Today, businesses face a dual challenge as employees are required to perform non-routine tasks that involve accessing data they need for decision-making, yet want to do so without having to go to IT each time a new requirement or question arises. This has led to a rift between businesses’ need to respond quickly to work demands, and IT’s failure to deliver data in an actionable time-frame. For this reason, organizations are asking: with growth in both complex work and data, how do more people independently explore business questions? A popular solution to this challenge is self-service BI.
Is it Really Self-Service?
To begin, let’s put things into perspective: providing real independence to business users is not trivial, so don’t be fooled by claims that a tool will ‘empower users to generate insights,’ or facilitate ‘IT’s role from enforcement to empowerment’. These superlatives mask the complexity of self-service BI.
To avoid poor technology decisions, it’s necessary to understand that self-service BI is not binary, which is either ‘off’ or ‘on’. Rather BI and dashboard software tools vary in the degree of self-service they offer. The implication of misjudging the level of a vendor’s self-service means you’ll need to invest in additional systems, learn advanced skills, and have more work down the road.
To illustrate this point, let’s take two self-service examples unrelated to the business world. Example one: Early in the morning, you step into your kitchen, open the top of the espresso machine, fill it with water, place a capsule in the top and a cup under the tap, and make the perfect espresso shot. Example two: Later in the day, you pass by a Starbucks. You go inside, wait in line, reach the cashier and describe your order—a tall caffè americano. After paying for the coffee, you wait to hear your order called, and soon receive your order. Both examples are a form of “self-service”, but the level of your involvement, independence and cost is vastly different. Likewise, a practical framework is needed to compare the extent of independence offered by different self-service options. To this end, we suggest an approach of ‘push and pull’ factors to better understand the conditions that enable independence against potential friction points that limit it.
Conditions for Self-Service
To be and remain independent, a user must be able to accomplish any data analysis task from start to finish within an actionable timeframe and without help from anyone. The conditions which enable this goal become clear by looking at the questions a user must answer to perform any analysis. Acquire: “Do I have access to all the raw data, and is it in the correct format for my tool?” Without access to relevant data, no analysis is possible. The complexity in this step arises from the need to manage any amount, form or type of data. Analyze: “Can I use the tool to create any queries I need and get accurate results?”
In technical terms, analysis relates to querying data which includes all formulas, visualizations and filters. The challenge is to generate queries without advanced skills like SQL to return correct results without delay.
Sustain: “Can I easily respond to changes in requirements?”
Your business environment is in a state of flux – data sets grow, more users perform analysis, and further data exploration is needed. Changes must be simultaneously supported to scale to large data sets with many users and varied analysis.
These three conditions to acquire data, analyze and sustain changes must be in place to support self-service.
Self-Service Friction Points
In establishing conditions for self-service, obstacles arise that prevent independence and result in the need for extra resources or abandoning the task. These friction points can be grouped in three categories.
Data: This relates to data access and preparation. It includes dealing with messy, inaccessible or large data sets, and includes challenges of joining multiple data sets together.
Skills: This relates to the need for advanced skills to get a task done. It occurs when the level of knowledge required exceeds that of a typical business user competent in Excel.
Infrastructure: Lack of appropriate hardware can restrict the ability to run queries, support more users, and process large data sets without a drastic decrease in performance. As far as possible in the process tasks should be supported by a standard laptop.
9 Questions for Self-Service
Conditions for self-service can now be balanced against the friction points that need to be addressed. This results in a matrix of nine types of questions to test the degree of self-service supported by a BI tool.
Let’s walk through a few examples of questions you can generate with this framework:
- Acquire – Data: Access recent data from all sources in the rawest format. Data access requires you to connect, join and synchronize diverse data sources. This gives rise to questions like: Is it possible to access all the required data sources? What needs to be done to join different data sources together? Can a connection be made directly to the raw data?
- Acquire – Skills: Skills to format data for analysis. Let’s assume you work with many data sources that have data quality issues. You need to check the level of skill to make corrections to the data before analysis. Relevant questions are: Does the tool require a professional DBA? Do users need to learn different skills to prepare different data source?
- Acquire – Infrastructure: Infrastructure to process the data for analysis. Working with large varied data sets is a common requirement. This needs to be supported by the appropriate technology infrastructure. You should ask: What hardware is required to process data sets in a reasonable time-frame?
- Analyze – Data: Ability to connect to data and run any type of analysis. You may need formulas to perform multiple aggregations on the data. For example, calculating ‘Average Total Sales per Month’ requires both an average and sum. You may ask: Can formulas be created with multiple aggregations without the need to first summarize data?
- Analyze – Skills: Perform desired analysis and easily explore results. Analysis requires both the ability to generate the formula and explore result sets. The leads to questions such as: Can the query be generated through the UI without the use of complex language like SQL? Is it easy to drill between the summary and detail levels of results?
- Analyze – Infrastructure: Hardware required to quickly run dashboards with many queries. For efficiency, you need to be able to display many queries in a single dashboard. A relevant question is: Does my laptop support dashboards that return results in a reasonable time-frame with many formulas, visualizations and filters?
- Sustain – Data: Hardware infrastructure to scale to large and varied data sets. Both the volume and variety of data grows over time. An assessment should identify if the solution supports large data sets, and what hardware changes are required as data grows.
- Sustain – Skills: Skills required for users to manage a single version of the truth. As the team grows, you need to ensure they get answers quickly and can make the necessary changes to the data and analysis. A critical question: is it easy for users to make changes, but still work from a shared data source?
- Sustain – Infrastructure: Hardware infrastructure to support many user needs. As teams grow, more users must interact, create and change dashboards and data. What is the appropriate hardware to maintain performance for users? Is the processing done on a central machine or the user’s own machine?
To choose the right self-service tool, vendors must be differentiated based on the extent it grants business users independence. The 9 Questions for Self-Service framework is a practical guide to raise important questions about BI options. Getting answers to these questions make it easier to test the degree of self-service and avoid decisions that demand more work and resources down the road.