Menu

Retail Forecasting GPT — User Manual

📘 Retail Forecasting GPT — User Manual

Version 1.0 — For Planners, Buyers, Analysts & Retail Operators

Table of Contents

Part I — About Retail Forecasting GPT

  1. What the system does

  2. Key capabilities

  3. Forecasting principles used

  4. What the system cannot do

Part II — How to Use Retail Forecasting GPT

  1. Starting a forecasting session

  2. Supported use cases

  3. Required user inputs

  4. How to ask for forecasts

  5. How to ask for buying, allocation, or inventory guidance

Part III — How to Read Forecasting Outputs

  1. Forecast structure

  2. Low / Base / High range definition

  3. Confidence ratings

  4. Assumptions and risks

  5. Recommended actions

Part IV — Troubleshooting

  1. Interpretation errors

  2. Unexpected forecast results

  3. Data quality warnings

  4. Common user mistakes

Part V — Data Requirements & Upload Guide

  1. Minimum required fields

  2. Recommended fields

  3. Optional fields

  4. Templates

  5. Upload instructions

  6. Data validation checklist

  7. Example datasets

─────────────────────────────────────────

PART I — ABOUT RETAIL FORECASTING GPT

─────────────────────────────────────────

1. What Retail Forecasting GPT Does

Retail Forecasting GPT is an advanced decision-intelligence system designed specifically for retail planning, demand forecasting, buying, inventory optimization, and allocation.

It provides:

  • Trend + seasonality forecasting

  • Promo uplift modeling

  • Category & SKU-level forecasts

  • Channel shift detection

  • Inventory planning (buy plans, reorder logic, WOS, safety stock)

  • Markdown recommendations

  • Diagnostic retail intelligence

  • Scenario modeling

All forecasting follows the standardized methodology defined in:

Retail Forecasting GPT — Master Specifications.

2. Key Capabilities

Retail Forecasting GPT can:

  • Forecast demand by SKU, category, channel, store, or entire business

  • Apply trend, seasonality, promo, elasticity, lifecycle, and macro interpretations

  • Recommend buys, replenishment, safety stock, allocation strategies

  • Diagnose causes of demand change (traffic, conversion, promo, macro, competition, stockouts)

3. Forecasting Principles Used

The engine uses only transparent, retail-standard time-series logic:

  • Moving averages

  • YOY with trend adjustment

  • Trend + seasonality decomposition

  • Promo uplift ranges

  • Channel allocation logic

  • Range-based forecasting (Low / Base / High)

It never uses black-box ML forecasting unless explicitly requested.

4. What the System Cannot Do

To set correct expectations:

  • Cannot forecast without date + units

  • Cannot interpret promo effects without promo flags or price signals

  • Cannot guarantee precise SKU-level accuracy (inherent volatility)

  • Cannot override missing or broken data patterns

─────────────────────────────────────────

PART II — HOW TO USE RETAIL FORECASTING GPT

─────────────────────────────────────────

5. Starting a Forecasting Session

Upload a dataset or paste a table into the chat.

Then specify the scope:
“Forecast Q3 for SKU A123.”
“Forecast total business for next 12 months.”
“Forecast Women’s Tops by channel.”

If no scope is specified, the system will ask clarifying questions.

6. Supported Use Cases

You can request:

Forecasting

  • Quarterly forecasts (primary method)

  • Monthly forecasts

  • SKU-level direction forecasting

  • New SKU forecasting

  • Category / channel forecasting

Inventory & Buying

  • Buy quantity recommendations

  • Safety stock levels

  • Reorder triggers

  • WOS interpretation

  • OTB planning support

Allocation

  • Initial allocation

  • Replenishment allocation

  • Inter-store transfer suggestions

Promo & Price

  • Promo uplift forecasting

  • Price elasticity interpretation

Diagnostics

  • Trend analysis

  • Seasonality detection

  • Promo effect breakdown

  • Channel-shift analysis

7. Required User Inputs (if not in data)

Retail Forecasting GPT must ask you for:

  • Lead time

  • Planned promotions

  • Service level targets

  • Whether overstock or stockouts are more important

  • Category role (basic, fashion, seasonal, hero)

  • Channel strategy (stores vs ecommerce priority)

8. How to Ask for Forecasts

Example prompts:

  • “Forecast next quarter for all SKUs.”

  • “Forecast Q1–Q4 for Women’s Dresses category.”

  • “Forecast at the channel level for ecommerce vs stores.”

  • “Create a high/low/base scenario for the next 6 months.”

The system will automatically analyze your data, detect frequency, identify trend, seasonality, and promo effects, then generate a structured forecast.

9. How to Ask for Inventory or Allocation Guidance

Example prompts:

  • “Based on the forecast, what should I buy for Q3?”

  • “Do I need to reorder SKU B241?”

  • “What allocation should I use for new styles?”

  • “What is my WOS and what should I do about it?”

Retail Forecasting GPT will produce:

  • Buy quantities

  • Safety stock

  • Replenishment logic

  • Markdown recommendations

  • Allocation guidance

─────────────────────────────────────────

PART III — HOW TO READ FORECAST OUTPUTS

─────────────────────────────────────────

10. Forecast Output Structure

Every forecast will include:

  1. Headline summary (1–2 sentences)

  2. Data insight summary (trend, seasonality, promo, channel)

  3. Forecast table (Low / Base / High)

  4. Assumptions used

  5. Risks & watchpoints

  6. Recommended actions

  7. Forecast confidence rating

11. Forecast Ranges

  • Low Case → downside protection

  • Base Case → most likely outcome

  • High Case → upside potential

Ranges widen when volatility or uncertainty is high.

12. Confidence Ratings

Forecasts include:

  • High confidence

  • Medium confidence

  • Low confidence

  • Very low confidence (short or noisy history)

13. Assumptions & Risks

Retail Forecasting GPT lists every assumption explicitly:

  • Trend applied

  • Seasonality applied or excluded

  • Promo uplift assumptions

  • Channel shifts

  • Macro conditions (if visible in data)

14. Recommended Actions

GPT converts forecasts into operational guidance:

  • Buy X units

  • Reduce replenishment

  • Add safety stock

  • Plan markdowns

  • Reallocate inventory

─────────────────────────────────────────

PART IV — TROUBLESHOOTING

─────────────────────────────────────────

15. Interpretation Errors

If the output seems incorrect:

  • Data may contain duplicates

  • Date frequency may be inconsistent

  • Promo flags missing or misaligned

  • Stockout periods suppress units

16. Unexpected Forecast Results

Check:

  • Were promotions marked in the dataset?

  • Was a seasonal pattern expected but unavailable?

  • Is the SKU new, volatile, or declining?

17. Data Warnings

The system will warn you about:

  • Gaps in data

  • Outliers

  • Unreliable seasonality

  • Decreasing trend due to stockouts

  • Excess volatility

18. Common User Mistakes

  • Mixing weekly and monthly data

  • Omitting promo flags

  • Mislabeling channels

  • Including subtotals or merged rows

  • Uploading pivot tables instead of flat tables

─────────────────────────────────────────

PART V — DATA REQUIREMENTS & UPLOAD MANUAL

─────────────────────────────────────────

19. Minimum Required Fields

To generate any forecast:

Field Description
Date Time period (must be consistent)
Units Sold Sell-through count
SKU / Category Required for granularity

20. Recommended Fields

Greatly improves accuracy:

  • Category

  • Channel

  • Price

  • Promo flag

  • On-hand inventory

  • On-order

21. Optional Fields

Enhances diagnostic intelligence:

  • Traffic / conversion

  • Returns

  • Marketing spend

  • Campaign type

  • Region

22. Data Templates

Minimal Template

Date,SKU,Units Sold

Category + Channel Template

Date,Category,Channel,Units Sold,Price,Promo Flag

Full Diagnostic Template

Date,SKU,Category,Channel,Region,Units Sold,Price,Promo Flag,Promo Type,On Hand,On Order,In Transit,Traffic,Conversion,Returns,Marketing Spend

23. Data Upload Instructions

  1. Prepare your data in CSV or Excel

  2. Ensure no merged cells, totals, or pivot tables

  3. Upload the file directly or paste data into chat

  4. GPT will automatically detect:

    • Frequency

    • Trend

    • Seasonality

    • Promo effects

    • Outliers

  5. GPT will summarize findings and ask clarification questions if needed

24. Data Validation Checklist

  • Consistent date frequency

  • No missing periods

  • Units represent sales, not shipments

  • SKU codes consistent

  • Promo flags provided

  • No subtotal rows

  • No formatting artifacts (merged cells, blank headers)

  • Inventory fields are numeric

  • Channels clearly labeled

25. Example Upload

Date,SKU,Category,Channel,Units Sold,Price,Promo Flag,On Hand
2023-01-01,A123,Tops,Ecommerce,82,29.99,0,120
2023-01-08,A123,Tops,Ecommerce,96,24.99,1,75

📘 1-Page QUICK START GUIDE — Retail Forecasting GPT

What Retail Forecasting GPT Does

Retail Forecasting GPT transforms your retail data into:

  • Forecasts (Low / Base / High)

  • Trend + seasonality insights

  • Promo uplift modeling

  • Inventory buys & reorder logic

  • Allocation & replenishment guidance

  • Risk scenarios & diagnostics

How to Get Started (3 Steps)

STEP 1 — Upload Data

Upload a CSV or Excel file with (minimum):

  • Date

  • Units Sold

  • SKU (or category if forecasting at category level)

Recommended optional fields:

  • Price, promo flag, on-hand inventory

  • Channel, category, region

  • Traffic, conversion, returns

STEP 2 — Tell GPT What You Want

Examples:

  • “Forecast Q3 for total business.”

  • “Forecast next 6 months for these SKUs.”

  • “Build a buy plan for Q2.”

  • “What is the WOS trend and what should I do?”

  • “Explain why sales dropped in January.”

GPT will automatically:

  • Detect trend, seasonality, and promo effects

  • Identify outliers, gaps, stockouts

  • Build forecasting ranges

  • Generate retail-ready insights

STEP 3 — Take Action

GPT provides:

  • 📦 Buy quantities (with lead-time logic)

  • 🔁 Replenishment recommendations

  • 🏬 Allocation strategy (initial + replenishment)

  • 📉 Markdown planning

  • ⚠️ Risks & watchpoints

  • 📊 Scenario modeling

Best Practice Prompts

  • “Forecast my dataset and summarize key drivers.”

  • “Create a buy plan using 95% service level and 8-week lead time.”

  • “Diagnose why SKU X is declining.”

  • “Model a 30% off promo and show upside/downside.”

What to Do if Data Has Issues

GPT will warn you about:

  • Missing dates

  • Promotional distortions

  • Stockouts masking demand

  • Inconsistent frequencies

  • Noisy or short history

Provide clarification when asked.

That’s it. You’re ready to forecast, buy, and allocate like a senior retail planner.

Views: 3