Managing Expectations

From The Foundation for Best Practices in Machine Learning
Technical Best Practice Guideline > Managing Expectations
Jump to navigation Jump to search


Hint
To view additional information and to make edit suggestions, click the individual items.

Managing Expectations

Objective
To effectively set and communicate realistic Product expectations to Stakeholders and obtain their buy-in.
Item nr. Item Name and Page Control Aim
9.1. Performance

Product management should attempt to set realistic Product performance expectations for Stakeholders through periodic stakeholder discussions on the following issues: (i) limited industry understanding of what tasks are difficult for the Product; (ii) difficulty of determining what type of modifications - network design, input Features, or training data - will create the greatest Product improvement; and (iii) Model improvement can stall significantly while experimenting with different variable modifications.

To (a) effectively communicate realistic Product performance expectations throughout the development process; and (b) highlight associated risks that might occur in the Product Lifecycle.

9.2. Timeframe

Product management should set expectations for long-term investment in the Product for Stakeholders, specifically focusing on: (i) the unpredictability of Product improvement; (ii) Product difficulties are traditionally hard to diagnose as they are often caused by subtle issues of intersecting inputs; and (iii) it is possible for the Product to completely stall with absolutely no discernible improvement in spite of significant time and effort.

To (a) effectively set Stakeholder expectations regarding the difficulty of locking down Product timelines; and (b) highlight associated risks that might occur in the Product Lifecycle.

9.3. Accuracy perception

The Product Team should work to ensure that the solution will be accurate enough to meet a variety of different Stakeholders' expectations, recognizing that each group of Stakeholders will have different views on what is 'accurate' based on their interaction with the Product. Product management should set and communicate expectations in-line with the achievable level of accuracy for each user group.

To (a) effectively communicate achievable accuracy levels, considering individual Stakeholder accuracy preferences; and (b) highlight associated risks that might occur in the Product Lifecycle.

9.4. POC-to-Production

The Product Team should effectively communicate that infrastructure is often the determining factor for the success of the POC-to-Production transition and rely heavily on the POC-to-Production Checklist, as discussed in Section 6 - Management & Monitoring, to set and align Stakeholder expectations of the transition process. The Product Team should set the expectation, before beginning the transition process, that novel problems will likely arise during the transition that may significantly affect the timeline and costs. The Product Team should be on alert for integration issues arising close to the final release of the solution, which the Product Manager should communicate to relevant Stakeholders, along with progress updates, at a progressively more frequent cadence.

To (a) uncover and communicate issues that may delay the transition of the solution from POC-to-Production or make that transition less feasible; and (b) highlight associated risks that might occur in the Product Lifecycle.

9.5. Production Costs

Review and analyze the finalized POC budget to determine a realistic Product implementation budget. Review the Product Cost Analysis as discussed in Section 4 - Problem Mapping to ensure its continued accuracy and reformulate as necessary. The Product Team should effectively communicate to Stakeholders that the budget for implementation will likely be in-line or more expensive than the cost to get through POC.

To (a) ensure realistic expectations for a sufficient Product implementation budget are communicated; and (b) highlight associated risks that might occur in the Product Lifecycle.