For the past 12 years, Steven has utilized his extensive knowledge in business intelligence and machine learning to offer clients a unique dashboard experience, which is often absent in today's ever evolving analytics environment.
Traditional reports typically focus on descriptive analytics, which provides businesses with valuable insights into current and historic trends and metrics. However, with the growing importance of predictive, simulation, and optimization modeling to gain greater insights, traditional reporting methodologies have only partially evolved to support such analysis.
Reporting teams have evolved to support advanced models such as machine learning and simulation to help stakeholders answer more complex business problems. However, these teams struggle to make these models interactive and easy to use for end-users. One common approach to allowing users to interact with model output is to generate output for all cuts of the data. For example, if a model can be run globally and for each region, the model will be re-executed for each region and its output will be merged together in the final output file. Then, users can select specific combinations of filters on a dashboard to retrieve pre-calculated model output. While this approach may be advisable in certain situations, it requires additional computational resources and larger output files to pre-calculate all possible scenarios for a user to select. Moreover, often scenarios are excluded from dashboards due to an inability to account for endless filter and parameter combinations, including continuous variables. To address these challenges, Steven created Interactive Models to set the standard for the future of interactive dashboards.
Steven works with clients to develop artificial intelligence and machine learning models that can be seamlessly integrated into advanced dashboards, providing end-users with the ability to adjust model parameters and assumptions in real-time and view output. Previously, models that required data scientists to re-run with stakeholder parameters and assumptions can now be executed in real-time by end-users, supporting an infinite number of scenarios. When paired with advanced dashboards, Steven's Interactive Models contribute to the longevity of artificial intelligence and machine learning models by repurposing past investments to support future initiatives. This approach enables organizations to extract additional value from existing models and continue leveraging them to enhance their analytical capabilities.
Steven has worked closely with clients to seamlessly integrate the following models into interactive dashboards:
- Predictive Models
- Neural Networks
- Supervised and Unsupervised Machine Learning Models
- Simulation Models
- Optimization Models
- Forecasting Models
These models offer a wide range of capabilities and can be tailored to suit specific business requirements, while remaining dynamic to support endless scenarios. Applications vary based on the client's environment, but often involve a combination of Tableau and Tableau Server (REST API and External Client integration), Power BI, Python, and/or R-Shiny, as well as DataIku, Data Bricks, and other cloud services for always-on model requirements.
View some of the examples below of Steven's Interactive Modeling and envision the potential applications for your projects.
Tableau Calculations Take on a New Form with Steven's Interactive Models
If you are proficient in Tableau Desktop, you still may not recognize the calculation syntax shown above. That's because the calculation above is specifically designed to initiate an external client call to the client's Python-Server, in order to execute an embedded risk model based on user-entered dashboard parameters. In this particular example, the end-user is updating risk thresholds for a Value at Risk (VaR) model, which is a widely used measure of risk that estimates potential losses that could arise under adverse market conditions, as determined by the user. Using dashboard parameters, the user can adjust the confidence level and time horizon for VaR calculations to reflect their specific risk appetite and modeling requirements. After the user inputs are sent to the client's Python-Server for model re-execution, the end-user can view updated charts and metrics based on their entered scenario.
To ensure the longevity of Steven's Interactive Models, Steven works with clients to adopt frameworks so model-driven calculations are centrally owned and governed. This approach allows for the cascading of the calculations' logic to all Tableau Dashboards that use them as new data sources and features are introduced, improving model accuracy and usefulness.
Steven's Interactive Models are custom-designed solutions that make it easy for end-users to engage with advanced modeling techniques without any specialized knowledge, aside from dashboard training. Watch Steven’s videos below to gain a better understanding of his unique Interactive Modeling approach.