Today’s column is written by Steve Marosi, director, client advisory, at 84.51°.
Why is it that many analysts – who have so much data, yet so little time – turn up their noses at reporting systems? Why do they lose patience and reach for the export button, doing their most interesting work in Excel, Tableau or other roll-your-own tools? They’re so quick to tell data/insight providers, “I don’t want your dashboards and reports – just give me your data!” Why?
A data analyst recently told me that she exports data from upward of 70 reports to prepare a monthly dashboard for her department. Extreme? Certainly. But not unusual.
This quest for raw data is a message heard repeatedly from consumer goods companies. I imagine it’s not much different in other industries. Yet, what is it behind this self-inflicted pursuit of raw data with all its messiness, hours of toil to clean and harmonize and high risk of introducing errors?
The answer, in one word, is personalization.
Here’s my working definition for personalization in the context of data analysis: the freedom to shape an analysis, with minimal restraints, to address a very specific business problem or question.
Data analysts prize personalization so much that they’re willing to endure endless complexity, tedium and manual effort because it lets them find and communicate insights tailored to the problem at hand. For many analysts, standard reports don’t deliver insights; they serve as data pumps – valuable only for the content they deliver through the export button.
Should you care?
If you’re a vendor of analytic insights, each export implies that your system has failed the user in some way – either to deliver the important insight or to enable the necessary formatting and context.
If you’re a decision-maker, your analysts are not only wasting precious time, they’re multiplying the chance for blunders, which can lead to wrong conclusions. Every analyst I know spends far too much time pulling and polishing data and too little time analyzing it for insights.
But is exporting data always a mistake? No. There are two good reasons for it. First, no system can anticipate every possible path or analytic that a user will want to investigate. And second, to find certain insights, an analyst may need to combine data from one system with other sources.
These reasons both support what I call “exploratory analysis.” In nearly all cases, exploratory questions emerge from answers to more common and less demanding questions. An analytic reporting system for a given domain should anticipate and deliver the common insights.
But no system can anticipate when, how or in what order an analyst will proceed through an investigation. The artful science of data analysis defies repeatable processes. Human intuition and imagination seep into every zig-zagging step, meandering along a path of questions that often lead far from where the analysis started. Wandering one’s own path is vital if an analyst is to find the essential insight. Equally important is communication, the ability to tailor insights to the decision-making audience in a way that leads to understanding and action.
I see three objectives for personalization.
Freedom To Wander One’s Own Path
Pre-built analyses deliver far more speed and precision versus starting with raw data. An analyst, progressing through common, early-stage questions, should be able to rapidly explore a variety of hypotheses, leading to promising new questions while dodging blind alleys. This leaves more time for custom number-crunching in the later stages of an investigation – where the analyst really earns his or her paycheck by exploring uniquely relevant and valuable questions.
Pre-built analyses will still require a bit of configuration, such as user-defined thresholds, for example, but, designed well, they could serve as a happy medium between rigid reports and unfettered raw data access.
Unobstructed Access To Raw Data
Why allow raw data access at all? Less-skilled analysts might combine data tables and columns that result in absurdities, or they might misinterpret a measure because they didn’t double-check the definition.
But these risks are no worse than the situation today where an analyst exports data from a wide variety of reports and data sources, combining them manually in a spreadsheet however they wish.
Flexibility To Shape Insights Into A Story
Every analysis spews out tables, charts and other artifacts that require pruning before they’re ready to share with decision-makers. Giving users the ability to edit and rearrange their story within the system itself avoids the need for data dumps and screen captures merely to format and annotate.
Striking A Balance
The future of data analysis must somehow enter the fluid state between solid (standard reports, dashboards and landing pages) and vapor (limitless raw data access). The opportunity is huge to strike a balance between these extremes, increasing the speed and impact of data analysis in every industry.
Follow 84.51° (@8451group) and AdExchanger (@adexchanger) on Twitter.
This post was syndicated from Ad Exchanger.