“Forecasts are always wrong.”
I have heard this quite often.
Yes, it is true. Forecasts are wrong because they are predictions about the future.
But, it is critical to know how wrong they are.
KPMG and Gartner indicate that companies can reduce finished goods inventory levels by 1%-7% for every 1% improvement in forecast accuracy.
This is an important improvement for operations and for the financials too.
The image shows 7 steps on how to create and improve a forecast.
Let’s dive in!
STEP 1: Start with Historical Data Collection & Cleaning
Action: Gather and clean at least 3 years of past sales or shipment data.
Remove outliers, fill gaps, and correct errors. This step might not feel glamorous, but it’s the foundation of everything. Bad input = bad output.
Tip: Watch for data distortions from stockouts, pricing changes, or system migrations. Flag them early.
STEP 2: Segment Your Demand
Action: Break down your demand into meaningful segments.
You can segment by product category, volume, value, customer type, region, or channel.
The right segmentation will depend on the business. We want criteria that make forecasting more accurate and actionable.
Tip: High-volume, high-variability SKUs may need more attention than stable, low-volume ones. Prioritize accordingly.
STEP 3: Generate a Baseline Statistical Forecast
Action: Use historical data to generate a starting forecast using statistical models.
Time series models like moving average or exponential smoothing are a good starting point. You’re not aiming for perfection. We want a solid base to build on.
Tip: Many planning tools have these models built in. You can use Excel to start. In the forecasting infographic bundle, I have included a step-by-step guide on how to use Excel for forecasting.
STEP 4: Apply Seasonality and Trend Adjustments
Action: Fine-tune your baseline with known seasonal patterns or trend shifts.
This includes promotional spikes, industry cycles, or ramp-ups for new products. You want your forecast to reflect reality; not just math.
Tip: Use seasonal indices and adjust for known events (like back-to-school, Black Friday, or peak flu season).
STEP 5: Collaborate & Fine-Tune in S&OP Meetings
Action: Align with cross-functional teams to turn your statistical forecast into a consensus forecast.
Marketing knows the campaign calendar and new products. Sales sees what customers are saying. Finance understands the revenue pressure. Bring it all together to reality-check and align.
Tip: Don’t let consensus become compromise. Use data to challenge assumptions and ground the discussion.
STEP 6: Adjust for Market Intelligence
Action: Layer in real-world insights that models can’t see.
This includes competitor moves, macroeconomic shifts, regulatory changes, or customer-specific insights. These are often the blind spots in pure statistical forecasting.
Tip: Create a simple intake process for Sales and Marketing to submit forecast-impacting information monthly. It is helpful to have regular check-ins for cross-functional alignment.
STEP 7: Incorporate Forecasts into S&OE (Sales & Operations Execution)
Action: Use the forecast to drive near-term execution decisions.
This is where planning becomes making it happen. S&OE helps teams translate the forecast into inventory, production, and service level decisions in real time.
Tip: The key is that the forecast drives action.
BONUS STEP: Build a Continuous Feedback Loop
Action: Track your forecast accuracy and improve continuously.
Compare forecast vs. actuals, investigate the misses, and adjust your model accordingly. This step transforms forecasting from a reactive chore to a strategic advantage.
Tip: Measure forecast accuracy at the right level (weekly, monthly, or item level) and always ask why you missed. Then, close the loop.
Conclusion
These 7 steps are your roadmap to creating a forecast to turn the data actionable, and to empower better decisions across the business.
You don’t need to get it all perfect from day one. Just start with Step 1 and build from there.
