---
title: "Demand forecasting with language models in the loop"
description: "Demand and sales forecasting appear on 66 recent papers in the operations research and AI journals, but only 9 of them carry a language-model, foundation-model or text keyword. Retailers now feed product descriptions, reviews, weather and news into forecasting systems. Does text improve forecasts for new products, where history is missing, and does it help where history is long?"
url: https://scholar4ai.com/ideas/demand-forecasting-with-llms
markdown_url: https://scholar4ai.com/ideas/demand-forecasting-with-llms.md
license: CC BY-NC 4.0 (https://creativecommons.org/licenses/by-nc/4.0/)
share_text: "Demand forecasting with language models in the loop: Demand and sales forecasting appear on 66 recent papers in the operations research and AI journals, but only 9 of them carry a language-model, foundation-model or text keyword."
date: "2026-10-23"
updated: "2026-10-23"
status: "seed"
tags: ["operations research", "retail", "ai"]
keywords: ["demand forecasting", "sales forecasting"]
---

# Demand forecasting with language models in the loop

Demand and sales forecasting appear on 66 recent papers in the operations research and AI journals, but only 9 of them carry a language-model, foundation-model or text keyword. Retailers now feed product descriptions, reviews, weather and news into forecasting systems. Does text improve forecasts for new products, where history is missing, and does it help where history is long?

## What prompted this

Since October 2025, 66 papers carry a demand-forecasting, sales-forecasting or demand-prediction keyword (31 in operations research journals, 22 in AI): "demand forecasting" on 42, "sales forecasting" on 7, "retail demand forecasting" on 3. Only 9 of them also carry a language-model, foundation-model, transformer, deep-learning or text keyword.

## Questions

- For new products with no sales history, do embeddings of descriptions and images predict first-season demand better than analogue-based methods, and by how much?
- Does adding text (reviews, news, promotions) improve forecasts for established products, or add noise?
- Do the forecast gains translate into lower inventory and fewer stockouts once ordering rules use them?

## Data that could help

Retail sales data with product descriptions and images (public datasets such as the M5 and fashion retail datasets), and inventory records from a retailer.

## Next step

Compare new-product forecast errors from an embedding-based model and an analogue method on a fashion retail dataset with product images and text.

## Related research activity

*The keywords this note follows, in the tracked finance, accounting, economics, management, operations research and AI journals as of 2026-10-26*

| Keyword | Papers, last 30 days | Rank | Papers per month, Oct 23 – Oct 26* | Papers, last 36 months |
|---|---:|---:|---|---:|
| demand forecasting | 10 | #121 | 3, 4, 5, 4, 2, 2, 3, 2, 4, 3, 3, 5, 7, 4, 5, 4, 2, 7, 6, 4, 0, 5, 7, 6, 7, 3, 2, 5, 3, 7, 6, 4, 6, 2, 4, 7, 7 | 153 |
| sales forecasting | 1 | #2881 | 1, 0, 1, 1, 2, 0, 3, 2, 2, 0, 1, 0, 1, 0, 1, 2, 0, 4, 1, 3, 1, 0, 2, 1, 1, 0, 0, 0, 1, 1, 2, 0, 3, 0, 0, 1, 1 | 38 |

## Cite this page

Scholar4AI. (2026). Demand forecasting with language models in the loop. Journal Research Rankings. Retrieved October 23, 2026, from https://scholar4ai.com/ideas/demand-forecasting-with-llms

```bibtex
@misc{scholar4ai-ideas-demand-forecasting-with-llms,
  author = {{Scholar4AI}},
  title = {Demand forecasting with language models in the loop},
  year = {2026},
  howpublished = {\url{https://scholar4ai.com/ideas/demand-forecasting-with-llms}},
  urldate = {2026-10-23},
  note = {Journal Research Rankings; data from OpenAlex (CC0); licensed CC BY-NC 4.0}
}
```

License: CC BY-NC 4.0 (https://creativecommons.org/licenses/by-nc/4.0/).