Research & Analysis / Trend Analysis
Extract underlying trend from seasonal patterns and irregular noise.
Why This Prompt Exists
December sales always look amazing — but is it growth or just Christmas?
You get:
- comparing December to November and panicking about the “surge” (it’s seasonal)
- comparing January to December and panicking about the “crash” (it’s seasonal)
- making year-over-year comparisons without adjusting for different holiday dates
- missing real growth because it’s hidden inside seasonal peaks
- forecasting that fails because it doesn’t account for repeating patterns
But time series have components:
- trend: long-term direction (up, down, flat)
- seasonal: repeating pattern (weekly, monthly, quarterly, yearly)
- cyclical: multi-year waves (economic cycles)
- irregular: random noise (can’t be predicted)
- calendar effects: holidays, leap years, trading days
Without decomposition, you mistake seasonality for trend.
This prompt separates what’s real growth from what’s just the time of year.
The Prompt
Assume the role of a time-series economist who decomposes seasonal patterns. Your task is to separate a metric into trend, seasonal, and noise components. Generate: 1. SEASONAL PATTERN IDENTIFICATION - Does seasonality exist? (Yes/No — confidence) - Period length (weekly, monthly, quarterly, annual) - Typical high periods (e.g., "December is 30% above average") - Typical low periods (e.g., "January is 20% below average") 2. TREND COMPONENT (seasonally adjusted) - Underlying direction (up / down / flat) - Rate of change (per period, after removing seasonality) - Is the trend statistically significant? 3. SEASONAL ADJUSTMENT - Raw value for most recent period: [value] - Seasonal adjustment factor: [multiplier] - Seasonally adjusted value: [value] - Plain-English: "After removing seasonal effects, the real change is X%" 4. CALENDAR EFFECTS (if applicable) - Holiday shifts (Easter, Black Friday) - Trading day differences - Leap year effects 5. WHAT THIS MEANS FOR DECISIONS - Compare performance to same period last year (not previous period) - Use seasonally adjusted numbers for trend analysis - Plan inventory/staffing around seasonal peaks INPUTS: Time-series data (at least 2 years recommended): [PASTE MONTHLY OR WEEKLY VALUES] Metric name: [E.G., "E-commerce revenue"] Business context: [E.G., "Retail, heavy holiday season"] Known seasonal factors (if any): [E.G., "December is typically 2x normal"] RULES: - Require at least 1.5 cycles of data to identify seasonality (18 months for annual pattern) - Flag when seasonality is changing over time (evolving pattern) - Note that seasonally adjusted numbers can still be noisy - Distinguish between calendar seasonality (e.g., weather) and business-driven patterns (e.g., back-to-school)
How To Use It
- Use at least 2 years of data — 1 year can’t distinguish seasonality from trend.
- Always compare to same period last year, not previous period, for seasonal metrics.
- Calculate seasonally adjusted metrics for internal performance tracking.
- For retail, account for holiday calendar shifts (e.g., Thanksgiving is different dates).
- Present both raw and seasonally adjusted numbers to stakeholders — explain the difference.
Example Input
Time-series data:
“Monthly revenue ($M): Jan 24: 8.2, Feb: 7.9, Mar: 8.5, Apr: 8.1, May: 8.4, Jun: 8.3, Jul: 8.6, Aug: 8.7, Sep: 8.9, Oct: 9.2, Nov: 12.5, Dec: 18.2, Jan 25: 8.5, Feb: 8.1, Mar: 8.8”
Metric name:
E-commerce revenue
Business context:
Retail with strong holiday season (Nov-Dec)
Why It Works
Most business metrics are compared to last month — which is almost always wrong for seasonal businesses.
This framework improves outcomes by forcing:
- seasonal pattern identification (know your rhythm)
- trend extraction (what’s actually changing)
- seasonal adjustment (apples-to-apples comparison)
- calendar effect detection (holidays move)
- decision implications (how to act on decomposed data)
Great seasonality decomposition doesn’t remove the pattern — it helps you see through it.
Build Better AI Systems
Subscribe for advanced prompt engineering, AI coding tools, debugging frameworks, and practical strategies for developers and engineers.