Building Ventana Models

Ventana® models are usually built through a process of rapid evolution with client participation. Improvements in modeling technology now allow major model enhancement and testing in hours to days, so the client can see and critique progress frequently and regularly.

Rapid evolution with frequent client feedback brings important project benefits: • The client can regularly inspect inputs, outputs, and detail granularity of the model to ensure relevance to the target decisions. • The client can contribute knowledge toward ongoing tests of the model framework, improving reliability. • Client participation in the model evolution develops familiarity and expertise, which are the basis of model credibility. • As client personnel become familiar with possible uses for the model, they can direct choices of data management, interface, and business process for best fit.

For more information on how Ventana's modeling process meets these objectives, please see Why Ventana Models Work.

Client Resources

Generally, two groups greatly benefit from regular participation: the decision team, and the model team.

• The decision team has direct responsibility for making the decisions supported by the model. These people typically participate in half-day meetings to guide model development once every 3 to 6 weeks over a 3 to 5 month period. Clients often feel these meetings, which focus on clarifying the outcomes to be measured and on the nature and requirements of decision-making, are extremely useful in and of themselves, independent of their contribution to model development.

• The model team has direct responsibility for model development and its implementation in the organization. During model development, these people work closely with Ventana to develop thorough knowledge of model capabilities, identify and provide relevant data, arrange meetings with stakeholders and knowledge resources, find answers to daily questions, and contribute their own expertise and critique. After delivery, the model team takes the lead to ensure the decision team gets full benefit of the tool.

Sometimes, the two teams are the same people. Sometimes a "team" is one person. In every case, however, the people in these roles develop a clearer sense of their business and their opportunities.

In addition to these people, Ventana invites to interviews or meetings client personnel who can provide useful information or data, and all other stakeholders in the model or the decisions it supports. These interviews and meetings, organized by the model team, not only ensure that the model profits from the broadest range of client knowledge, but also foster organization-wide involvement in and ownership of the resulting business intelligence.

Project Elements

Five elements work together to keep the delivered model up to date and connected to ongoing decisions:

• Knowledge - a documented and tested collection of client expertise on how each factor will affect other factors
• Data - a reviewed collection of historical data, and a system to manage ongoing data to connect to the model
• Model - the system of equations used to calculate the outcomes associated with decision options
• Business Process - The job descriptions, schedules, and assigned responsibilities by which the model gets used to inform decisions
• Interface & Training - The system connecting the model to human users or to automated decision systems, and any required training

Project Sequence

In a typical project, the five project elements - Knowledge, Data, Model, Business Process, and Interface & Training - proceed in parallel by iteration and critique. A specific plan for each project is drafted at project start, with milestones evaluated by the client at each review. The figure represents a typical plan.

Step 1: Project Start
A facilitated workshop with the decision team and the model team to introduce the project and complete the project definition

Agenda:
• Introduce decision team and model team to the project objectives and project plan.
• Begin work on project elements:

° Knowledge: Collect initial expectations on cause-and-effect mechanisms active between actions and outcomes.
° Data: Identify and discuss relevant data sources.
° Model: Develop first-pass structure, focusing on decisions to be made and outcomes to be managed.
° Business Process: Determine who will maintain and run the model, and how questions of and insights from using the model will be communicated.
° Interface & Training: Make a first assessment of requirements to support the business process design.

For more information on possibilities for knowledge and data management, business process, interface, and training, please see Using Ventana Models.

Step 2: Collect Information
In collaboration with the model team, begin shaping all five project elements:

Knowledge: Initial collection
• Conduct one-on-one or group interviews of people with direct experience of how things work and why things happen the way they do, possibly including the decision team.
• Begin recording expertise in Reality Check(r) libraries.

Data: Initial data examination
• Collect relevant data.
• Interview people with expertise on how and why data were collected and what potential problems they may contain.
• Begin "cleaning" process:

° Aggregate to level of detail appropriate to modeling goals.
° Compare different data sets to identify and understand inconsistencies.
° Understand and document likely biases and uncertainty in measurements.
° Prioritize unmet data needs to pursue.

Model: Begin forming rough draft of model equations from initial descriptions of decisions and outcomes to be managed.

Business Process: Initiate further conversations as required to establish business process for model use.

Interface & Training: Begin interface development and draft initial plan for training.

Step 3: Enhance System
Prepare a working draft of all elements for client review.

Knowledge: Track down and clarify any conflicts with data or other inconsistencies.

Data: Initialize a rough system for management and use of the data collected to date. Refine the data set and the data management system in subsequent revisions.

Model: Build a working prototype, including:
Decision levers
• Outcomes / outputs
• All major relevant aspects of the business situation.
In subsequent revisions, increase fidelity and robustness.

Business Process: Develop conversations and procedures to define the process.

Interface & Training: Enhance interface to meet need. Training ongoing as appropriate.

Step 4: Review System
The review is ideally a facilitated workshop for the decision team and the model team together.

Data & Knowledge Report:
• Review any findings with significant policy implications.
• Discuss any major shortcomings and options for redress.

Model Critique:
• Review behavior of current model draft in different scenarios.
• Discuss causes of behavior and judge realism.
• Examine current framework for specifying decisions and observing outcomes.
• Evaluate for usefulness.

Business Process Review:
• Outline planned business process and plan for implementation.
- Discuss revisions if necessary.
Interface & Training Review:
• Discuss interface options for breadth of application, and for focusing output to highlight the most cared about results.
• Progress report on training and discussion as required.

Step 5: Handoff
With the model, Reality Check library, data management, and interface complete, and with client personnel oriented and trained to apply the model to decisions, a project typically ends in a facilitated workshop with decision team and model team. Key findings to date are presented, possibly by model team. This marks the beginning of using the model for ongoing decision support.

Deliverables

Project deliverables are determined with the client at project start, but typically include:
• Reality Check archive of documented organizational knowledge,
• Cleaned historical data set appropriate for model use with process in place for ongoing upkeep,
• Model, usually documented equations in Vensim software,
• Documented data & knowledge management system and user interface, and
• Documentation of planned business process for using model to improve decisions.

Project Duration

The first working model prototype usually appears 1-3 weeks after project start, typically followed by several rounds of testing, critique, and enhancement. The amount of time to a finished model depends greatly on the scope and intensity of the effort. Past projects have ranged from 6 weeks to 12 months, with most projects lasting 4-6 months. Scope, intensity, and schedule are all developed with and approved by the client before project start.