EN
Back to the archive

The encyclopedia · Software & IT · Operational decision · 1984–1986

Citgo wrapped an optimizer in a knowledge base so managers could ask why

Citgo paired a supply, distribution and marketing optimizer with an expert system, adding about $70m a year in profit.

Citgo Petroleum Corporation · Southland Corporation

the move

Citgo Petroleum, the nation's largest independent downstream marketer of refined products with about $3.8 billion in 1985 sales, faced short-term planning and operational problems spanning supply, distribution and marketing.

The firm built SDM, an intelligent decision-support system that combines an optimization-based model, a state-of-the-art database and a modeling language with a knowledge-base component that uses artificial intelligence and expert-system techniques.

The knowledge base lets analysts and managers verify and modify inputs and explore what-if and why questions about the solution, and the deployed system helped produce a profit improvement documented at roughly $70 million a year and changed how the company did business.

why it works

  • The supply-distribution-marketing problem crosses too many functions to solve by hand.
  • A bare optimizer gives an answer with no way to question it.
  • A knowledge base makes the solution explainable and testable, so it is trusted.
  • Measured: about $70 million a year of profit improvement.
the payoffMake the optimizer answer 'why', not just 'what'clever

what transfers

An optimization model is adopted only if the people who must trust it can ask why it says what it says; pair the optimizer with a knowledge base that explains, verifies and stress-tests its answers.

what came after

The SDM system became central to Citgo's planning and was recognized with the 1986 Franz Edelman Award; the project showed how an expert system can make an optimization model usable in practice rather than remain a desk study.

references

spotted an error? The archive wants to know.

same kind of clever