Plan to Predict (P2P)
In most organisations data collection is siloed, with different stakeholders determining what data is relevant or what end value it can generate. The Decision Cycle concept emphasises that the utility of data lies solely in its capacity to fuel predictions models that lead to decisions and actions that generate measurable value for meeting an organisation’s goals and objectives. P2P is an essential process that determines to what degree an organisation’s data and current usage meet these goals. It establishes how data is used in existing Decision Cycles, which then serve as a benchmark against which improvements can be objectively measured. The predictability in the data for a given Decision Cycle is quantified by generating Hybrid Intelligent prediction models and using their performance to estimate the potential ROI that can be generated by the corresponding improved Decision Cycles. If the cost-benefit is favorable the models can be incorporated into an Intelligent Middleware which can then be integrated into current systems and workflows.
The P2P process consists of the following elements:
- Specification of the required organisational goals and corresponding KPIs that are used to measure them.
- Analysis of the available actions (“interventions”) that are used or could be used to reach the goals.
- Analysis of current key decision points associated with the actions.
- Analysis of how predictive models using Artificial Intelligence and/or Human Intelligence are currently used to support decisions.
- Evaluation and analysis of the organisation’s data to determine its degree of predictability with respect to the desired goals.
- Estimation of potential ROI that could be generated through using the data to power Hybrid Intelligent prediction models.
- Road map for the potential creation of Intelligent Middleware and integration into the organisation’s Decision Cycles.
The process requires an intensive collaboration between Presage’s team and key stakeholders in the organisation and typically takes 6-12 weeks depending on the complexity of the organisation and its data. It can represent between 50% and 75% of the work needed to create a production-level prediction model.