New Book Review: "Embedding Analytics in Modern Applications"
New book review for Embedding Analytics in Modern Applications: How to Provide Distraction-Free Insights to End Users, by Courtney Webster, O'Reilly, 2016, reposted here:
In her introduction to this book, Webster comments that on any given day people use more than 20 different software applications, but while more and more employees are expected to make data-driven decisions, only about 20% to 25% of workers have access to BI (business intelligence) products, and users often do not want to use a "BI tool". The premise for this book is that the new trend is to embed analytics into applications that are already used every day, rather than using standalone dashboards, and this book is intended to act as a a guide to delivering analytics in this manner.
To set the stage, the author first discusses the possible drivers for an embedded solution, with the given that resources have already been invested in analytics: (1) focus groups report that users value the analytics in your application, (2) an opportunity exists to monetize the data captured by your application, (3) the desire exists to offer more sophisticated analytics, or customers are reporting some level of dissatisfaction with the current analytics or reporting being made available, (4) sufficient ad hoc or self-service capabilities are lacking, resulting in too much development time to deliver custom reports or queries, (5) reporting provided by competitors is superior, resulting in customer loss, and (6) a migration to SaaS is planned, but unsurety exists as to whether the current analytics solution will meet the needs for a multitenant environment.
If one or more of these drivers is true, the discussion then turns to the buy or build decision, and determination of the type of experience that is desired for users. In her subsequent discussion, the author presents several types of such interfaces involving static data (often involving REST APIs or reporting libraries), interactive data (often involving BI tools which offer dedicated tabs or pages, or custom development), and self-service exploration (often involving an API-based BI tool, BI scripting framework, or custom development).
From my experience, the buy or build decision can often enter the picture for application types already offered in the marketplace, but heavily custom application requirements not addressed by commercial products, often involving market differentiation and some level of innovation, will likely lean in the direction of a decision to build. And in the scenario targeted by this book, where a custom application already exists, the author comments that once such an application has been painstakingly designed, built, and optimized over time, it can be difficult to imagine an out-of-the-box analytics product which meets all needs.
Based on surveys of non-BI software vendors, the author walks the reader through the top-5 reasons cited to build instead of buy, and the top-5 reasons cited to buy versus build, and subsequently presents the 7 challenges and their best practice solutions around choosing the right embedded analytics tool: (1) customization (will it look like the rest of the application, and can it easily be customized?), (2) usability (will it please customers, and will it provide a seamless experience?), (3) capabilities (can it meet business needs?), (4) multitenancy (can it support needed security and access permissions?), (5) scalability (can it scale with the custom application?), (6) data structure (will it work for associated data structures and data streams?), and (7) performance (will it slow down the custom application?). Good overall presentation delivered in a small package.