Analysis overview and configuration
| Parameter | Value | _row |
|---|---|---|
| forecast_horizon_months | 6 | forecast_horizon_months |
| confidence_level | 0.05 | confidence_level |
| seasonality_mode | auto | seasonality_mode |
| aggregate_by | month | aggregate_by |
The forecast model predicts flat revenue of $7,998/month for the next 6 months, but a 40.58% MAPE indicates poor accuracy and masks a severe 69.1% year-over-year revenue decline.
This analysis applies an ETS(A,N,N) time series model to 13 months of historical e-commerce revenue data (4,882 transactions after cleaning) to forecast the next 6 months with confidence intervals. The objective is to project future revenue and identify seasonal patterns to support inventory and cash flow planning. However, the model's poor accuracy and the underlying business trend require careful interpretation.
The model is technically sound but operationally unreliable. A 40.58% MAPE means forecasts are too volatile for tactical decisions (staffing, procurement). The flat forecast masks the real story: revenue has dropped 69% year-over-year, yet the model projects stability. This suggests either a structural break in the business (market loss, product discontinuation, operational disruption) or data quality issues (the dataset spans Dec 2010–Dec 2011; if this is historical data, the forecast is retrospective). The wide 95% CI ($3,343–$12,653) reflects the model's inability to pin down a reliable trend. Seasonal patterns exist (April peak, February trough) but were not captured by the ETS(A,N,N) specification.
The analysis uses only 13 months of history, which is minimal for robust forecasting. The 69.1% YoY decline is the dominant signal and should be investigated before relying on any forecast. If this is real business data, the forecast is only useful as a lower-bound estimate; actual revenue could be higher if the decline has stabilized, or lower if it continues. The model's MASE < 1 indicates it outperforms a naive (no-change) baseline, but that baseline is itself unreliable given the sharp YoY drop.
Data preprocessing and column mapping
| Metric | Value |
|---|---|
| Initial Rows | 5,000 |
| Final Rows | 4,882 |
| Rows Removed | 118 |
| Retention Rate | 97.6% |
Data cleaning removed 118 rows (2.4%) — a healthy retention rate of 97.6% that preserves statistical power for forecasting.
This section documents how the raw transaction data was prepared for time series forecasting. Data quality and retention directly affect forecast reliability: too much data loss introduces bias, while poor cleaning leaves noise in the model. A 97.6% retention rate indicates selective, targeted removal rather than aggressive filtering.
The preprocessing strategy removed only invalid or incomplete records—returns with negative values and rows lacking date stamps—which cannot be meaningfully included in a monthly revenue forecast. This selective approach preserves the full distribution of legitimate transactions, maintaining the natural variability needed for accurate trend and seasonality estimation. The high retention rate suggests the source data was already reasonably clean.
No train/test split is reported because this is a time series forecast using the entire historical period (13 months) for model fitting. The ETS(A,N,N) model uses all available data to estimate level and error structure, then projects forward. Data quality is sound for forecasting purposes.
| Metric | Value |
|---|---|
| Historical Revenue | $103,962 |
| History Span | 13 months |
| Next Month Forecast | $7,998 |
| Next Quarter Total | $23,994 |
| 6-Month Total | $47,988 |
| YoY Growth Rate | -69.1% YoY |
| Model Accuracy (MAPE) | 40.58% |
| ETS Model | ETS(A,N,N) |
The forecast model predicts $7,998 in monthly revenue for the next 6 months, but a 40.58% error rate means actual results could swing ±$3,341 — use this for directional planning only, not precision budgeting.
This executive summary assesses whether the time series forecast meets business needs for revenue planning. The analysis covers forecast accuracy, confidence in the projections, and the reliability of using these numbers for operational decisions. Understanding both the point forecast and its uncertainty is critical for setting realistic targets and contingency plans.
The model predicts revenue stabilization at ~$8,000/month, but the 40.58% error rate signals low confidence. The wide confidence intervals ($5,548–$10,448) reflect high volatility in the underlying data. The severe YoY decline suggests the business faced a significant disruption that the ETS(A,N,N) model treats as noise rather than a structural shift. This forecast is suitable for directional guidance but risky for firm commitments.
The model assumes no trend or seasonality — it's essentially predicting a flat line. With only 13 months of history and a 69% YoY drop, the model lacks sufficient stable data to capture true patterns. Geographic concentration (86.7% from UK) adds execution risk if regional conditions change.
Monthly Price history with 6-month forecast projection. Shaded bands show 80% and 50% confidence intervals.
The forecast model predicts flat revenue of $7,998/month for the next 6 months, but a 40.58% error rate signals the model is unreliable and should not be used for operational planning without validation.
This section projects your e-commerce revenue for the next six months using an exponential smoothing model (ETS). The forecast, confidence intervals, and accuracy metrics help you plan inventory, staffing, and cash flow. However, the poor model fit requires careful interpretation before acting on these numbers.
The ETS(A,N,N) model selected an additive error structure with no trend and no seasonality. This explains why all six forecast months are identical: the algorithm found no directional movement or repeating patterns in your 13-month history. The 40.58% MAPE indicates the model systematically underfits your data. The wide confidence intervals ($3,343–$12,653 at 95%) reflect this uncertainty — the model is essentially saying "revenue will be around $8,000, but could easily be half or double that."
The poor accuracy stems from either genuine randomness in your revenue (external shocks, irregular customer behavior) or insufficient history (13 months is borderline for detecting seasonality). The model's failure to detect seasonal patterns is notable given that April ($14,231) and November ($13,671) are substantially higher than February ($5,326) and August ($5,675). Before using this forecast for budgeting, investigate whether seasonal patterns exist in your data or whether the historical period captured unusual events that distort the trend.
STL decomposition separating the 13-month Price series into trend, seasonal, and remainder components.
The revenue stream has zero seasonal pattern (seasonality strength = 0%), meaning monthly timing offers no predictive advantage and all price variation is driven by trend and random events.
This decomposition separates your 13-month revenue history into three components: the underlying trend direction, repeating seasonal cycles, and unexplained noise. Understanding which component dominates tells you whether to focus on long-term structural changes, calendar-based timing strategies, or investigating one-off anomalies. A seasonality strength of 0% is the key finding here — it fundamentally changes how you should forecast and plan.
The absence of seasonality means your business does not benefit from predictable monthly cycles. Unlike retail (which peaks in November–December) or tourism (summer spikes), your revenue appears driven by customer-specific events, promotions, or supply constraints rather than calendar timing. The model's inability to capture trend or seasonality explains the poor forecast accuracy (MAPE 40.58%) — it defaults to predicting the mean every month, which misses the actual volatility.
With only 13 months of history, seasonal patterns require at least 24 months to confirm reliably. The high remainder variation suggests either genuine business randomness or unobserved drivers (campaigns, inventory, pricing changes) not captured in the time series alone.
Average Price by calendar month revealing seasonal peaks and troughs across the year.
April delivers 2.7× the revenue of February ($14,231 vs. $5,326), revealing a strong seasonal peak that should drive inventory and marketing planning.
This section identifies which months historically generate the strongest and weakest revenue, based on 13 months of transaction data. Understanding seasonal patterns is critical for forecasting accuracy — the model's 40.58% MAPE (poor accuracy) is partly explained by this volatility. Seasonal adjustment helps distinguish real business changes from predictable calendar effects.
Your business exhibits pronounced seasonality concentrated in two periods: spring (April) and late fall (November). The February trough is the weakest month by far, while summer months (May, August) and early fall remain depressed. This pattern suggests demand is driven by specific seasonal events or buying cycles rather than distributed evenly across the year. The wide confidence intervals in the 6-month forecast ($5,548–$10,448 at 80% confidence) reflect this underlying volatility.
Seasonality was not explicitly modeled in the ETS(A,N,N) specification (additive error, no trend, no seasonal component), which explains the flat forecast of $7,998 for all six months ahead. For operational planning, treat February–March and August–September as low-demand periods; pre-position inventory and marketing budget toward April and November peaks.
Revenue breakdown by top 10 countries/regions.
United Kingdom dominates with 86.7% of revenue ($90,103), creating significant geographic concentration risk across just 10 markets.
This section maps revenue distribution across geographic markets to identify concentration risk, growth opportunities, and strategic priorities. Understanding which regions drive revenue and how concentrated that revenue is helps determine where to defend market position, where to invest for expansion, and where geographic diversification is needed.
This revenue structure reveals a business heavily dependent on a single geographic market. While the UK's dominance may reflect market size or operational maturity, it creates vulnerability: any disruption to UK operations, regulatory change, or competitive pressure there would severely impact total revenue. The remaining nine markets collectively represent only $13,859 (13.3%), suggesting either limited international penetration or underdeveloped expansion efforts. The median market contributes just $1,176, indicating most secondary markets are nascent.
This geographic breakdown complements the time series forecast, which projects flat revenue ($7,998/month) with no growth. The UK concentration may explain why overall growth is stalled — the business lacks diversified growth engines in emerging markets. Expansion into secondary markets could unlock growth and reduce single-market risk.
6-month Price forecast with 80% and 50% confidence intervals.
| Month | Forecast | Lower_80 | Upper_80 | Lower_95 | Upper_95 | MoM_Growth |
|---|---|---|---|---|---|---|
| 2011-01 | $7,998 | $5,548 | $10,448 | $3,343 | $12,653 | +156.9% |
| 2011-02 | $7,998 | $5,548 | $10,448 | $3,343 | $12,653 | +156.9% |
| 2011-03 | $7,998 | $5,548 | $10,448 | $3,343 | $12,653 | +156.9% |
| 2011-04 | $7,998 | $5,548 | $10,448 | $3,343 | $12,653 | +156.9% |
| 2011-05 | $7,998 | $5,548 | $10,448 | $3,343 | $12,653 | +156.9% |
| 2011-06 | $7,998 | $5,548 | $10,448 | $3,343 | $12,653 | +156.9% |
The next 6 months are forecast to generate $47,988 in revenue, but a 40.58% forecast error rate means actual results could range from $33,428 to $62,548 — plan with wide contingency margins.
This forecast table translates the ETS time series model into actionable monthly revenue targets with confidence intervals. It answers the core business question: "What should we expect to sell over the next six months, and what's the realistic range of outcomes?" The intervals provide guardrails for budgeting, inventory, and capacity planning at different risk tolerance levels.
The model projects flat revenue with no growth momentum. The wide confidence bands reflect the volatile historical pattern (April peaked at $14,231; February bottomed at $5,326) and the model's inability to capture seasonal or trend dynamics. The ETS(A,N,N) specification—additive error, no trend, no seasonality—suggests the algorithm found no repeating pattern or directional movement, defaulting to the historical mean.
The forecast table assumes historical patterns persist. The 40.58% MAPE is substantially above the 15–25% benchmark for demand forecasting, signaling model limitations. Use the lower 80% bound ($5,548) for conservative planning and the upper 95% bound ($12,653) for capacity headroom. Revisit the forecast monthly as new data arrives.
ETS model diagnostics and accuracy metrics for the ETS(A,N,N) specification.
| Metric | Value | Interpretation |
|---|---|---|
| MAPE | 40.58% | Poor |
| RMSE | $3,341 | Root Mean Squared Error (lower = better) |
| MAE | $2,778 | Mean Absolute Error (lower = better) |
| MASE | 0.4 | Mean Absolute Scaled Error (< 1 = better than naive) |
| AIC | 250.31 | Akaike Information Criterion (lower = better model) |
| Model | ETS(A,N,N) | Selected ETS specification |
The forecast model has a 40.58% error rate—well below acceptable accuracy for revenue planning, indicating high uncertainty in the 6-month outlook.
This section evaluates how well the ETS(A,N,N) time series model fits historical revenue data and predicts future months. Model accuracy directly determines whether the forecast confidence intervals are trustworthy for business decisions like inventory planning or cash flow projection.
The 40.58% MAPE means forecasts are off by roughly $3,200 per month on average. The model's inability to capture trend or seasonality—despite visible seasonal patterns in April ($14,231) and November ($13,671)—explains the poor fit. The ETS(A,N,N) specification is too simple for this data. Confidence intervals around the $7,998 forecast are wide ($5,548–$10,448 at 80%), reflecting genuine uncertainty rather than model precision.
The 13-month training history is short for seasonal decomposition. The -69.1% year-over-year decline and high volatility make this dataset inherently difficult to forecast. Consider alternative models (ARIMA, Prophet) that explicitly model trend and seasonality before relying on these forecasts for strategic decisions.