Deploying Embedded BI
According to an ongoing survey conducted by the Business Application Research Center (BARC), the top three impediments to successful BI implementations are “lack of resources on [the] project team,” “unclear requirements,” and “data migration.” This guide offers best practices in navigating those and related challenges so that you can reach your embedded analytics deployment goals.
Selecting an Embedded BI Launch Strategy
There are effectively three different approaches to an embedded BI launch, which we’ll call accelerated, incremental, and comprehensive.
An accelerated launch prioritizes expediency over thoroughness. Lean teams facing an imminent deadline will often adopt this approach and shoot to launch a minimally viable implementation.
Comprehensive embedded BI deployments take the opposite approach, prioritizing thoroughness over expediency. Larger enterprises with lots of runway and budget tend to gravitate toward this strategy, preferring to launch as ideal an implementation as possible.
The incremental deployment, a midpoint between the two extremes, aims to cover about 80% of use cases in the initial launch. A moderate timeline gives application administrators ample runway while reducing chances of the project becoming stalled.
Enterprises should adopt the rollout method most suited to their timelines, resources, and objectives. This guide, however, will concentrate on best practices for the incremental launch, since it is recognized as the industry standard and the method most often met with success.
Best Practice for an Incremental Launch
Each SaaS vendor’s training program is uniquely suited to its product and user base, and there are a host of strategies to consider as you design yours. Here is a list of tips and ideas sourced from our customer base.
Phase 1: Implementation
Prepare Your Data
User trust in your BI solution is paramount to its success down the line, and that trust starts with the data itself. You cannot over invest in this step, so take the time to ensure that the data you release to users is organized, reliable, comprehensible, and secure.
There is more to say on this topic than we can cover here, so please refer to the following to get started:
The Four Data Management Mistakes Derailing Your BI Program: Discusses common impediments to successful BI implementations, from poor data stewardship to mishandling of dimension information.
An Introduction to ETL Data Transformations: Describes data manipulations you might consider in building a data warehouse or library of programmable data objects.
Focus on the Report Library
SaaS providers looking to enhance their products with embedded analytics are often swamped with reporting requests from users. If this is the case for your IT team, your first priority is reducing that workload. Take stock of the most frequently requested reports and set about building those in the application.
If you focus on providing canned, parameterized, and scheduled reports for the initial deployment, you will have more time to plan and develop formal user training materials. Your team will also have many fewer reporting requests coming in and be better able to assist with the self-service phase of your embedded analytics launch.
Plan Your Feature Configuration
Decide which features you will expose in your initial release. Confining users to tools requiring little to no training is one tactic, but you may also opt to test more advanced tools on select customers or user groups. Beta releases are very often used for user research of this kind. Anticipate what implementation questions will be coming down the pike after the initial launch, and plan your tests accordingly.
Phase 2: Launch
Whether conducting explicit beta testing or not, challenge your product team to gather feedback from customers regarding the initial release. Allow that input to inform your next product update.
How long you wait before moving on to the feedback stage will depend on user adoption, so consider setting up a monitoring system for tracking BI traffic. Keep your timeline in mind, but also give users a chance to get familiar with the new tools before soliciting first impressions. When it’s time, probe for:
- Points of confusion or frustration
- Functionality users are requesting
- Unexpected feature behavior
- Unexpected user behavior
While the solution is in beta, you can either work on other projects, freeing you up for the next phase of your embedded BI deployment, or you can work on BI tasks unlikely to be influenced by public response to the beta. Either way, use the downtime to get ahead.
Phase 3: Regroup and Reiterate
Remember that you may also have the option of soliciting your BI provider for guidance regarding your implementation.
Lastly, be sure to make initial users available to your marketing team as well so they can source testimonials, petition for product reviews, get quotes for future press releases, and author case studies. Keep all teams responsible for publicity up to date on your roadmap plans and united in their messaging.
When your team is ready to introduce more advanced BI features to the product, it’s time to think about how you will train your end users.
Tips on Designing Your End User Training Program
Follow these recommendations to keep user adoption high after deploying your embedded analytics.
1. Segment your courses.
Enterprise BI can do a lot, and it’s counterproductive to rush through the material, so help ensure that your end users get through all the necessary information by breaking up the training into segments. Matt Smyrl, Manager of Training Services at FM:Systems, divides his course into three half-day segments, which allows him to ramp up the complexity gradually and give his students breaks between segments.
Lomesh Shah, President of NonProfitEasy, likewise subdivides his training into one-hour drop-in forums. Each forum focuses on a different level of report designer, either basic, intermediate, or advanced.
2. Hold in-person training sessions whenever possible.
SIS Product Manager Bryce Lee hosts monthly training webinars for his company’s Exago BI implementation, but he also leads in-person regional workshops for whole groups of insurance agencies using PartnerXE. Exago customers generally agree that in-person classes are always preferable to virtual ones because they allow for better communication and more interactivity. It’s also easier to observe users as they explore the application.
3. Start with your data architecture.
That’s right, break out those ERDs. Both DLGL Data Extraction Analyst Jonathan Giles and Smyrl say they begin their training with an introduction to the data itself: what data categories are available, how they are organized, and why they are organized that way. This gives users a better understanding of what data they can access and might even get them excited to start exploring it! If you have a particularly large data set, consider confining your introduction to an especially relevant subset of data objects to avoid overwhelming new users with information.
4. Provide a familiar point of reference.
One Exago BI client discovered that it was helpful to explain the report designer in terms of a more familiar tool, and that tool ended up being Excel. Exago BI doesn’t work quite like Excel does, and explaining how they differ actually helped this product manager get some of the more unfamiliar concepts across. Spreadsheet cells, for example, only hold one piece of information at a time, but some of Exago BI’s report designer cells hold whole fields of information that expand to display detail on output. These kinds of comparisons can help users gain a foothold more easily.
5. Give users an objective toward which to work.
Smyrl says he’s found the most success with individual and group classes wherein attendees describe a report they’d like to make and then spend the session building it. Not only do they learn along the way, but they have something useful to take with them at the end. This is also a great way to get users working through use cases they might realistically encounter in their daily work, giving them a supportive environment in which to work through design challenges.
6. Leverage canned reports.
A well-constructed canned report can be as powerful a learning tool as a blank canvas. One product manager said his first step is always getting users comfortable with modifying existing reports for their own purposes, starting with stylistic changes and working up to structural adjustments and new calculations. Giles will tell his end users to take inspiration from the report library and try reverse-engineering the reports or recombining elements to create something new.
7. Provide refresher courses.
As one panelist pointed out, there’s often a lag between when end users attend training and when they build their first reports. It’s easy to forget some of the finer details in the interim, so help your end users get back on track either by providing a refresher course or by making training videos available on demand.
In Shah’s experience, some end users come to him exasperated because they needed a custom report made yesterday and didn’t have the skill set to do it themselves. In the interest of time, Shah’s team will typically build the report on the client’s behalf and charge for the work unless the client attends a training session, in which case the report would be pro bono. This has proven an effective means of advertising the training program to those who need it most.
Once you’ve deployed, find out from your BI provider what you can do to become more involved in the solution’s evolution. Perhaps there are opportunities for you to weigh in on the product roadmap or share implementation tips with others using the same solution. The more involved you become, the better your embedded BI vendor will be able to accommodate you and others with similar challenges.
Congratulations on your embedded BI deployment! We hope this guide proved helpful and invite you to explore our whitepapers and ebooks, along with the rest of our content library.