5 mins read
During my training in data school, we worked on several data projects. Each project involved requirement gathering with clients, understanding the dynamics of data, performing ETL operations and creating visualizations. The deadline for each project was 1 week. It is always a race against the clock, but if we follow the best practices, then success is warranted.
In this blog, I will be covering the best practices to follow while working on a data project:
1) Understand the requirement CLEARLY
The first step of creating a dashboard is to clearly understand the requirements. The golden rule to follow here is : Ask, Ask and Ask questions from stakeholders.
Ask these questions to the stakeholder:
- What is the business about ?
- Information about the datasets ?
- Do they have a data model ?
- Source systems and any access relation issues that need to be resolved ?
- Clients availability to answer any issues
2) Understand the data
We often make the mistake to start building the dashboard or figuring out insights from data without understanding the data. Understanding the data is the most important step of building a dashboard. This enables us see data at a wider angle and build more insightful dashboards.
Ask these questions to yourself:
- What is the data about?
- How many datasets do we have ? What are the facts and dimensions ?
- Do we have a data model for reference ?
- What is the grain of the data ?
- What is the relationship between datasets ?
- How is the Data Quality ?
- Are we able to figure out the approach on how we will be able to meet the clients requirements ?
- Are there any potential issues or blockers that we need to raise to the client in the next meeting ?
- A questionnaire for client
3) Set clear GOAL – building the Minimum Viable Product(MVP)
We usually get lost when we try to figure out the best visualization that the stakeholders will love. This is where I have often failed. The whole and sole purpose is to create a minimum viable product in dashboard first, get feedback from stakeholder and enhance it later. The idea here is FAIL FAST by getting constant feedback from clients.
4) Data Engineering work
By now, we will have a clear picture of the relationships and type of transformations that will be required to develop the dashboard. We use majorly either Tableau Prep or Alteryx for the work and create a final consumption view for our dashboards.
The most important question to ask: Are the numbers looking right ?
We often do the transformations without testing the final results. It is essential to show the correct numbers in the dashboard else no matter how beautiful the dashboard artwork looks, it will not be of any value if there are no business insights.
5) Data Visualization work
The best way to build a dashboard is to approach it by creating a sample design and building minimum efforts(bar graphs are fine) to meet the layout requirements. This is when we have to target creating an MVP than spending a lot of time making the best chart in the world. At the end of the day, the numbers or insights must help business to derive benefit out of it.
This is the stage I feel that I have reached a good state to improve my work, that includes:
a) Changing charts
b) Changing colours and fonts
c) Making sure that dashboard is interactive
d) Add any other insightful information that we can present to the client that they are not thought before
During the entire process, we work in teams, split the work, we collaborate, help each other, and try to present the best to the client on Friday presentation day.