How to Unstick Your Analysis

Sometimes people hit a wall when working with data. Bottlenecks arise and next steps are unclear. In this post I will look at bottlenecks that can leave you feeling stuck with your analysis, as well as share tips on how to get unstuck.

When Analytics Maturity Slows You Down

Often, analysts get stuck due to reaching a plateau in their capabilities and resources. Analysts move along an analytics maturity continuum throughout their careers, as their  skills and capabilities grow. People have studied this phenomena over the years, with one well-known maturity model coming from Stephane Hamel in 2009 (he has an updated version of the model here). When an analytics pro hits that plateau, it leads to that stuck feeling — when that continuum on improving skills and capability is brought to a halt.

So how do you get unstuck? A starting point is to recognize being stuck isn’t a workflow failure. It’s an opportunity to rethink how to best deploy resources and skills. Having the awareness that you’re stuck means you are aware that the continuum exists and that better processes are possible. It is a reflection of one soft skill I feel analysts must have: curiosity.

To refine that curiosity into useful processes, use the following ideas to get past the bottlenecks.

Challenge Assumptions Baked Into Your Analysis

Part of being stuck can come from setting up an analytics project too closely to old processes, which fails to give the project the forward momentum needed.

Check the assumption you’ve made that supports the intended data model. Is the starting point the same as a previous analysis? How do the initial assumptions compare to current conditions? Do new data conditions exist that did not exist before? Even differences in data cleaning techniques, such as the treatment of NA observations, can be cause to reassess assumptions.

Being stuck too long can push projects into technical bankruptcy, a variation of technical debt that is recognizable by a lack of agency to move forward. To avoid this, visualize your project’s milestones and ask what you need to reach these milestones. Doing so will help you identify smaller, achievable tasks. View this as a retrospective, not as an opportunity to assign blame for something that perhaps went undone.

Related Article: What Analytics Trends Should Marketers Expect in 2022?

Avoid ‘Data Puke’ – Put Your Analysis Back in Focus 

Scattered analysis work sounds antithetical to the organizational nature of analytics, but it happens. Not bringing order back to that mess will ultimately overwhelm your audience with “data puke,” when you needlessly report anything and everything.

When you are overwhelmed, take a step back and see if you can focus on a subset of data where plausible answers seem to be forming. Ask an analytics so-what question: how does this data discover tie back to the customer journey or objective of the analysis? Focusing on a subset reduces repetitive tasks that lead to overwhelming feeling of ongoing aimless tasks.

Another aspect is to look at the framework being applied to advanced analytics modeling. It is easy to get lost in a sea of frameworks and lose sight of the purpose they serve in structuring a relationship within the data. 

Related Article: How to Choose the Right Data Visualization

Rubber Duck With Your End Users

A lot of times, the audience for your analytics results won’t be as proficient as you are in the technical elements of analytics and data modeling. For example, they’ll know what SEO is, but are unable to look at code and recognize that metadata or analytics tags are missing. A common scenario is when a colleague needs insights, but are unable to describe the technical elements of where to obtain the insights. This is no different than going to the doctor and describing pains, but not knowing the medical terms for the body parts that ache. If your end user is stuck in the analytics quicksand, you’re stuck, too.

To climb your way out of the quicksand, go through a point by point review. Developers call this review process “rubber ducking” — a way to articulate a problem in non-technical language. A rubber ducking review can sometimes help identify how the results can better speak to your colleague and lead to adjusting the important steps to get an analysis done. It is also a great learning opportunity, a blameless “gut-check” to know when to refine an analysis workflow.

Related Article: How to Improve Data Literacy Among the Non-Quants in Your Organization

Build a Supportive Resource Community

Sometimes the bottleneck arises when you are using a tool or a solution feature in an inefficient way. If you think this might be the case, do a sanity check with people who have encountered similar issues. Analysts often share questions or concerns around specific tools in online communities. The solutions providers often host these forums, including IDEs such as RStudio or dashboard platforms like Tableau. You can also find independent help online, be it through a hashtag on Twitter or through a hosted groups in Discord. Analytics communities have existed around specific platforms for a while, so they typically know all the ins and outs regarding updates and past feature bug history.

You can also use this process to identify potential documentation needs for the software you use. Documentation sometimes goes overlooked with upgrades and changes. Reviewing documentation can highlight if functionality questions about current software are being properly addressed and keep resources up to date.

Is the Automation in Your Platform Stuck?

Analysts have a number of automation choices for ingesting data, such as macros in Excel, API libraries for R and Python programs, and collection features within tools such as Google DataStudio, Tableau and Power BI. So look towards any existing automation workflows if a bottleneck arises. Make sure your datasets and visualizations are being kept up to date with the latest information. When implemented correctly, gathering data from automated steps should consume less effort over time so you can give deeper attention to analysis.

Review your frameworks to see if better feature solutions exist. At minimum, a review will give rise to questions that will inform your research. To learn how to form those questions, see my post “How to Create Dashboard Frameworks That Support Marketing Analysis.” No one perfect solution exists, but the time invested in choosing and refining a solution will help unstick your analysis.

Use Bottlenecks as a Time to Address Privacy, Security and Accessibility Needs

While bottlenecks can be a drag, you can use these slowdowns as an opportunity to review data security, privacy compliance and, for data generated from web UI, accessibility — since you’re already reviewing data and data sources. The inputs for these processes dovetail into steps for ingesting and processing data. For example, adding analytic tags to a website page is an opportunity to check the accessibility of that page content.

Remember to Celebrate Your Victories

If it feels like you still have a lot of work after expending all this effort to unstick your analytics, celebrate the victories from your accomplishments just the same. The last two pandemic-dominated years can feel like an ongoing exercise in avoiding burnout. But any effort you make that prevents a small analytic problem from scaling into a larger one is a good reason to rejoice.