The Optimization Effect: why 84% of companies still make decisions with Exce

Having data is no longer enough. The real competitive advantage lies in turning it into better, faster and more traceable decisions.

That was one of the main conclusions of The Optimization Effect, the event organised by DECIDE | Linkroad to analyse how artificial intelligence and mathematical optimisation are changing the way companies address operational and regulatory complexity without falling into common market mistakes.

Below, we summarise the three strategic themes that defined the event:

1. Regulatory certainty as an investment asset

Contrary to the perception that regulation restricts development, the analysis of European and international regulation shows the opposite. A clear legal framework stabilises the long-term investment environment.

Before deploying complex solutions, organisations should ask themselves one essential question first: do we really need an AI system to solve this problem?

When it is needed, development must strictly comply with three fundamental requirements: a human-centred approach, shared responsibility throughout its lifecycle and full traceability to avoid opaque systems.

2. Excel and mathematical optimisation: allies, not substitutes

Data from the 1st Barometer of the State of Mathematical Optimisation in Spain and Portugal reveals a critical disconnect across the business landscape:

• 98.5% of organisations identify clear room for improvement in their operational efficiency.
• However, 84.5% still manage complex decisions through basic analytics or spreadsheets.

Excel serves its purpose: it is versatile, accessible and supports the day-to-day management of any organisation. However, its architecture is not designed to solve combinatorial problems. When critical variables such as capacity allocation, production planning, inventory or logistics routes intersect, the number of possible scenarios rises into the millions. At that point, spreadsheets lose their prescriptive capability.

Mathematical optimisation closes that gap. By integrating these analytical engines into daily operations, companies achieve initial returns on investment ranging from 5% to 20%. The viability of this technology is well established: 90% of the models developed move beyond the pilot phase and are successfully deployed in the company’s real production systems.

3. Models do not have to be perfect: they have to be useful

Another key learning from the event was the need to reduce the fear of complexity.

Today’s computing capacity and the maturity of algorithmic solving engines have completely removed the technical barrier to processing large volumes of variables.

For companies, implementation risk can be reduced with a clear strategy: move forward through pilot projects and outsource complex modelling to specialists instead of overloading internal teams.

The mistake of trying to design a perfect model that replicates 100% of reality from day one should be avoided.

Instead, the objective should be to model the 80% of variables that generate the real economic impact. This makes it possible to put the technology into operation quickly and safely, with the option to improve it step by step.