It’s time to decide: You know the problem, proposed solution, and how to measure success. So, continue or not? Stopping now is cheap. Stopping after the data work, or in production, won’t be.
A CPMAI can frame this decision for you with nine questions split into three groups: business feasibility, data feasibility, and execution feasibility. Together the answers come down to one question: Is the reward worth the risk?
The review activity
The activity uses a stoplight grading for each question:
- Red = High risk. Right now, a blocker or a material impact if the risk manifests. The project needs a work package to address any red before other work goes ahead.
- Yellow = Caution. You may decide you can carry this while working the project, but it’s on the risk register and regularly monitored.
- Green = Go for it. Projects are not slam dunks, but there’s one less worry for now!
You want to consider all the reds and yellows especially. The worksheet in the sidebar can help you review them. The reds and yellows represent the blockers and risks your iteration faces. Is it possible to continue? If the risks occur, are the impacts bearable? These require your decision as an executive.
Keep in mind that an overall “no” is not a judgment on the idea. At this point the reds and yellows represent “known unknowns.” How to address them is your next strategic move.
If you have multiple reds, drastically reshaping or even shelving the project is a serious consideration. Also remember that perfect knowledge is not a practical goal; there will always be some ambiguity. The key question is whether you know enough to act with confidence.
If you have multiple yellows, it will help to take a very high view: Is there something in the team organization, business-as-usual process demands, or technological platforms that is having a broad impact? A “five whys” root cause analysis would be recommended at this point to try and discover it.
The nine questions
Business feasibility focuses on whether the business conditions for success exist.
| Question | How to answer | |
|---|---|---|
| 1. Do you know the problem? | State it clearly, in terms of business objectives. | |
| 2. Will the organization invest and change? | The business owner commits, verbally or in writing, to putting the result into production. | |
| 3. Is the return worth it? | Apply ROI analysis, as with any business venture. |
Data feasibility focuses on whether the data can support the project.
| Question | How to answer | |
|---|---|---|
| 4. Does the data measure what you care about? | Name the data the solution needs and the outcome it measures. | |
| 5. Is the data available and accessible? | Confirm there is enough of it and that you can get to it. Who is the business owner? | |
| 6. Is the data good enough? | A best guess is fine for now; Phase 2 will evaluate more fully. |
Execution feasibility focuses on whether the project can be built and run as required.
| Question | How to answer | |
|---|---|---|
| 7. Do you have the technology and skills? | Identify from where the tech and skills will come and the funds and approvals to use them. | |
| 8. Can it run fast enough? | Fast enough for the problem, without a major new investment in infrastructure? | |
| 9. Does it make sense where you plan to use it? | Technically, operationally, and financially. |
An example
Continuing with the example of a research chat agent, Morty, the following are example answers:
| Question | Morty | R | Y | G |
|---|---|---|---|---|
| 1. Do you know the problem? | Staff time spent searching and building indices; three goals written down | ✓ | ||
| 2. Will the organization invest and change? | The broker worked on the prototype with me; staff already used Discord | ✓ | ||
| 3. Is the return worth it? | Staff time is the cost; for a small office, verbal surveys and chat channel stats are the measure; the prototype beat PDF search | ✓ | ||
| 4. Does the data measure what you care about? | The regulations themselves | ✓ | ||
| 5. Is the data available and accessible? | Public documents the office already had | ✓ | ||
| 6. Is the data good enough? | Dense; revisit in Phase 2 | ✓ | ||
| 7. Do you have the technology and skills? | No-code prototype, then a hosted bot with a vector database and an LLM service | ✓ | ||
| 8. Can it run fast enough? | Criterion: a research response within 1 minute | ✓ | ||
| 9. Does it make sense where you plan to use it? | Fits where used: yes, in Discord. | ✓ |
Eight green, one yellow. Go. Data quality gets a second look in Phase 2, Data Understanding, when more is known.
Immediately following the Go
Refine the project plan, then on to Phase 2, Data Understanding. The plan for the current iteration lists the stages ahead, with duration, resources, inputs, outputs, and dependencies for each. It also notes the risks to the schedule, where phases may repeat (modeling and evaluation often do), and the initial choice of tools.
To Do Next
- Use the attached worksheet. A vague answer is a yellow.
- Review it with your CPMAI.