- The design of all analytics dashboard beings with a fundamental problem statement about the business.
- All analytics dashboard need to be actionable. Shows the user an alert and prompts the user to perform a follow up action
- The dashboard should leave nothing to user’s interpretation. The action to be taken needs to be very clear
- Each well defined user profile will come with his own pre-defined dashboard configuration.
- A single page dashboard that users do not even need to tweak or configure
- Have sales person understand what the user’s business question is and then design the dashboard exactly to support that narrowly defined need.
- Sales person to train the user on how to use the dashboard, what goes where and how to interpret the signals
- Show users only the shortlisted options for actionable items. However it can not be too minimal, need to give users the impression that a lot of work went on to get to the insight. User’s has nothing to triangulate against if no none-options are presented
- Reduce the number of user profiles we are catering to so as to get to the 80% faster use case faster
- False positives and false negatives signals are a big issue when it comes to analytics dashboards. It becomes problematic when trying to monitor an environment that is very noisy such as COVID travel bubble since the landscape changes very fast based on situation
- Different profiles have different needs some profiles are not very comfortable when presented with a lot of quantitative signals. It is better to just have a IF / ELSE signal
- Palantir went down a very long journey of figuring out all the corner cases in a specific industry before finally generalizing the modules. It was only then did they achieve profitability
- A lot of companies in Singapore just survive on government handouts. Most just shut down the project when fail to get government grant
notes, learnings and reflections for future self reference