What is data-driven decision making and why is it important? Evidence and method (2026)

Companies that adopt data-driven decision making have output and productivity 5% to 6% higher than would be expected from their other inputs. That is the conclusion of the "Strength in Numbers" study by Erik Brynjolfsson (MIT), Lorin Hitt and Heekyung Kim, which analyzed 179 large publicly traded companies.

In this article we explain what data-driven decision making is, why it matters according to the evidence, which mistakes it helps avoid and a five-step method to apply it.

5% to 6%

Higher productivity

179 large companies

3%

Higher value added

more than 18,000 US plants

30%

Plants deciding with data in 2010

up from 11% in 2005

39.85%

EU companies that analyze data

2025

Source: Brynjolfsson, Hitt and Kim, 2011; Brynjolfsson and McElheran, 2016; Eurostat, 2025

View the data as a table

Higher productivity	5% to 6% | 179 large companies
Higher value added	3% | more than 18,000 US plants
Plants deciding with data in 2010	30% | up from 11% in 2005
EU companies that analyze data	39.85% | 2025

What data-driven decision making is

Data-driven decision making (DDDM) is a process in which decisions rest mainly on data and analysis, not only on experience or intuition. It is the core practice of a data-driven company.

It doesn't mean data decides on its own. It means that, before deciding, you define what evidence you need, measure it and use it to choose between options. For the full definition of the concept, see what data-driven means and how to apply it in your company.

Why data-driven decision making is important: the evidence

There are three key studies, all based on data from real companies:

1. Higher productivity. In the study by Brynjolfsson, Hitt and Kim, companies that decide with data have 5% to 6% higher output and productivity. The relationship also shows up in other indicators, such as asset utilization, return on equity and market value.
2. Higher value added. Using US Census data on more than 18,000 plants, Brynjolfsson and McElheran found value added about 3% higher in plants that adopted the practice. The improvement appears after adoption, not before.
3. Fast adoption. In the same research program, the share of plants deciding with data almost tripled in five years.

200511%

201030%

Source: Brynjolfsson and McElheran, American Economic Review, 2016

View the data as a table

2005	11%
2010	30%

The authors themselves note a limitation: the first study shows an association and does not prove on its own that data causes the improvement. The second, with more plants and before-and-after data, strengthens the case for a real effect.

The mistakes data-driven decision making helps avoid

Deciding by intuition alone exposes you to systematic errors of judgment. Psychologist Daniel Kahneman received the 2002 Nobel Prize in Economic Sciences for integrating psychological research into economics, especially regarding how we judge and decide under uncertainty. Three common biases in companies:

Bias	What happens	How data helps
Confirmation	We look for information that confirms what we already believe	Define in advance which result would change the decision
Anchoring	The first figure we hear shapes the decision	Compare with historical or market data
Availability	We overvalue what we remember easily, like the latest complaint	Look at all cases, not the anecdote

How to apply data-driven decision making in 5 steps

1. Frame the question. For example: "Should we raise the price of this product by 5%?"
2. Define what data you need and which result would change your decision. For example: "If sales drop more than 10%, we roll back."
3. Gather and check the data. Verify where it comes from and whether it is complete. To choose tools, see the most used data-driven platforms in data analysis.
4. Decide and document. Write down what you decided, based on which data, and what you expect to happen.
5. Measure the result and learn. Compare what happened with what you expected and adjust.

To get the whole team working this way, read our guide to data-driven culture.

The limits: when data is not enough

• Poor-quality data: if data is incomplete or badly recorded, the decision inherits the error.
• Correlation is not causation: two figures rising together doesn't mean one causes the other. Controlled tests (for example, comparing two versions of a page) help separate cause and effect.
• Too few sources: only 15.86% of European companies analyze data from three or more sources, according to Eurostat. A single source gives a partial view.