three outputs prove it's working:

• One demand plan in units AND dollars. If the demand team plans units and finance plans revenue separately, you have two forecasts and zero accountability.
• A supply plan that says yes, no, or yes-but. Capacity, materials, and labor reconciled against the demand plan with named constraints.
• A gap-closing decision. Where demand exceeds supply or the revenue plan misses budget, the executive review picks the action. Build inventory ahead, add a shift, expedite materials, or cut the forecast.

The five-meeting monthly cadence

The classic five-step cadence still beats every shortcut I've seen. Run it on a fixed calendar so people plan their month around it.

The pre-S&OP is the meeting most mid-market teams skip, and it's the one that makes the executive review short. Walk into the exec meeting with three scenarios and a recommendation, not raw data. If your executive S&OP runs longer than 90 minutes, your pre-S&OP failed.

Metrics: track these four, ignore the vanity ones

• Forecast accuracy (or MAPE/WMAPE) by family. Measure at the level you plan, not at the SKU-DC level where noise drowns signal. A $250M business should know its WMAPE at family level cold. Mid-market manufacturers typically land at 50-65% item-level accuracy; the best hit 75%+ at family level.
• Forecast bias. More dangerous than error. Persistent over-forecasting inflates inventory; under-forecasting starves service. If bias runs one direction for three months, someone's sandbagging or hoping.
• Plan attainment. Did supply actually deliver the committed plan? If attainment is 70%, your supply review is fiction.
• Inventory vs. plan, split by purpose. Cycle, safety, and excess/obsolete (E&O) tracked separately. Stranded inventory hides inside a healthy-looking total.

Skip days-of-inventory as a headline metric. It moves for reasons that have nothing to do with planning quality.

The mistakes that kill mid-market S&OP

Consensus by averaging. Sales says 100, ops says 80, so the plan is 90. That's not consensus, it's splitting the difference to avoid a fight. Real consensus means surfacing the assumptions behind each number and choosing one.

No single source of demand. Sales has a CRM pipeline, marketing has a campaign forecast, finance has a budget, planning has a statistical baseline. Four numbers, four owners, nobody wrong. Pick one demand plan and make everything else reconcile to it.

The CFO override. When the executive meeting routinely throws out the consensus for a top-down target, the process dies in six months because everyone learns it doesn't matter. The fix: let finance set the revenue target as a gap, then make S&OP own closing it with named actions.

Planning the whole catalog the same way. A 4,000-SKU manufacturer can't forecast every item with equal care. Segment by ABC-XYZ. Plan the A/X items (high value, stable) statistically and review them; let the C/Z items (low value, erratic) run on simple rules and reorder points.

Where AI demand forecasting actually helps

Statistical forecasting in spreadsheets caps out fast in the mid-market. AI demand forecasting earns its place when you have enough history and external signals (promotions, pricing, weather, macro indicators) to beat a human-adjusted baseline. The honest test: run the model and the planner in parallel for two quarters and compare WMAPE and bias. If the model wins on the A/X items, automate those and free your planners to work exceptions.

This is where a platform like Pigment changes the economics. Demand, supply, and finance plan on the same model, scenarios run in seconds instead of overnight, and the units-to-dollars translation happens automatically. The point isn't the tool. It's that the consensus number stops living in twelve disconnected spreadsheets.

A 90-day rollout that sticks

1. Days 1-30: Fix the data and definitions. Agree on family hierarchy, planning level, and the four metrics. Baseline your current accuracy and bias.
2. Days 31-60: Run the five-meeting cadence manually, even in spreadsheets. Build the pre-S&OP discipline first.
3. Days 61-90: Add scenario planning and start the AI-vs-planner bake-off on A/X families.

Don't buy software in month one. Earn the process, then automate it.

See where your process and inventory actually stand

We run a free planning-maturity and stranded-inventory teardown for mid-market manufacturers. You'll get a benchmark of your S&OP cadence against the five-step model, a read on your forecast bias, and a dollar estimate of cash trapped in excess and obsolete inventory. Book a call and we'll walk your numbers together, no slideware.