Siloed analytics - What is it and why is it bad?

[date-stamp]Many applications, both web applications and desktop applications, have excellent analytics capabilities. You can drill down into the data in the application, narrow your view to one single customer and see trends and forecasts. All that is great. But this application is not your whole company, what about the data in the other applications you're using? What is happening here is "siloed analytics", where data analytics are unable to see data that are outside of their applications' control.


What is siloed analytics?

Siloed analytics are analytic capabilities that operate on data from one application only. This can be the CRM system, the support system, email or any of the other applications you're using. Analytics in one single application are created with the best intentions: To give at-a-glance overviews and the ability to drill down into the available data. But these analytic capabilities only see the data in one application, or even only in parts of one application, so it does not give a comprehensive view of the information available in the company, only a part of it.

One example is your support system- it may give beautiful graphs over customer communication and the number of incidents reported, but what if some customers email one of your employees directly when they have a problem instead of using the support email address? These emails fall through the cracks and are not included in the reports from the support system since these emails never entered that system. If this is a widespread scenario, the analytics in the support system give a false impression of the number of support requests.

Any analytic functionality needs a complete view of data. If only a part of the data is available, the analytics are misleading.


The invention of siloed analytics

99% of all companies use more than one application in order to run their business. There is an email system, a CRM application, a file storage, a support system, maybe a custom production system, social media software and so on. Some of these are off-the-shelf software, while others are tailor-made for the business, either by an external consultancy or in-house. And this is normal, most companies have it this way. And many of these applications have analytics capabilities, allowing you to see very impressive graphs and overviews. When a vendor decides to add analytics support to their product, they only think of the data that is available in their product, and I cannot really blame them. The result, however, may be a visually stunning experience but with data from only one application. Duh. While that may be great for some businesses, after some time using it, most companies will discover the limitations with this approach. Any data needs to be loaded into this application so its analytic capabilities can "see" the data, otherwise the analytics reports will not be of sufficient quality to base a decision on it. And what's the point with analytics if you cannot take decisions based on it? Take a social media application that reports how many times the links you have posted is clicked on, showing which one is clicked the most. This kind of report clearly states which of the topics you post about on social media are most popular, and this allows you to decide to post more links about these topics. But imagine that you post to Facebook, LinkedIn and Twitter, and your analytics application only tracks Facebook? The analytics report would be inaccurate and not fit for basing a decision on. This illustrates the problem with siloed analytics: They are great on their own, but not in a multi-faceted world where companies operate in multiple applications simultaneously.

There is one more reason why there is so much siloed analytics, and there is the lack of any standard to share data with other applications. This is the case for web applications, mobile apps and especially for desktop applications. Most web applications have REST APIs, but maintaining connectors for these can be time-consuming for a software vendor (believe me, I have been there!) so most web applications have connectors for a limited number of other web applications. For mobile apps, there are no standards for exchanging information between business applications in Android or iOS to my knowledge. For desktop applications, there are certainly no standards.  So this lack of standardization for data exchange clearly affects which data is available in applications, and thus also affects the quality of data analytics in these applications.


How to break the silo

Earlier this year I wrote a blog post about data silos: Are you stuck with a data silo but don't realize it? This post describes some ways of getting rid of data silos in general, without touching on the data analytics challenge for data silos. The best advice for how to break the silo is to be very careful when choosing which applications you use in your company. There are applications which talk to each other and exchange data, so be sure to choose these. But the fact that applications are able to exchange data does not mean that their analytics functionality can use this data, unfortunately. A CRM system and its analytics package cannot automatically start generating reports for a support system even though support requests can be seen from the CRM system. So even though data exchange helps, it is not a magic bullet.

A way to break the silo is to use cross-application analytics, which extracts data from multiple applications and uses this data to create reports and overviews. In theory, this should generate a more accurate foundation for making decisions.


Why does it matter?

If your applications do not talk together and cannot generate reports from data across multiple applications...then so what? Why is this a problem? The problem is that the analytics are not complete, since it covers only one application, it can give false or misleading information about aspects of your company. One reason for using analytics is to track developments as they unfold, and then make adjustments. But making adjustments based on incomplete reports risks doing more damage than good, and represents a potential risk for your company. If the analytics have missed one important fact which is stored in a different application, it can be plain wrong. So this is something to take into account when using siloed analytics, it may not be 100% (or even 50%) accurate.