Advanced Analytics

Sales Forecast Example

A six-month sales forecast built from 36 months of U.S. sales history, using a seasonal year-over-year method, backtesting, and planning considerations.

United StatesUSDJanuary 2023 – June 2026
BUSINESS DATA ANALYSIS
Project Resources

Why the analysis was needed

A business had three years of monthly sales history but relied mainly on recent performance when planning upcoming periods. The objective was to create a forward-looking view that could be used as a planning input while keeping the assumptions and limitations visible.

BUSINESS QUESTION

What does the historical sales pattern suggest for the next six months, and how does the forecast compare with the most recent six-month period?

What the analysis used

Thirty-six months of monthly sales history followed by a six-month forecast period.

RECORDS36 monthly historical observations
FORECAST HORIZON6 months
FIELDSMonth and revenue
PERIODJanuary 2023 – December 2025 historical; January 2026 – June 2026 forecast

How the source was made analysis-ready

  1. 01

    Sorted the monthly revenue series chronologically and confirmed one observation for each month from January 2023 through December 2025.

  2. 02

    Checked the 36 historical values for missing months and unusually large discontinuities before calculating year-over-year changes.

  3. 03

    Calculated same-month year-over-year growth for 2024 versus 2023 and 2025 versus 2024 so recurring calendar position is preserved.

  4. 04

    Kept 2025 separate as a validation period for the initial backtest before using all available history for the final six-month projection.

How the analysis was structured

  1. 01

    Reviewed the 36-month series for trend and recurring same-month behavior rather than extrapolating only the latest month.

  2. 02

    Backtested a seasonal year-over-year growth rule on 2025: each 2025 month was estimated from the corresponding 2024 month using the growth observed from 2023 to 2024.

  3. 03

    Evaluated the 2025 holdout using mean absolute percentage error (MAPE), obtaining 0.7% MAPE.

  4. 04

    For January–June 2026, averaged the two observed same-month year-over-year growth rates and applied that average to the corresponding 2025 month.

  5. 05

    Compared the six-month forecast total with the immediately preceding six actual months and treated the result as a planning baseline, not a guaranteed outcome.

Historical Revenue$10.425M36 months of historical sales data
Forecast Horizon6 monthsJanuary 2026 – June 2026
Forecast Revenue$2.272MTotal seasonal-growth forecast for six months
Backtest MAPE0.7%2025 holdout backtest — 0.7% MAPE; not evidence of future accuracy
Historical Sales Trend36 months of historical monthly revenue
$000s
Jan 23Jul 24Dec 25
Sales ForecastHistorical sales followed by six-month forecast
$000s
HistoricalForecast
Jul 25Aug 25Sep 25Oct 25Nov 25Dec 25Jan 26Feb 26Mar 26Apr 26May 26Jun 26
01

The historical series has a sustained upward pattern

Monthly revenue increases from $210K in January 2023 to $382K in December 2025, with positive same-month year-over-year growth throughout the period.

02

The seasonal growth method performs well in the backtest

Applying 2023-to-2024 same-month growth to estimate the 2025 holdout produces 0.7% MAPE. This result measures performance on the sample dataset and should not be generalized to other datasets or future forecasts.

03

The six-month planning baseline is $2.272M

The January–June 2026 forecast totals $2.272M, compared with $2.172M across July–December 2025, a difference of approximately 4.6%.

04

The method preserves month-specific movement

The forecast ranges from $355K in January to $403K in June and does not force every month into a straight-line increase; May is below April because the method retains same-month historical structure.

Forecast Period
MonthForecast Revenue
Jan 2026$355K
Feb 2026$364K
Mar 2026$374K
Apr 2026$390K
May 2026$386K
Jun 2026$403K
Validation Summary
CheckResultInterpretation
2025 holdout backtest0.7% MAPELow backtest error on the sample dataset
Jan–Jun 2026 forecast$2.272MSix-month planning baseline
Previous six actual months$2.172MJul–Dec 2025 comparison period
Forecast vs. previous six months+4.6%Comparison, not a causal growth claim

What the numbers mean

The forecast provides a transparent planning baseline derived from historical sales patterns. The backtest evaluates performance on a held-out year, while the final projection preserves month-specific growth behavior. Planning decisions should also consider pricing, promotions, launches, capacity, and market conditions.

What was checked before interpretation

  • Confirmed the historical series contains 36 consecutive monthly observations with no missing months.
  • The 2025 holdout backtest was completed before generating the final 2026 projection and produced 0.7% MAPE on the sample dataset.
  • The six forecast months sum to $2.272M and the Jul–Dec 2025 comparison period sums to $2.172M.
  • The reported +4.6% comparison was recalculated from the two six-month totals.

How the findings could inform the next decision

01

Use $2.272M as a baseline, not a commitment

Treat the six-month total as a reference scenario for planning and compare it with management assumptions before setting inventory, staffing or sales commitments.

02

Add known future business events

Adjust or scenario-test the baseline when promotions, pricing changes, launches, closures or capacity changes are already known but absent from the historical series.

03

Reforecast as actuals arrive

Compare each new month with the forecast, record the error and update the model when the recent pattern materially departs from the historical relationship.

What this example does not establish

  • The synthetic dataset exhibits stable year-over-year behavior, contributing to the 0.7% backtest MAPE. Forecast performance will depend on the characteristics of the data being analyzed.
  • Only monthly revenue and calendar position are modeled. Promotions, price changes, product launches, economic conditions and capacity constraints are not explanatory variables.
  • The method uses a short three-year history, so structural changes or new seasonal patterns may not be represented.
  • The point forecast does not include a probabilistic prediction interval. Additional uncertainty estimates can support planning decisions when appropriate.

DELIVERABLE

What a comparable engagement could produce

01Six-month sales forecast
02Historical trend visualization
03Forecast comparison
04Monthly forecast table
05Method and interpretation notes

Why the output matters

The example shows a complete forecasting workflow: prepare the time series, define a transparent method, test it on a held-out period, produce the forecast, state its limitations and connect the result to a planning decision.

NEXT

Have a similar business question?