Database Reference
In-Depth Information
provided “$8.5M in annual cost savings” is much more compelling than
saying it has “great value.”
•
Make the benefits of the project clear and conspicuous:
After
calculating the benefits of the project, make sure to articulate them clearly
in the presentation.
12.2.10 Providing Technical Specifications and Code
In addition to authoring the final presentations, the team needs to deliver the
actual code that was developed and the technical documentation needed to support
it. The team should consider how the project will affect the end users and the
technical people who will need to implement the code. It is recommended that
the team think through the implications of its work on the recipients of the code,
the kinds of questions they will have, and their interests. For instance, indicating
that the model will need to perform real-time monitoring may require extensive
changes to an IT runtime environment, so the team may need to consider a
compromise of nightly batch jobs to process the data. In addition, the team may
need to get the technical team talking with the project sponsor to ensure the
implementation and SLA will meet the business needs during the technical
deployment.
The team should anticipate questions from IT related to how computationally
expensive it will be to run the model in the production environment. If possible,
indicate how well the model ran in the test scenarios and whether there are
opportunities to tune the model or environment to optimize performance in the
production environment.
Teams should approach writing technical documentation for their code as if it
were an application programming interface (API). Many times, the models become
encapsulated as functions that read a set of inputs in the production environment,
possibly perform preprocessing on data, and create an output, including a set of
post-processing results.
Consider the inputs, outputs, and other system constraints to enable a technical
person to implement the analytical model, even if this person has not had a
connection to the data science project up to this point. Think about the
documentation as a way to introduce the data that the model needs, the logic it
is using, and how other related systems need to interact with it in a production
environment for it to operate well. The specifications detail the inputs the code
needs and the data format and structures. For instance, it may be useful to specify
