Skip to main content

It’s an open secret that companies are failing their efforts to become data-driven. Investments in data are rising but successful business adoption of data initiatives are falling [1]. The leading obstacle hindering adoption of a data-driven culture is not what you might think. It’s not broken tools, it’s not a lack of data, and it’s not a dearth of qualified people; it’s “process.”

A leading process problem is that the existing paradigm by which organizations access insights – the pull paradigm – is broken. The pull paradigm is the notion that you must manually go to a tool to pull the information that you need when a question arises. Typically, expending significant effort since this has to be done repeatedly to get disparate pieces of information and then stitch them back together. This hurts both decision-making speed and quality.

Let’s illustrate the problem with an example: Suppose you are a manager selling widgets online and at a monthly business review you want to report on sales performance. In preparing for the presentation you check the Tableau dashboard and find that sales have dropped 12%. “Uh oh,” you think to yourself. Is 12% a lot or is that expected? Does that typically happen this time of year or is it because of the recent change to the checkout process? You decide to dig-in to the Google Analytics conversion funnel. You see many custom reports/transaction-successful, /signup-complete, /purchases_v3. Gah! Afraid of using the wrong data, you raise an amber alert and send out an email to the analyst asking for help, cc-ing the product manager as FYI.

This situation may seem contrived, but it is all too common. Instead of selling widgets, you could be renting scooters, hosting a cryptocurrency trading platform, or selling pet food, the experience is a common one. It illustrates two major problems – untimely anomaly detection and tedious root cause analysis. To confirm this, we surveyed professionals at data-driven organizations and found that their experience echoes the example above.

Untimely Anomaly Detection

Existing tools fail to highlight what’s important. Having to pull updated metrics instead of having it delivered to you results in fire drills.

Tedious Root Cause Analysis

When a pattern of interest is identified, explaining why it happened is hard. Having to pull additional context means wading through many views/metrics resulting in dashboard fatigue.

Fig 1: Where today’s analytics is failing business and product owners

Let’s dig into the downstream impact of these problems.

A failure of tools to highlight trends and alert you to what’s important means that you are going to see the issue late, typically inadvertently when prepping for some recurring meeting, resulting in lost customers and sales, which could have been saved. Additionally, the subsequent fire drills lead to thrash, late office nights, and hurt employee satisfaction. Often the analysis ends up being rote, repeatable, and standardizable, but automating this work is not seen as a priority because of the time and effort required. The resulting delayed decisions mean that your invested effort in hiring people and setting up data infrastructure is getting wasted. It’s like buying a great steak but serving it cold and overcooked, you won’t go hungry, but you definitely aren’t getting all the value out of your purchase.

Difficulty in explaining causal drivers of metric movements hurts decision quality. The inability to readily find the data you need means you have to go to multiple views, sources, or tools. It could be that the dashboard you are looking for exists, but is hard to find, likewise, the metric may be on-hand but is tough to interpret and contextualize. Alternatively, the required view may not be readily available and you may need to filter, download, pivot and mashup data across many dimensions and time periods – this can be both times consuming and error-prone. If you make a mistake you end up using the wrong data, if you decide to ignore some dimensions you end up having an incomplete picture. Most product managers or general managers don’t have the ability or time to analyze so many dimensions. And if you need an analyst, you’re hamstrung by their speed, ability, and business context.

In the end, these are tractable problems that can be addressed with the right solutions. Our respondents resoundingly agree that addressing the twin issues of providing high quality alerting and offering insightful machine-driven root cause analysis would be a huge impact on their business. Much more so than applying deep learning, providing natural language answers, or improving the visual quality of existing solutions.

Fig 2: What users really want from their analytics solutions

Out of the Blue™ is building a solution to address the shortcomings of today’s existing pull-based tools and to empower data-driven professionals to give them what they need when they need it. Stay tuned for more.


[1] https://hbr.org/2020/05/is-your-business-masquerading-as-data-driven

[2] We surveyed 82 Bay Area Tech professionals, primarily product owners and business managers, to understand how analytics impacts their organization.

Learn how we do it, Get in Touch

Author

  • Mohit is a former product manager turned data scientist. He led analytics at Opower, Nest, and Google, successfully using data to build products, launch features, and grow users. He has an MBA from the University of Chicago, Booth School of Business and a BS in Engineering from University of California, Berkeley.

Leave a Reply