The biggest misconception in demand planning is that you need sophisticated tools and specialized expertise to forecast demand. You don't. What you need is a systematic approach and the discipline to follow it.
Why Most Brands Overcomplicate Forecasting
Enterprise software vendors have convinced everyone that demand forecasting requires machine learning, AI, and teams of analysts. That might be true if you're Walmart managing millions of SKUs across thousands of stores.
But for a CPG brand with 50-500 SKUs? The fundamentals work just fine.
The brands that struggle with forecasting usually aren't struggling because they lack sophisticated tools. They're struggling because they don't have any consistent process at all. They're making decisions based on gut feel, last month's sales, or whatever number feels right in the moment.
A simple, consistent approach beats a sophisticated approach that never gets implemented.
Start With What You Have: Historical Sales Data
Your sales history is the foundation of any forecast. Before you do anything else, get your data organized.
Pull at Least 12 Months of History
You need enough data to see patterns. Twelve months captures seasonality. Twenty-four months is better if you have it. More than that can include outdated patterns that no longer reflect your business.
Organize by SKU and Channel
Your forecast should be at the level you make decisions. If you order inventory by SKU, forecast by SKU. If different channels have different patterns (DTC vs. wholesale vs. Amazon), separate them.
Clean Out the Noise
Some historical periods don't represent true demand. Stockout periods show artificially low sales—you didn't sell less because demand dropped, you sold less because you ran out. One-time spikes from viral moments or press coverage aren't repeatable.
Flag these anomalies so they don't skew your baseline.
The Simple Moving Average Method
For most products, a simple moving average is a reasonable starting point.
How it works: Average the last N periods (weeks or months) to predict the next period.
Example: Your last 12 weeks of sales for a SKU were: 45, 52, 48, 50, 55, 47, 53, 49, 51, 54, 46, 50
Average: 50 units per week
Your forecast for next week: 50 units
That's it. No algorithms, no software, just arithmetic.
When Moving Averages Work Well
- Products with stable, consistent demand
- SKUs without strong seasonal patterns
- Short-term forecasting (next 4-8 weeks)
When Moving Averages Fall Short
- Highly seasonal products (holiday items, summer goods)
- Products with strong growth or decline trends
- New products without history
Adding Seasonality With Index Factors
If your products have seasonal patterns, you can adjust the moving average with seasonal indices.
Step 1: Calculate Your Baseline Average
Take your average weekly (or monthly) sales across a full year.
Example: Total annual sales: 2,600 units / 52 weeks = 50 units/week average
Step 2: Calculate Seasonal Indices
For each period, divide actual sales by the average to get an index.
Example:
- December sales: 100 units/week
- December index: 100 / 50 = 2.0 (twice the average)
- July sales: 30 units/week
- July index: 30 / 50 = 0.6 (60% of average)
Step 3: Apply Indices to Your Forecast
Multiply your baseline forecast by the seasonal index.
Example:
- Baseline forecast: 55 units/week (from recent trend)
- December forecast: 55 × 2.0 = 110 units/week
- July forecast: 55 × 0.6 = 33 units/week
This simple adjustment captures seasonal patterns without complex modeling.
Incorporating Known Events
Statistical methods only see the past. You see the future—at least the parts of it you're planning.
Adjust your forecast for known events:
Promotions
If you're running a 20% off sale and historically that lifts sales by 40%, apply that lift to your baseline forecast for the promotion period.
New Distribution
Landing a new retail partner? Estimate the additional volume based on their projected door count and velocity.
Marketing Pushes
A significant advertising investment should increase demand. Estimate the impact based on past campaigns or industry benchmarks.
External Events
Industry events, holidays, competitor stockouts—anything you know about that could affect demand.
The key is documenting your assumptions. When actuals come in different from forecast, you can trace back to which assumptions were wrong and improve next time.
Building Your Forecast Cadence
A forecast sitting in a spreadsheet that nobody updates is worthless. Build a regular rhythm:
Weekly Review (30 minutes)
- Compare last week's actual sales to forecast
- Note any significant variances
- Adjust near-term forecast if needed
Monthly Deep Dive (2 hours)
- Review forecast accuracy for the past month
- Update baseline forecasts with new data
- Adjust seasonal indices if patterns are shifting
- Incorporate new information about upcoming events
Quarterly Planning (half day)
- Look further ahead (6-12 months)
- Plan for major seasonal periods
- Align with sales and marketing on promotions and launches
Consistency matters more than perfection. A forecast that gets reviewed weekly will outperform a sophisticated model that gets updated quarterly.
Tools You Already Have
You don't need specialized software to start. Use what you have:
Spreadsheets (Excel/Google Sheets)
Perfectly adequate for brands with under 500 SKUs. Set up templates for moving averages and seasonal calculations. The formulas are straightforward.
Your Inventory System
Many inventory and order management systems have basic forecasting built in. It might not be sophisticated, but it's already connected to your data.
Planster
When spreadsheets start breaking—formulas overwritten, multiple versions floating around, nobody trusts the numbers—that's when purpose-built tools earn their keep. Planster handles the calculations automatically and keeps everyone working from the same source of truth.
Common Mistakes to Avoid
Forecasting Too Far Out
Accuracy degrades with time horizon. Focus your energy on the periods that drive near-term decisions (usually 4-12 weeks out). Longer-range forecasts can be rougher.
Ignoring Forecast Error
Every forecast is wrong. The question is how wrong. Track your accuracy so you know how much buffer you need in safety stock.
Treating All SKUs the Same
Your hero products deserve more attention than your long tail. Focus forecasting effort where it matters most—the SKUs that drive the majority of revenue.
Never Revising
New information should change your forecast. If sales are trending 20% above forecast, update the forecast rather than waiting to see if it continues.
Key Takeaways
- You don't need data scientists or expensive tools to forecast demand
- Start with historical data organized by SKU and channel
- Simple moving averages work for many products
- Add seasonal indices for products with predictable patterns
- Adjust for known events like promotions and new distribution
- Build a consistent cadence: weekly reviews, monthly deep dives
- Track forecast accuracy to improve over time
Frequently Asked Questions
Q: How accurate should my demand forecast be?
For CPG brands at the SKU level, 70-85% accuracy is typically good. Perfect accuracy isn't realistic—the goal is being close enough that your inventory decisions work out most of the time.
Q: What if I don't have 12 months of history for a product?
Use whatever history you have, supplemented by data from similar products. A new flavor of an existing product line can borrow patterns from established flavors.
Q: Should I forecast in units or dollars?
Forecast in the units you order. If you order by case, forecast by case. Dollars are useful for financial planning but units drive operational decisions.
Q: How do I forecast for a brand new product?
Use analogous products, pre-order data, and conservative assumptions. Plan for multiple scenarios and start with smaller initial orders until real sales data comes in.
Q: When should I upgrade from spreadsheets to software?
When spreadsheets start causing problems: formulas breaking, version confusion, lack of trust in the numbers, or when the time to maintain them exceeds the time to use a purpose-built tool.