I’m developing a return on investment (ROI) calculator for data warehousing (DW) appliances, using the Forrester Total Economic Impact methodology.


At the heart of that is a conceptual ROI model that can be applied to any decision support infrastructure, not just DW appliances (though indeed high-quality decision support is the raison d’etre for DW appliances).


That said, and not wanting to bog down forthcoming syndicated TEI study with a lot of this conceptual material, here are the core principles of this conceptual model , plus a discussion of how, net-net, they map to the key benefits of a DW appliance:


  • In the attention economy, every decision is a business investment: Every decision is an investment of attention by a decision agent, also known as an information worker.
  • Every bad decision has a calculable do-over cost. The basic monetary value of a decision to an organization is the cost of provisioning that decision agent (i.e., his or her fully burdened annual salary) divided by the number of decisions that he or she makes in a year, which is the same as the replacement cost of that decision (i.e., the do-over cost of that person’s time if they make the wrong decision based on an suboptimal decision-support infrastructure, such as one with no data warehouse or an underperforming DW).
  • Every optimal decision has a calculable monetary return. The potential monetary value of a decision to an organization is the expected top-line payoff from making the right decision with the optimal decision support infrastructure, especially one with a high-performance DW appliance). The average decision, like any investment (in this case, of an information worker’s attention), will yield payoffs in line the internal rate of return, factored against the basic monetary value (fully burdened salary divided by number of decisions per year).
  • Every decision can be optimized by supporting it with decision-support infrastructure that enables enhancements in the following key areas:

    • Decision velocity: This depends on the number of questions that a user can pose (via queries, reports, dashboards, etc.) against an information set within a given time period prior to making decisions and taking any subsequent actions. Decision velocity depends on the speed and scale of the available analytics infrastructure. This depends, in turn, on the scale of the underlying DW infrastructure. A DW appliance expands decision velocity through scalability and performance features that enable more questions to be asked against more information in a shorter period of time. The key DW scalability features in this regard are:

      • Query response: more queries, reports, charts, dashboard elements, and other view are possible per minute per user per deployed DW appliance instance
      • Mixed query workloads: broader variety of concurrent queries, reports, charts, in-database model functions, and other analytical query workloads can be executed in parallel per deployed DW appliance instance
      • In-database analytics: greater volume, variety, and complexity of predictive models can be scored, executed, validated, and otherwise processed in parallel per deployed DW appliance instance
      • Pipeline processing capacity: greater volume, variety, complexity of extract, transform, cleansing, loading, and other data management pipeline functions can be executed in parallel per deployed DW appliance instance
      • Data capacity: greater volume and variety of usable information (compressed) can be stored and managed per deployed DW appliance
      • Load capacity: greater volume and frequency of source-data loads, both batch and real-time, can be processed per day per deployed DW appliance instance
    • Information quality: This depends on the extent to which the user can access a decision-relevant information set that has the following characteristics:

      • Consolidated: all relevant information has been consolidated into a physically and/or logically integrated analytic database
      • Conformed: all relevant information has been virtualized and harmonized to a common data model, semantic model, vocabulary, schema, dimensions, and hierarchies across all subject areas
      • Cleansed: all relevant information has been transformed, matched, merged, corrected, and enhanced prior to loading into the analytical database
      • Current: all information has been extracted, processed, and delivered in real-time from source applications to queries and other applications
      • Multidimensional: all information has been delivered into a full range and mix of query, reporting, dashboarding, search, visualization, predictive models, and other analytic apps in support of flexible multidimensional exploration
      • Deep: all information embodies a depth of summaries, aggregates, materialized views, and details, as well as a depth of historical, longitudinal, and time-series information
  • Every DW appliance supports optimal decision making.A DW appliance, as an engine of decision support, expands the velocity of decisions that a given information worker can make in a year, thereby reducing the basic value (replacement cost) of each decision. However, a DW appliance, by improving decision velocity and information quality, also enables each decision to reach its potential value. It does so by enabling information workers to take into consideration more relevant options and leverage more current, correct, and comprehensive information than if there were no DW appliance.
  • Every DW appliance has a calculable potential decision-support productivity benefit. A DW appliance, as a decision-support investment, delivers productivity enhancements by enabling, for any given set of users, aggregate decisions to reach their expected potential value. The potential productivity payoff is the aggregate cost of provisioning all decision agents (i.e., the sum of fully burdened salaries of all users who require access to the DW appliance) multiplied by the internal rate of return.
  • Every DW appliance maximizes productivity through scale and quality.A DW appliance can only help users achieve full expected decision-support potential if the appliance enables balanced, order-of-magnitude scaling over prior DW or pre-DW analytic environments, and only through implementation a comprehensive, multifaceted information quality initiative.


Those who are paying attention will notice the conceptual hooks that would allow me to extend this beyond decision support environments to also address the ROI of decision automation and decision management environments.


What do you think? I’d love your feedback as I tune my TEI model.