Is self-service all that? The problem with self-service analytics

Neil Raden Profile picture for user Neil Raden December 4, 2023
Self-service isn't as straightforward as some would think...


The term ‘self-service’ has a beautiful ring to it. In the context of using information in your work, this seems like a pretty good idea. Going back decades, you only had two choices for gathering data for reporting and analysis: putting in a request to IT for a report and waiting six months or more to get it or getting a microfiche reader.

Compared to that, wrangling some data, and developing a report or a dashboard or a visualization was a huge step forward. Self-service works unless it's something you don't want to do yourself. If you consider the term self-service outside of the context of information work, the sentiments are mixed.

Do you like doing your own self-service check-out at the grocery store? I don't. If you like being monitored like a criminal, then maybe this is for you. Moreover, if you pay by debit card and you forget your pin, or you take out your card half a second before you technically should, the machine shouts out in front of the whole store..."CARD DECLINED".

What I would prefer is to walk through the store, have the things I want dropping in my cart, tallied as I go and alerted to things I overlooked ("Neil, you passed the Crispix again; do you still want a box?") instead of coming home without something I needed. Information is provided to me as I pass things that may be of interest. Topping off the CX (Customer Experience) is a friendly associate to bag them for me. However, what would you call this? It isn't self-service—it's beyond that.


Self-service analytics (also known as “democratization” or “embedded” analytics uses function-rich toolsets so that business users can pose questions and get answers, typically on extensive collections of data, without the need to request assistance from IT. Efforts at this with Natural Language interfaces were clumsy and very limited, but LLMs, particularly interactive, conversational chatGPT, blew the doors off this feature.

What if you could have something beyond self-service analytics?

How many "clicks" do you make in a day? After twenty-five years or more of "self-service", what would you think of "click-less" analytics? Like beaming technology in the grocery store alerting you to items, discounts, or coupons that are guaranteed to interest you, what if the organized information you need simply finds you in the context of what you're doing?

My colleagues at Ventana Research use the term “Embedded Analytics Buyers Guide 2023 Vendor and Product Assessment”:

When analytics are embedded in business processes and applications, analyses are easier to perform and more accessible to line-of-business personnel…because the application collects and assembles data. Our research shows that data preparation can be the most time-consuming step in the analytical process; embedded analytics can dramatically reduce or eliminate this step. Analysis is also easier to consume because there is no need to switch context between the business very complex

It's a little more complicated than that. 

  • Logic: Data doesn’t speak for itself. Analytic models take time and patience because they can be very complex
  • Data: Embedded Analytics depends on a ready, quality corpus of data.
  • Knowledge Graph: facilitates intelligence up and down the organization. A knowledge graph is a navigable abstraction that consolidates disparate sources and presents a view that simplifies the process of leveraging data across the enterprise and beyond and is the best option to securely expose information effortlessly, contextually, and in an easily digestible manner.

My take

Developing embedded analytics will, for the time being, be developed by application vendors to operate cleanly within their applications. A DIY approach to embedded is currently beyond the capabilities of most enterprises.

A grey colored placeholder image