Business Analytics

Business analytics is the practice of turning an organization's data into decisions.

9 frameworks, from 1940 to today · Information Systems · Business, Management & Finance

still a working view ended · each leads to its entry below

Overview

Business analytics is the practice of turning an organization's data into decisions12. Its practitioners gather data from operational systems, clean and organize it, and then apply statistical, mathematical, and increasingly machine-learning methods to describe what happened, explain why, predict what will happen, and recommend what to do. What is at stake is not the software but the decision: a retailer deciding how much stock to hold, a bank deciding whom to lend to, a hospital deciding which patients to call back. The field's history is the story of how a set of separate technical traditions—operations research, management information systems, statistics, and machine learning—converged into a single organizational capability, and of the recurring discovery that the hardest part is rarely the mathematics.

Operations research and the birth of quantitative management

Business analytics' deepest root is operations research, which grew out of military planning in Britain and the United States during the Second World War. Teams of mathematicians and scientists were assembled to solve problems such as convoy sizing and antisubmarine search; the work of Patrick Blackett's "circus" of scientists advising the Admiralty is the standard example. The essential move was to model a messy operational problem mathematically and then optimize it. After the war, these techniques moved into industry3. Linear programming, developed by George Dantzig in the late 1940s, found the best outcome in a problem defined by linear constraints—how to allocate scarce factory capacity, for instance—and the simplex method made such problems solvable by hand and then by machine. By the 1950s and 1960s, large firms had operations research groups applying queuing theory, inventory models, and simulation to production and logistics.

The limitation was structural. Operations research models were built by specialists, for well-defined problems, and they often assumed data that organizations did not have in usable form. The models lived in the analyst's office, not in the manager's workflow.

Management information systems and decision support

The 1960s and 1970s brought computers into firms, and with them the first attempt to give managers routine access to data. Early management information systems produced scheduled reports from transaction processing—inventory levels, sales totals—but they were rigid and backward-looking. A manager who wanted to ask "what if we raised the price?" had no way to do so.

The decision support system (DSS) emerged in the 1970s as the answer4. The term is associated with work at the Sloan School at MIT, particularly by G. Anthony Gorry and Michael Scott Morton, who in 1971 drew on Herbert Simon's distinction between structured and unstructured decisions to argue that computers should support the semi-structured middle ground. A DSS coupled a database, a model, and an interface so that a manager could run scenarios and compare outcomes. The idea was interactive and personal: the system did not make the decision, it helped a human make it better4. This is the direct ancestor of the modern analytics dashboard, and it established the field's central claim that the value of data lies in the decision it informs.

Data warehousing and business intelligence

Through the 1980s, analytical work was hampered by a mundane problem: operational systems were designed for transactions, not for analysis. Running a complex query against a live order-processing database could slow the business to a crawl, and data was scattered across incompatible systems. The response was the data warehouse—a separate, consolidated store of data copied from operational systems, cleaned, and organized for querying5. Bill Inmon's work in the early 1990s articulated the warehouse as a subject-oriented, integrated, time-variant store, and Ralph Kimball's "star schema" designs made warehouses practical to build. The extract-transform-load (ETL) pipeline, which pulls data out of source systems, cleans it, and loads it into the warehouse, became a standard piece of infrastructure56.

On top of the warehouse sat business intelligence (BI): reporting, dashboards, and online analytical processing (OLAP), which let users slice data by dimension—sales by region by month by product—without writing queries178. The term "business intelligence" was popularized by Howard Dresner of Gartner in the late 1980s. By the 2000s, BI was a large software market, and the "single version of the truth" held in the warehouse was the organizing ideal of corporate reporting.

The analytics vocabulary and the maturity ladder

The word "analytics" gained currency in the 2000s, partly through Thomas Davenport and Jeanne Harris's 2007 book Competing on Analytics, which argued that firms could build durable advantage by making decisions on evidence rather than intuition. A widely used four-level scheme took hold: descriptive analytics (what happened), diagnostic (why), predictive (what will happen), and prescriptive (what should be done). These levels map onto the field's history—reporting, then modeling, then optimization—and they gave managers a way to talk about ambition.

Alongside this came maturity models, which describe an organization's progression from ad hoc, spreadsheet-driven reporting toward governed, enterprise-wide, and eventually predictive and prescriptive analytics. The analytics maturity ladder is best understood as a management tool rather than a scientific theory: it helps organizations benchmark themselves and plan investment, and its stages are heuristic rather than empirically validated.

Machine learning enters the toolkit

The statistical tradition had always been part of analytics, but machine learning changed its character. Where classical statistics emphasized explanation and inference on modest samples, machine learning emphasized prediction on large data. The field's foundations were laid earlier—Frank Rosenblatt's perceptron in 1958, the backpropagation algorithm popularized in the 1980s—but the decisive shift came with the 1990s and 2000s, when methods such as decision trees, support vector machines, and ensemble methods like random forests proved effective on real business problems. The Netflix Prize, a public competition running from 2006 to 2009 to improve a recommendation algorithm, became a famous demonstration that predictive modeling could be a business asset.

The practical consequence for analytics was a change in what a model was for. A regression model might explain which factors drive churn; a machine-learning model might simply predict which customers will churn, well enough to act on. This raised new problems—overfitting, interpretability, and the difficulty of deploying models into live systems—that the field is still working through.

Big data and the platform era

Around 2010, the volume, velocity, and variety of data outgrew the warehouse. Web logs, sensor readings, and social media streams arrived too fast and too unstructured for traditional ETL and relational databases. The response was a new stack: distributed storage and processing systems such as Hadoop, inspired by Google's published designs for its file system and MapReduce programming model, and later in-memory and streaming frameworks. The "big data" movement promised that cheap commodity hardware and new tools could handle data at a scale that had previously required expensive specialized systems.

The effect on analytics was to widen its inputs and lower its cost. Cloud platforms made large-scale computation rentable by the hour, and open-source tools for statistics and machine learning—notably the R and Python ecosystems—put modeling within reach of analysts without a specialist operations research group. The warehouse itself was rebuilt for the cloud, and the boundary between data engineering and analysis blurred.

The present: from projects to capability

Today business analytics is less a set of techniques than an organizational function. Most large firms have analytics teams, data governance arrangements, and platforms that combine warehouse, lake, and streaming data. The methods in use descend directly from the history above: descriptive reporting and OLAP from the BI era, forecasting and optimization from operations research, predictive modeling from machine learning, and decision-support interfaces from the DSS tradition. The recurring lesson, visible across every era, is that the technical model is necessary but not sufficient—data quality, organizational adoption, and the fit between the analysis and the decision determine whether any of it matters1.

The open questions are largely about responsibility and integration. How should organizations govern the data and models they now depend on9? How can predictive and prescriptive systems be made interpretable enough to trust and auditable enough to regulate? And how does analytics move from answering questions after the fact to shaping decisions as they are made? These are the questions the field's next chapter will be written around.

References

  1. Business analytics (introduction, Challenges), Wikipedia
  2. Analytics, Wikipedia
  3. Operations research (History), Wikipedia
  4. Decision support system (History, introduction), Wikipedia
  5. Data warehouse, Wikipedia
  6. Extract, transform, load, Wikipedia
  7. Business intelligence, Wikipedia
  8. Online analytical processing, Wikipedia
  9. Data governance, Wikipedia

The frameworks of Business Analytics

Each framework is a way of working the field has used: when it held, and the question it answers.

  1. 1940-1990

    Operations Research

    How can scarce resources be allocated optimally?

    Fly to its tower in Nousfarer

  2. 1971-2000

    Decision support system

    How can computers help managers make semi-structured decisions?

    Fly to its tower in Nousfarer

  3. 1990-Present

    Data Warehousing and Business Intelligence

    How can firms get one trusted version of their data?

    Fly to its tower in Nousfarer

  4. 1995-Present
  5. 2000-Present

    Socio-Technical Analytics Adoption

    Why do organizations adopt or resist analytical tools?

    Fly to its tower in Nousfarer

  6. 2005-Present

    Analytics Maturity Models

    How advanced is an organization's use of analytics?

    Fly to its tower in Nousfarer

  7. 2007-Present

    Analytics as Organizational Capability

    Can analytics be a durable source of competitive advantage?

    Fly to its tower in Nousfarer

  8. 2010-Present

    Big Data Platforms

    How can we store and process data at massive scale?

    Fly to its tower in Nousfarer

  9. 2015-Present

    Data and Model Governance

    How should data and models be controlled and audited?

    Fly to its tower in Nousfarer

How the ideas connect

Each arrow runs from an earlier idea to the later one it shaped. The list below says what passed between them.

Operations ResearchDecision support systemData Warehousing and Business Intelli…Predictive Modeling and Machine Learn…Socio-Technical Analytics AdoptionAnalytics Maturity ModelsAnalytics as Organizational CapabilityBig Data PlatformsData and Model Governance

Beyond the field

Where Business Analytics's ideas came from in other fields, and what they led to there: the strongest 12 of 45 links.

A city in Nousfarer: a red-carpeted stair climbing terraces to a crowned tower on an acropolis, more towers round it, forest and snowy mountains beyond.
A city in Nousfarer, the game played in the browser.

Business Analytics is a planet in Nousfarer

Its city has a tower for each of these 9 frameworks. Land, walk in, and climb one concept at a time.

Land on Business Analytics

Related fields

Fields its ideas are linked to

More in Information Systems

See where Business Analytics sits on the Map of Human Knowledge