Analytics models are centered on the “protagonist”: the beneficiary or customer of the program.
Start by mapping the steps the protagonist must take to reach a desired outcome. Model the “events” the protagonist must experience or perform to move from their current state to the desired state. Then add the activities that make the protagonist more likely to reach that outcome.
Each event or activity block is a discrete step in the pathway, and can be given a quantitative input-output relationship, a probability, or a modifier function. The result is a diagram that clearly communicates the program’s logic and how it solves the problem, with the added ability to predict impact, run sensitivity analysis, and estimate uncertainty.
There are four essential pieces for analytics modeling:
Social impact work should begin with a focus on the individual being served or helped, much like human-centered design. In analytics modeling, the client, customer, beneficiary, or target population is called the “protagonist.” This is the human at the center of the design.
Next, the modeler identifies the desired outcome for the protagonist. The outcome must include either a direct, objective measurement or a measurable surrogate indicator.
Event blocks define the Protagonist Pathway: the series of conditions that must be true for the protagonist to reach the desired outcome. The pathway frames the problem from the protagonist’s perspective by asking what must happen for that outcome to occur. By examining these required steps, the modeler can see where the pathway breaks down and where an intervention may be needed.
Once the problem is framed, the modeler begins to model the theory of change: one or more activities that modify the step in the Protagonist Pathway where the intervention is meant to act.
For more complicated systems, these additional blocks can help:
This is where resources, such as money or labor, enter the model. Impact is a measure of the outcome produced by those input resources.
An explicit relationship that shows the probability of an event occurring when two or more previous steps must all be true.
An explicit relationship that shows the probability of an event occurring when any one of two or more previous steps is true.
Makes it explicit that a following event acts on the case where a previous step did NOT occur.
Lets the modeler divide resources or protagonists onto different pathways.
Here is how the pieces combine, step by step, in an example bed net program:
Description coming soon.
Description coming soon.
Open the demo, fork a template, and replace the numbers. No install, no SDK, no spreadsheet duct tape.