A Beginner’s Guide to Embedded Data Analytics

This abridged guide will cover the essential things to look out for in selecting and purchasing embedded analytics software. The full guide is available here.

Whether you’re producing automation software, SaaS products or cloud applications, it’s likely to assume you’re collecting a lot of data in the process.

This abridged guide will cover the essential things to look out for in selecting and purchasing embedded analytics software. The full guide is available here.

Whether you’re producing automation software, SaaS products or cloud applications, it’s likely to assume you’re collecting a lot of data in the process.

With an increasing number of companies and individuals understanding the value of using data to improve different aspects of their business, the ability to offer a powerful data analytics and BI feature within your existing application can give your product the competitive edge that it needs and greatly improve the value you offer to customers (see: Does the Future Lie with Embedded BI?)

Here are 4 things to consider before getting started with embedded analytics:

#1 In-House or Out of the Box: Buy vs Build

Once you’ve decided to add an embedded BI feature to your product, The first thing you’ll want to consider is whether to buy existing embeddable software and integrate it in your own app, or to develop an analytics platform in-house.

In a world of unlimited resources, you’d probably want to keep everything in-house to retain full control of your product and include the exact kind of functionality and UX you feel that you or your clients require. However, the reality of the matter is that business intelligence is not a core competency for most businesses, and developing a full-fledged business intelligence platform could take years of R&D work and huge financial investments.

NB: A simple visual interface does not present a particular challenge in development terms. But it’s important to note that Business Intelligence is about more than just displaying fancy visualizations on the user’s screen: It handles joining multiple data sources, running fast queries on large datasets and allowing users to explore their data by questioning it in a wide variety of ways.

This type of analytics platform is no cakewalk to develop. Building a robust BI system that can handle the demands of big or complex data would require immense resources (in terms of time and money) and might still fail to achieve the same level of functionality as an out of the box solution.

#2 The Caveat: Ease of Implementation

Having said that, you also shouldn’t overlook the possible hidden costs and time-sinks that come with some embedded solutions.

Problems with integration between your own software and the embedded analytics platform of your choice have the potential to greatly increase your costs and production time. This could mean prolonged periods that will have to be devoted to development and iterations between you and your BI provider.

Additionally, some BI software is so complex to implement and use that it will require extensive training on your end before the system is actually up and running, further extending your costs, time to market and proving to be a major headache on its own sake.

In other words, choosing external embedded BI will not necessarily guarantee you faster time to value. It’s important to try the software out for yourself, on your own data, before making the decision.

#3 Defining your Requirements

There’s a seemingly endless amount of BI products in the current marketplace, and to the untrained eye they could all appear to be promising the same essential things.

However, closer inspection – which might actually require downloading a trial version of the software or requesting a proof of concept – will reveal substantial differences between the different types of software. For example, Front end tools such as data visualization software focus on dashboard reporting, whereas end to end tools also handle data preparation and have a built-in querying and analytics engine.

The type of tool you’ll require depends, among others, on the volume, variety, and velocity of the data you plan to process. Things you need to consider include:

• Size: How much data will you need to handle? Hundreds of megabytes? Gigabytes? Terabytes? Some BI tools’ performance can suffer when handling large datasets.
• Reporting: Will it be enough to generate a few pre-determined reports, or will you want users to be able to generate custom queries and reports?
• Security: Which permissions will you be able to set, and how difficult will it be to do so? Can you set permissions on database, table and row levels?
• Data complexity: Is your data fairly organized and structured, or are you dealing with complex data coming from multiple sources?

#4 Don’t Underestimate Your Future Needs

Even after thoroughly defining your exact plans for your embedded analytics, don’t forget that Business Intelligence is, to a large extent, the realm of the uncertain. The amounts and types of data we collect today would have been incomprehensible a few years ago, and there’s no reason to believe they will remain identical in a few years’ time.

To avoid the need to repurchase, re-implement and re-train your staff when you discover the solution you’ve chosen can no longer fully satisfy your requirements, make sure that whichever embedded analytics platform you choose will be scalable. Assume your datasets will grow and your querying and reporting needs will also expand, and make sure that the software you integrate will be able to handle the larger workload.

Learn More: Evaluation, Implementation, Building a Support Model

Download our free whitepaper: 5 Do’s and Dont’s of Embedded Analytics to get more essential tips on choosing between vendors, the implementation process and how to assure your customers get the best support in the process.