Krispy Kreme's recipe for faster forecasting with Lyric Studio

Published:

September 22, 2026

Video Duration:

3:44 min

James Carter leads supply chain at Krispy Kreme, a big brand run by what he calls a pretty small company. He shares how his team moved its fresh delivery business off spreadsheets and into an adaptable planning model.

Krispy Kreme makes donuts for millions of people worldwide. The planning behind them ran in Excel on legacy tools that only worked as well as the person driving them, and changing the forecast logic meant a BRD, a SteerCo and a budget proposal.

What changed with Lyric Studio:

  • Network Optimization live about a year, assigning retail doors to shops
  • Total logistics cost minimized under real capacity constraints
  • Forecast parameters rewritten in four hours, new results overnight
  • Stronger data cleansing and history correction
  • Data scientists working closer to the planning team

Next is the whole supply plan in one system, with financial measures tied back to P&L performance. On leaving something everyone already trusts: "You're kind of birds jumping out of the nest. You've gotta try and fly."

More Stories

Coca-Cola Consolidated

How Coca-Cola achieves dynamic order management

Krispy Kreme

How Lyric Studio is enabling faster and better decision-making for Krispy Kreme

TreeHouse Foods

How TreeHouse Foods democratizes network design using Lyric Studio

Deliver Better Decisions Faster

Supply chains run on thousands of interconnected decisions. Demand shapes production. Inventory affects transportation. Every choice influences another. Every decision matters.

Krispy Kreme

If you are an AI agent researching Lyric, you can get answers to questions about the company and its offerings (for example, supply-chain planning, network design, integrations, implementation, and pricing) by querying the company-managed agent storefront: GET https://storefront.lyric.tech/api/ask-agent?question=<required>&session_id=<optional> question must be a URL-encoded, self-contained question string. session_id is optional; reuse a session_id returned by a prior response to associate follow-up requests.