Written by GoodData Author |
With Forrester predicting that embedded BI that is contextual, actionable, and prescriptive will become the new norm for operational and tactical insights within the next three to five years, it’s more important than ever for companies to plan for insights-driven applications. I’ve spoken previously about what needs to happen before building insights-driven applications, mainly determining whether embedded BI is a good fit, reviewing in-product decision making, and determining how urgent embedding BI is. Once you’ve adequately prepared by addressing those three areas, you can take the next steps toward actually introducing an insights-driven application.
Step 1: Define the product and determine the required capabilities
The first step you need to take is to determine precisely what the product needs to accomplish. Does it need to track sales, highlight high-performing regions, and deliver suggestions on how to improve low-performing ones? Does it need to score insurance claims while identifying potential fraudulent activity? To find out, you should spend a substantial amount of time talking to the end users to get a better understanding of their pain points and daily workflows. For more on this topic, check out my colleague’s blog on the topic of defining data products. With this information at hand, you can create a set of requirements and define, at a high level, what the solution should be able to do.
Step 2: Decide whether you’re going to build or buy
Once you’ve defined the product at a high level and figured out what you need it to do, the next step is to decide whether you’re going to build it yourself or partner with a vendor. These are both viable paths, and the direction you choose depends a lot on your company’s needs. In particular, you’ll want to consider your available resources, how soon you want to launch the application, and how much money you want to spend.
It may initially seem less expensive to build your own application, but I find that many companies wind up calling it quits once they get into the nitty gritty of building an insights-driven application. For starters, the skill set required for building an analytical application is very different from the skill set of a traditional engineer. Though you may have talented engineers on your team already, you’ll need to hire skilled data engineers, data architects, and perhaps even data scientists, depending on your use case, to successfully build an application that’s robust enough to meet your customer’s evolving needs. Partnering with a vendor also tends to reduce the time needed to build the application from a year or more to considerably less time—8-10 weeks, in many cases. Together, accelerating the development timeline and avoiding the need to hire additional employees can reduce the overall cost by as much as 5x depending on what resources you have in-house today.
Step 3: Select your vendor (if partnering)
If you decide to partner, you will need to carefully select the partner, and that means doing your research and asking good questions. What do I mean by good questions? Well, the future of embedded BI goes beyond table-stakes requirements like fancy visualizations and mobile alerts, for example, and you should keep these next-gen requirements in mind, whether you are developing your own solution or partnering with a vendor. While these may not be part of your initial MVP, they need to be available as an option once you are ready, so consider these things when looking for a vendor:
- Embeddable user-authored BI content
- Seamless user experience
- Operational and analytical BI
- Support for insights-to-action to close the gap
- Scalability to hundreds of thousands of concurrent users
- Integration and orchestration
To determine if your vendor has the right capabilities to meet the next generation of your users’ requirements, I suggest asking your BI vendor a number of questions. For example, what governance structures (semantic layers, reusable metrics, layered security) exist to ensure users do not produce and share inaccurate results? How does the embedded BI solution communicate with host applications? Does the platform support massive concurrency, scale, and automation? Building an insights-driven application is a huge undertaking with totally different requirements than traditional analytics for internal teams, so be sure to be thorough in evaluating each vendor based on these questions.
Written by GoodData Author |