How to Build a Demand Planning Process from Scratch

If you've just been tasked with "fixing the forecast" or "starting S&OP," you're likely staring at a blank page—or worse, a folder full of broken spreadsheets and a sales team that refuses to talk to operations.

Building a demand planning process from scratch is less about choosing the perfect algorithm and more about change management. It's about moving an organization from reactive firefighting to proactive planning.

This guide outlines a practical, 8-step framework to build a demand planning process that works. We'll skip the academic theory and focus on the "crawl, walk, run" approach that actually survives first contact with reality.

Step 1: Define Your "Demand Planning Universe" (Scope)

Before you calculate a single number, you need to define the boundaries of your process. A common mistake is trying to forecast everything at every level immediately.

You need to answer three questions:

1. Granularity: Will you forecast at the SKU level, product family level, or brand level? (Tip: Start at the family level for consensus, disaggregate to SKU for execution).
2. Horizon: How far out do you need to look? 12 months is standard for financial budgeting; 18 months is often needed for long-lead-time supply chain planning.
3. Frequency: Will you update the plan monthly or weekly? For most businesses starting out, a monthly cadence is the right balance between responsiveness and stability.

Step 2: Identify Stakeholders and Build the RACI

Demand planning is a cross-functional sport. If you build the forecast in a silo, it will be ignored. You need to identify who provides the inputs and who relies on the outputs.

Use this RACI (Responsible, Accountable, Consulted, Informed) matrix to clarify roles:

| Activity | Demand Planner | Sales | Finance | Operations | | :--- | :--- | :--- | :--- | :--- | | Provide sales forecast input | I | R | C | I | | Generate statistical baseline | R | I | I | I | | Review forecast assumptions | R | C | C | C | | Approve consensus forecast | A | C | C | C | | Measure accuracy | R | I | C | I |

• Responsible: The person doing the work (usually you).
• Accountable: The person with veto power (often Sales or Ops leadership).
• Consulted: People who provide input (Finance for budget targets, Marketing for promos).
• Informed: People who need to know the result (Supply Chain for purchasing).

Step 3: The Data Audit (Garbage In, Garbage Out)

You cannot build a robust process on bad data. Before implementing a model, conduct a data audit:

• Sales History: Do you have at least 24 months of clean shipment or order history?
• Master Data: Are product lifecycles accurately tracked? Do you know which SKUs are active vs. discontinued?
• Outliers: Have you flagged one-time events (like a bulk order or a stockout period) so they don't skew the forecast?

Practitioner Note: Don't wait for perfect data. It doesn't exist. Clean the top 20% of SKUs that drive 80% of your volume and start there.

Step 4: Choose Your Forecasting Approach

When starting from scratch, simplicity beats complexity.

1. Start with a Statistical Baseline: Use a simple method like moving averages or exponential smoothing to generate a "computer number." This removes emotion and political bias from the starting point.
2. Layer on Intelligence: Use human judgment to adjust for things history can't see—new customer wins, upcoming price changes, or competitor actions.

Avoid the temptation to use complex Machine Learning (ML) models on Day 1. If you can't explain why the forecast went up, stakeholders won't trust it.

Step 5: Establish the Planning Cadence

A consistent rhythm is the heartbeat of demand planning. A typical monthly cycle looks like this:

• Day 1-3 (Preparation): Demand planner cleans data and runs the statistical baseline.
• Day 4-10 (Sales/Marketing Input): Sales reps or regional managers review the baseline and add their market intelligence.
• Day 11-15 (Consensus Review): A formal meeting to reconcile the statistical view with the sales view.
• Day 16-20 (Executive Sign-off): Final numbers are published to Supply Chain and Finance.

Step 6: Build the Consensus Process

This is the hardest step. The "Consensus Meeting" is where the tension between Sales (optimism) and Operations (realism) is resolved.

To make these meetings effective:

• Focus on exceptions: Don't review every SKU. Focus on items where the statistical forecast and sales forecast differ significantly.
• Document assumptions: If Sales predicts a 20% bump, write down why (e.g., "New placement in Target"). This allows you to review accuracy later.
• Talk in dollars and units: Sales speaks revenue; Ops speaks units. Your job is to translate.

Step 7: Implement Accuracy Tracking

If you don't measure it, you can't improve it. You need to track how well your process is performing.

Common metrics include:

• WAPE (Weighted Absolute Percentage Error): Good for measuring overall error magnitude.
• Bias: Tells you if you are consistently over-forecasting or under-forecasting.

Deep Dive: Read our guide on Forecast Accuracy Metrics to choose the right KPIs for your business.

Step 8: Iterate and Improve

Your first forecast will be wrong. That's okay. The goal of the process is continuous improvement.

Every month, run a "post-mortem" on the previous month's performance. Did the promo lift materialize? Did the new product launch on time? Use these learnings to refine the parameters for the next cycle.

Common Pitfalls to Avoid

• Tool before Process: Buying expensive software before you have a defined process is a recipe for disaster. You'll just automate chaos.
• Analysis Paralysis: Don't get stuck trying to forecast the tail (C-items). Let the statistical engine handle the low-volume items and focus human effort on the A-items.
• Lack of Sponsorship: If the VP of Sales or COO doesn't back the process, it will turn into a theoretical exercise that nobody follows.

Maturity Model: Crawl, Walk, Run

Don't try to be Amazon on day one. Be realistic about your maturity stage.

| Stage | Forecasting | Process | Technology | | :--- | :--- | :--- | :--- | | Crawl | Simple statistical baseline (Moving Average) | Ad-hoc inputs, heroic effort by one person | Spreadsheets (Excel) | | Walk | Robust statistics + defined seasonality | Monthly consensus meetings, defined RACI | Basic Demand Planning Software | | Run | Machine Learning + Demand Sensing | Integrated Business Planning (IBP), financial integration | Advanced platform with automation |

How Software Accelerates the Process

While you can start with spreadsheets, you will eventually hit a wall. Excel is great for prototyping but terrible for collaboration.

Dedicated software helps by:

• Automating the baseline: freeing you up to talk to sales.
• Handling hierarchy: allowing you to plan at the family level and automatically allocate down to SKUs.
• Visualizing data: making it easier for non-technical stakeholders to see trends.

Conclusion

Building a demand planning process from scratch is a journey. It requires a mix of data science, diplomacy, and persistence. Start small, get a quick win by improving accuracy on your top products, and build credibility over time.

Progress is better than perfection. The best process is the one your team actually uses.

---

Ready to move out of the "Crawl" phase? Schedule a demo to see how DemandPlan can automate your baseline and streamline your consensus process.