Your forecast says next week should be 1,000 units. But yesterday's sales were 200 units—double the daily average. Is the forecast wrong, or was yesterday an anomaly?
This is where demand sensing comes in. It's not the same as forecasting, but it complements forecasting in important ways.
Demand Forecasting: Looking Forward
Demand forecasting predicts future demand based on historical patterns, trends, and known events.
Characteristics of Forecasting
Time horizon: Weeks to months out
Data sources: Historical sales, seasonal patterns, planned promotions, new distribution
Update frequency: Weekly or monthly
Best at: Capturing structural patterns—seasonality, growth trends, planned events
What Forecasting Does Well
Long-term planning: Production schedules, capacity planning, budgeting—all require looking months ahead.
Seasonal preparation: Building inventory for holiday peaks needs forecasts generated months in advance.
New product planning: Launches require forecasts before any sales data exists.
Strategic decisions: Forecasting supports decisions about suppliers, warehousing, and inventory investment.
Where Forecasting Struggles
Short-term accuracy: What will sell this week? Forecasting tools aren't designed for this precision.
Unexpected events: Competitor stockouts, viral moments, weather events—forecasts can't predict the unpredictable.
Rapid changes: When demand shifts suddenly, forecasts lag behind reality.
Demand Sensing: Reading the Present
Demand sensing interprets real-time signals to understand what's happening now and adjust near-term expectations.
Characteristics of Sensing
Time horizon: Days to a few weeks
Data sources: POS data, website traffic, social mentions, weather, competitor activity
Update frequency: Daily or real-time
Best at: Detecting shifts as they happen, short-term course corrections
What Sensing Does Well
Pattern breaks: Sensing catches when current demand deviates from forecast—and helps you respond.
Short-term adjustments: This week's replenishment decisions can incorporate what happened yesterday.
Early warning: A sudden demand spike shows up in sensing before it becomes a stockout.
Event response: External events (weather, news, competitor actions) affect demand sensing picks up changes forecasting might miss.
Where Sensing Falls Short
Long-term visibility: Sensing reads the present, not the future. It can't tell you what demand will be in three months.
Structural planning: You can't plan production capacity or negotiate supplier contracts based on this week's data.
Noise vs. signal: Is yesterday's spike a trend or an anomaly? Sensing provides data; interpretation requires judgment.
How They Work Together
Forecasting and sensing aren't competing approaches. They're complementary:
Forecasting Sets the Baseline
Start with a forecast that captures structural patterns: seasonality, trends, planned events.
Sensing Provides Course Corrections
As time passes and real data comes in, sensing adjusts the near-term forecast.
Example
January 1: Forecast says February week 3 should be 1,000 units (based on seasonality and planned promo).
February 1: Sensing shows current velocity is running 15% hot. February week 3 estimate adjusts to 1,150 units.
February 10: Sensing picks up a competitor stockout driving extra traffic. Week 3 adjusts to 1,300 units.
February 17: Week 3 begins. You're prepared for elevated demand because sensing updated your plan.
Practical Demand Sensing for Mid-Market Brands
Enterprise companies invest millions in demand sensing platforms with AI and real-time data feeds. You don't need that complexity to capture most of the value.
Simple Sensing: Daily Sales Review
Look at yesterday's sales vs. recent average. Significant deviation (±20%+) warrants investigation.
Example: Daily average is 100 units. Yesterday was 130. If today is also 125+, you're seeing a trend, not noise. Adjust your near-term expectations.
Watch for Pattern Breaks
Train yourself to notice when current reality diverges from forecast:
- Velocity increasing when it should be flat
- A SKU suddenly moving faster than usual
- Channel mix shifting unexpectedly
Each observation is a sensing signal.
Monitor External Signals
Weather: Extreme weather affects demand for many CPG products. Hot week → sunscreen demand up. Cold snap → soup demand up.
Competitive activity: Track competitor stock levels on Amazon. Their stockout becomes your opportunity.
Social media: Mentions of your brand or product spiking? Something's happening.
Weekly Sensing Review
Spend 30 minutes weekly comparing this week's actuals to last week's forecast:
- Which SKUs are deviating?
- Which channels are different than expected?
- What external factors might explain the deviation?
Adjust next week's expectations based on what you learn.
When Each Approach Matters More
Sensing Matters More When...
Lead times are short: If you can reorder in days, sensing enables reaction. With 12-week lead times, it's less actionable.
Demand is volatile: High variability means forecasts go stale quickly. Sensing keeps you current.
You sell perishables: Shelf life pressure means you can't hold excess inventory. Near-term accuracy is critical.
You have good data: Sensing requires near-real-time visibility into sales, inventory, and external signals.
Forecasting Matters More When...
Lead times are long: You're planning orders 8+ weeks out. Sensing's short-term view isn't directly actionable.
Demand is stable: With low volatility, forecasts stay accurate longer. Sensing adds overhead without much benefit.
You're capacity-constrained: Production and supplier capacity need to be planned months ahead.
You're budgeting: Financial planning requires looking forward, not just reading the present.
Building Sensing Into Your Process
If you want to incorporate sensing:
Start Simple
Daily or weekly review of actuals vs. forecast. Investigate significant variances. Adjust near-term orders based on what you learn.
Add External Data Gradually
Weather correlations are often the easiest to establish. Competitor monitoring is valuable but takes effort. Start with one external signal and add more over time.
Automate the Obvious
Set up alerts for:
- SKUs running more than 20% above or below forecast
- Stockouts on key competitors (Amazon alerts, manual checks)
- Velocity changes that trigger reorder point reviews
Preserve the Forecast
Don't let sensing override long-term forecasts. Use sensing for near-term adjustments while keeping the structural forecast intact for planning horizons beyond a few weeks.
Key Takeaways
- Demand forecasting predicts future demand using historical patterns and planned events
- Demand sensing reads real-time signals to detect current demand shifts
- They're complementary: forecasting sets the baseline, sensing provides course corrections
- Sensing matters more with short lead times, high volatility, and perishable goods
- Forecasting matters more with long lead times, stable demand, and capacity planning
- Start simple: daily/weekly reviews of actuals vs. forecast, with investigation of variances
Frequently Asked Questions
Q: Do I need specialized demand sensing software?
For most mid-market brands, no. A disciplined process of reviewing actuals vs. forecast catches most sensing value. Software helps when you have massive SKU counts or need real-time automation.
Q: How often should demand sensing update my plans?
Sensing can inform daily decisions for very near-term (next 1-2 weeks). It shouldn't override longer-term forecasts unless you see a sustained pattern change.
Q: What signals are most useful for sensing?
Start with your own sales velocity—it's the most direct signal. Then add weather (if relevant to your category), competitor stock status, and website traffic.
Q: Can sensing replace forecasting?
No. Sensing tells you about now. Planning production, negotiating with suppliers, and building for seasonal peaks all require forward-looking forecasts.
Q: How do I know if a spike is a trend or an anomaly?
Look for persistence. One day above average might be noise. Three days suggests a pattern. A week confirms it. Investigate the cause to understand if it's sustainable.