Ten things to do (or don’ts) to succeed in a self-service BI project

Companies face a number of challenges as they try to spread BI widely among their employees. Here are some tips on good (and bad) practices to ensure a smooth and successful implementation of self-service BI tools.

  1. Associate employees with data experts

Self-service analytics require making data accessible to all employees. It also allows users with no data management experience to work hand-in-hand with experts, “said Rob Perry, vice president of product marketing at ASG Technologies, an IT service provider. Bringing novices to BI experts can bring a new perspective and exit traditional approaches to development and analysis.

  1. Do not let users do it all alone

It’s not always a good idea to let users search for and buy their BI tools.

“The data manager (or the Data Officer team) must manage the selection of solutions and provide the training and support structure needed to make users productive with these selected tools,” adds Rob Perry.

  1. Create a reference system

Create datasets that have been validated and verified to make them a reference system. It is also necessary to define a validation process that can recalibrate data sets that do not meet the defined quality constraints. A good strategy is to find a way to audit these data early in their life cycles so that errors can be corrected at the source, says Sreeni Iyer, CTO at LevaData, a cognitive services provider.

According to a Harvard Business Review study, 47% of newly created data has at least one critical error. The data quality process should not be isolated. It must be ubiquitous and function almost like an antivirus to ensure adequate quality across the organization, before the data reaches the stage of analysis.

In addition, the trades must be responsible for the quality of their data, because they are the ones who know them best, advises Isabelle Nuage, director of Big Data marketing at Talend. Trades must help IT in this quest for quality.

  1. The best is the enemy of good

In his report “The maturity of analytical tools and their impact on employees” dated 2017, Deloitte notes that it is not possible to wait until all the data are clean to take action.

The best is to identify a data set that is just “pretty” good, advises Chris Havrilla, vice president at Bersin. For example, if only 10 of the 1,000 rows in an employee dataset appear to be inaccurate or out of date, it remains relevant to use them for analysis.

  1. Arbitrate between different conceptions of data sharing

Implementing self-service decision-making tools requires balancing the different concepts of data sharing, warns Sreeni Iyer of LevaData. On the one hand, there are those who believe that data control gives power – which can lead to reluctance to share information or invite partners to the analytic process. On the other hand, there are those who believe that all data must be democratized, open, shared to create value. Between these two extremes, clear governance of access to data must be put in place.

  1. Do not focus on one tool

It would be great to have all the analytics in one tool, but that’s not realistic, “said Andrew Roman Wells, CEO of Aspirants, an analytic- driven management consulting firm . Most companies are changing rapidly. New data and new analyzes are needed continuously to adapt to the changing landscape. “Unfortunately, every analytical need requires specific tools,” he warns.

  1. Consider how analysts work

Analysts are very loyal to their tools. They will be reluctant to change unless they see a compelling reason and clear benefits for them. Business analysts will use data sources more easily and more efficiently if they can be integrated into existing tools, says Naras Eechambadi, CEO of Quaero, a customer data platform provider.

If it is necessary to replace existing tools, change management, advanced training and business support are essential. “You’d be amazed at how many analytical tools are left unused on employees’ PCs, just because the decision makers did not understand how the analysts whowork, “says Naras Eechambadi.

  1. Do not (immediately) combine self-service BI tools with traditional BI tools

“You would create duplicates, and any change to an element will require an impact analysis,” says Gal Ziton, co-founder and technical director of Octopai, a metadata management tool.

A change in the Extract, Transform, and Load ( ETL ) process will affect existing systems and self-service BI tools. The different groups should ensure that all reporting systems reflect the same results. Reverse engineering a bug will take a long time due to the multiplicity of duplicate systems and elements.

  1. Start department by department

Start by implementing a self-service BI tool for each branch, rather than a centralized reporting tool. Using the self-service BI tool as the primary reporting tool can only be done after designing the data item requests for each department in the organization.

  1. Train users with sexy data

Start training users with their own data, or those of their service. It’s even better if trainers can make sample reports with this data. Otherwise, the adoption may remain low, and users will be more tempted to use an export to Excel, warns Dominic Go, director of analysis at Olivela, a luxury shopping site.

Author: 
    author

    Related Post

    Leave a reply