How Mo's Forecasting Works

circle-info

Quick Answer: Mo analyzes your historical sales data to generate a directional forecast—a statistically grounded starting point. Because no model can capture promotions, market shifts, or product launches on its own, Mo is designed for you to layer in business logic through forecast guidelines and targeted edits.

The Baseline Challenge

Every forecast starts with the same fundamental question: what will demand look like in the future?

Mo answers this by analyzing your historical sales data—identifying trends, seasonal patterns, and demand signals across your catalog. The result is a directional forecast: a data-driven projection that reflects what your sales patterns suggest will happen next.

The challenge: Historical data tells you what has happened, not what will happen. Statistical models are excellent at detecting patterns, but they cannot account for:

  • Upcoming promotions that will spike demand

  • New product launches with no sales history

  • Market shifts like a competitor exiting or a viral trend

  • Supply-side changes like a key supplier delay

  • Strategic decisions like entering a new channel or discontinuing a line

This is why Mo generates a directional forecast as a starting point, not a finished plan.


From Directional Forecast to Demand Plan

The forecasting process in Moselle follows three stages:

Stage 1: Generate a Directional Forecast

Mo processes your historical data and produces a baseline projection. You choose the approach:

Approach
How It Works
Best For

Top-Down

Starts with a revenue target and distributes units across SKUs

Financial alignment, new brands, stable demand

Bottom-Up

Analyzes each SKU individually and aggregates upward

Rich historical data, seasonal products, complex catalogs

Both approaches give you a statistically grounded starting point. Most teams use both and reconcile the differences.

Stage 2: Layer In Business Logic

This is where your expertise transforms a directional forecast into an accurate demand plan. Mo supports two refinement methods:

  • Mass Updates via Forecast Guidelines — Apply broad business rules across categories (e.g., "summer products increase 40% May through August")

  • Surgical Edits — Make targeted adjustments to specific SKUs through conversation with Mo

Stage 3: Finalize and Execute

Once your forecast reflects both the data and your business knowledge, lock it and move into replenishment planning. Moselle translates your finalized forecast into production plans and purchase orders.


Why Directional Forecasts Need Business Logic

A common misconception is that a good forecasting model should "just get it right" without human input. In practice, even the best statistical models produce directional guidance that requires refinement.

Here's why:

Models see patterns, not context

Mo can detect that your sunscreen sales spike every June. But it doesn't know you're planning a 30% off promotion this June, or that a competitor just launched a similar product at a lower price.

New products have no history

When you launch a new SKU, there's no historical data to analyze. Mo can use comparable products to estimate demand, but your knowledge of the product's positioning, marketing support, and target audience is critical.

Business strategy evolves

Your go-to-market strategy changes over time. Entering a new marketplace, adjusting pricing, or shifting marketing spend all affect demand in ways that historical data cannot predict.

One-time events distort patterns

A viral social media post, a weather event, or a supply shortage can create anomalies in your historical data. Without context, the model may over- or under-weight these events in future projections.

circle-check

What Mo Considers When Forecasting

When generating a directional forecast, Mo evaluates several dimensions of your data:

  • Trend direction — Is demand for this SKU growing, declining, or stable?

  • Seasonal patterns — Are there recurring peaks and valleys tied to time of year?

  • Demand volatility — How much does demand fluctuate week to week or month to month?

  • Sales velocity — What's the typical rate of sales, and how has it changed recently?

  • Channel behavior — Do different sales channels show different demand patterns for the same product?

Mo selects the appropriate forecasting approach based on the characteristics of your data—you don't need to manually configure model parameters.


The Feedback Loop

Forecasting improves over time. Moselle tracks forecast accuracy using MAPE (Mean Absolute Percentage Error) so you can see where your projections are landing relative to actuals.

MAPE
Interpretation

< 10%

Highly accurate

10–20%

Good

20–30%

Reasonable

> 30%

Needs improvement

Use the Forecast Performance Report to identify which categories or SKUs need better guidelines, and refine your approach each cycle.


Frequently Asked Questions

Can Mo forecast without historical data?

Mo needs historical sales data to generate a directional forecast. For new products without history, you can assign product comparables—existing items with similar demand patterns—to give Mo a reference point. You can also manually set projections for new launches.

How far back does Mo look at historical data?

Mo uses all available historical data in your account. Generally, more data leads to better pattern detection, but recent trends are weighted more heavily than older data.

Does Mo automatically account for promotions?

Not automatically. Promotions are a key example of business logic you should layer in through forecast guidelines or surgical edits. Tell Mo about upcoming promotions, and it will adjust the affected SKUs accordingly.

How often should I regenerate my directional forecast?

Most teams regenerate quarterly or when significant data changes occur (e.g., after a major selling season). Between regenerations, use Mo's refinement tools to keep your forecast current.

What if my directional forecast looks completely wrong?

Start by checking your data inputs—are all sales channels connected and syncing? Are there data gaps or anomalies? If the data looks clean, the forecast may reflect patterns you don't expect. Ask Mo to explain the reasoning behind specific projections before making broad changes.


How to Create a Forecastchevron-rightRefine Your Forecastchevron-rightBuild your forecast with Mochevron-rightDemand Forecast Performancechevron-right

Last updated