Pivot From Spreadsheet Anarchy to Analytics Governance

By: on October 27, 2016

I’ll begin with a personal story.

When I started my current role at Software Advice (we’d just been acquired by Gartner at the time), I was introduced to our company’s custom-built business intelligence (BI) system.

It offers a crisp dashboard overview of business performance using reporting tables and pie charts. It’s refreshed with new data not just daily but on an impressive near real-time basis, and allows users to drill a few dimensions down in the data.

Sounds pretty slick for a small business like we were, right? Well, yes and no.
The system worked perfectly for our executive leadership, as well as our software advisers who consult with buyers via phone and web chat.

However, users in my own department frequently needed to perform more detailed analyses than the dashboard could handle. This meant that we had to export data from the BI system to CSV files and then analyze those files in Microsoft Excel or Google Sheets.

Over time, a whole spreadsheet library fed by data from the BI system piled up. There were spreadsheets with data from lots of different sources and points in time, and several different versions of some of the most important spreadsheets. Methods of analysis frequently differed significantly from spreadsheet to spreadsheet. Sound familiar?

Small business BI: A synonym for pivot tables in Microsoft Excel (Used with permission from Microsoft)

We’ve subsequently overhauled our department’s approach to analytics. However, most small businesses don’t even have the luxury we had of a centralized reporting system, which at least ensures that the data comes from the same source and shares the same format.

In this article, we’ll explain how this anarchic situation can be avoided with software that enables a governed approach to analytics.

What Is Analytics Governance?

The easiest way to understand analytics governance is to think of analytics as a spectrum, stretching from no central control over how employees analyze data to total IT control over analytics:

The Analytics Governance Spectrum

Self-Service Analytics

Self-service analytics is the most decentralized, anarchic end of the analytics spectrum.

Essentially, this means that employees do analytics themselves, without involving the IT department. Another synonym for self-service BI is “data discovery,” since users “discover” new insights via their interactions with the data.

Spreadsheets—like in my example—are the most well-known form of self-service BI, and one of the purest. However, there are now much more advanced BI tools on the market that allow users to perform analysis by dragging and dropping measures and dimensions into visualizations such as charts and graphs. Self-service BI is where the market is heading, and is thus also known as “modern BI.”

According to Gartner’s Best Practices for Driving Successful Analytics Governance, by Thomas W. Oestreich and Joao Tapadinhas, “while the widespread dissemination of data discovery provides more freedom for users, this comes at the expense of governance.” (This content is available to Gartner clients.)

“Self-service BI” isn’t always an antonym for analytics governance, since many self-service BI tools offer governance features that we’ll examine below. However, a purely self-service approach to analytics will be too anarchic for nearly all businesses.

IT-Driven Reports and Dashboards

IT-driven reports and dashboards, aka “traditional BI,” are at the other end of the spectrum, representing total IT control over analytics instead of business user freedom.

If self-service analytics verges on anarchy, then traditional BI is more like fascism. IT rules with an iron fist, developing dashboards and scheduled reports that are rolled out to the rest of the company (like the custom-built system in the example I gave in the introduction).

If someone needs an answer to a specific question that isn’t covered by the report, he or she will have to either get IT to write a custom report or resort to Excel.

If total self-service is anarchy, and total IT control is fascism, then you probably want to strive for liberal democracy by positioning your own BI program somewhere in the middle of the spectrum. Gartner recommends that you “think of governance as an enabler, not a restrictor.”

But how can you achieve this goal? For many small businesses, the answer can be found in modern, self-service BI tools rather than traditional BI systems.

Four Steps for Governing Analytics With Self-Service BI Tools

Small businesses tend to suffer more from analytics anarchy than from tightly and centrally controlled analytics. Let’s look at the steps that small businesses drowning in spreadsheets can take to get their heads back above the water:

1. Create Role and Department-Specific Dashboards

This one might seem a bit confusing, since in my introductory example we saw that dashboards won’t always provide enough insights to power your business. However, dashboards are certainly a first step.

The problem with the dashboards in my example was that they weren’t customized for my specific role/department. Instead, we had one dashboard for the whole organization.

When you’re rolling out dashboards at your small business, don’t just decide on a single set of metrics to use for reporting business performance. Instead, talk to leaders across the departments that will be served by dashboards (sales, marketing, advertising, business development, customer service etc.) to determine which specific metrics they need to track.

BI systems typically do support role- and department-specific dashboards, as in the below example:

Sales manager dashboard in Microsoft Power BI (a self-service platform)

Such customization is worth the time it takes you to set up, even if your organization is on the smaller side. After all, why even bother setting up organization-wide dashboards if they only give a basic, highly general overview of company performance? Most employees don’t need real-time insight into overall performance, even if they still need timely insight into metrics that are more closely linked to their roles.

2. Curate and Share Data Sources

Out of all forms of self-service BI, spreadsheets are the most anarchic, because they’re typically lacking in collaboration features, as well as features for connecting to data sources and extracting information from them.

Dedicated self-service BI tools such as Tableau and Qlik Sense, on the other hand, have a number of analytics governance features built into them. They need these features, because they serve as an alternative or supplement to traditional, IT-governed BI systems.

In a modern self-service system, any user can connect directly to a data source. They can then begin joining tables in the data sources or between multiple data sources.

For example, customer information from certain databases can be blended with accounting information in other databases and e-commerce information in still others. This is essentially the creation of a data model—a schematic representation of the categorical relationships between data sets—on the fly.

With traditional systems, however, data modeling is done upfront via IT-controlled mechanisms such as the data warehouse (see our explanation here) and the semantic layer (explanation here), and the data model remains under IT control after it’s defined.

The rigidity of data models in traditional BI systems means that blending data sources takes a deep knowledge of SQL and the specific technical details of every single data source.

In a modern system, however, data blending can be done through drag-and-drop gestures without any programming: Simply plop two data sources into the interface, select the type of join you want from a menu and you’re done.

Working with data sources in Alteryx (a self-service platform)

Once you’ve developed an original blend of data—for example, HR and accounting data, or social media and sales data—this pre-configured data source can be shared with other, less experienced users in your organization. They can then start analyzing the data in the ways they need to without having to develop data models from scratch.

This is a much more powerful way to enable your employees to perform their own analyses than merely sharing and co-authoring spreadsheets.

In addition to helping less experienced users analyze data, pre-configuring and sharing data sources also minimizes the risk of employees combining data in incorrect ways to arrive at false insights.

3. Build and Share Advanced Analytics Modules

Self-service tools allow power users to share complex analyses as well as data sources:

Advanced analytics module in Alteryx (a self-service platform)

Obviously, only power users like dedicated business analysts can set up a complex analysis involving multiple predictive algorithms, clustering algorithms etc. Gartner notes that advanced analytics pose specific dangers: “Analytics users and decision-makers need to understand emerging predictive and prescriptive analytic capabilities, which require data scientist-type skills.”

However, skilled users can still share their analyses with other users who can then at least modify the end steps of the workflow to arrive at new insights. Moreover, analyses can be shared within self-service BI tools in the form of easy-to-interact-with visualizations rather than complex scripts in statistical programming languages, such as R.

4. Configure Role-Based Access to Visualizations and Data Sources

Traditionally, the word “governance” in IT contexts translates into setting up specific access privileges for certain user groups.

In a large enterprise, the question of who should get access to which data sources is quite complex. An organization-wide governance policy is necessary, and sensitive data sources will absolutely need to be cordoned off from certain users.

Small businesses, however, have more basic governance needs. Primarily, you’ll need to designate:

  • Administrators who handle the technical details of the system
  • “Power users” who can configure data sources and advanced analytics modules and publish them to the rest of the organization
  • Everyone else who uses free-form analytics to spot issues with business performance
  • Dashboard viewers

Once you’ve authored the governance policy, self-service solutions allow you to divide users into groups with different privileges.

Users who need free-form analytics capabilities, power users and administrators will frequently need more expensive desktop/personal licenses for the tool than mere dashboard viewers, who will typically only need server licenses.

However, creating an endless set of rules for which departments get access to which databases, and deciding how senior someone needs to be to access a dashboard tab, can waste time and become an exercise in petty tyranny at a small business. Remember: you’re authoring an analytics governance policy, not a Kafka novel.

Conclusions and Next Steps

The essential difference between analytics governance in traditional and modern BI platforms is that traditional governance models are top-down. IT controls everything, including the most crucial part of the picture: data models.

Modern BI solutions however represent a bottom up, “crowd-sourced” approach to analytics governance. Power users create data models on the fly as they access and prepare data, with minimal input from IT. These data models can then be shared across the organization.

This bottom-up approach to governance is more feasible for small businesses that lack the IT resources to effectively implement the top-down model. Self-service BI solutions allow your users the freedoms they need to analyze their data, while also giving you a number of capabilities for governing their analyses.

Self-service BI is thus the best way for small, IT-starved organizations to become data democracies.

Get more information on and read user reviews about self-service products with the governance features discussed in this report on our data discovery page.

You may also like:

Traditional vs. Modern BI Buyer Trends Report

A Small Business Guide to Self-Service Data Preparation

What Is Data Blending, and Which Tools Make It Easier?

Compare self-service BI tools