As you see in the cartoon above, translating analytics to action is fraught with problems. Not the least of which is the political nature of organizational decision-making.
When it comes to ROI, we see lots of organizations jumping on the analytics bandwagon—happy to have more insight into which pages, which messages, which segments are driving returns. A whole host of digital tools can transform your data into pretty dashboards to help improve ROI.
Overall, businesses do a great job of monitoring, understanding, and using ROI information—translating it into effective actions. These relatively simple metrics guide managers on which messages, segments, products, and channels are working best, allowing them to tweak strategy to improve market performance. Still, you sometimes run into issues of data quality and politics that impede implementation of the right actions based on available data.
Advanced analytics and big data
The problem arises when it comes to using more advanced analytics and big data to improve ROI indirectly. Let’s take a look at the conversion funnel to see exactly what I mean:
Notice, that conversion is only one step (in the middle) of the process. Advanced analytics consider the entire funnel (as well as external factors that impact funnel conversions) to improve market performance.
For instance, consider this distinction between simple and advanced analytics:
Simple: We’re getting a higher conversion rate from traffic coming through Twitter. So, we put more money into our Twitter advertising strategy.
Advanced: Twitter is just the “last touch” in a multichannel conversion process that started with an Adwords campaign and included multiple visits to our website through remarketing. Multichannel attribution models are an example of advanced analytics.
Increasingly, C-suite managers realize the value of advanced analytics and big data, they’re just unsure how to proceed. Consider this from Harvard Business Review:
Even so, our experience reveals that most companies are unsure how to proceed. Leaders are understandably leery of making substantial investments in big data and advanced analytics. They’re convinced that their organizations simply aren’t ready. After all, companies may not fully understand the data they already have, or perhaps they’ve lost piles of money on data-warehousing programs that never meshed with business processes, or maybe their current analytics programs are too complicated or don’t yield insights that can be put to use. Or all of the above. No wonder skepticism abounds. (Source)
Using advanced analytics
This cartoon, from Avinash Kaushik of Occam’s Razor and data guru at Google, shows how even really smart people don’t fully appreciate the role of predictive analytics in translating analytics into action.
By their very nature, predictive analytics are probabilities. That means that they’re right more often than they’re wrong, but they’re not ALWAYS right.
Again, here’s from Harvard Business Review:
Data are essential, but performance improvements and competitive advantage arise from analytics models that allow managers to predict and optimize outcomes. More important, the most effective approach to building a model rarely starts with the data; instead it originates with identifying the business opportunity and determining how the model can improve performance.
Many of the worst predictive models follow a data-first approach. This includes data mining, which is still fashionable in many analytics circles and a feature of many analytic tools. Data mining looks at existing data and seeks correlations (relationships) among the fields within the data. Thus, it’s a rather mindless approach that totally discounts any understanding of marketing concepts or consumer behavior. Not surprisingly, the resulting algorithms are often totally useless. The bigger the dataset, the bigger the risk you’ll produce a useless algorithm.
Combining conceptual knowledge in determining which data to include in your model building results in much greater utility.
For instance, I once created an algorithm using readership among subscribers to predict who would be a sales lead for the organization. I used conceptual understanding building on the conversion funnel shown above to score articles based on content–content further along the conversion funnel got a higher score than general interest content. Using this method, we assigned a score to each subscriber and, once the subscriber passed a threshold it generated a lead for a sales person. This resulted in better quality and higher quantity of leads to the sales force.
Translating analytics to action with data
Translating analytics to action, especially when you’re talking about advanced analytics involves developing understanding and trust among users—often something in very short supply.
First, there’s a serious shortage of managers who understand analytics and big data. Notice in the graphic below, finding folks with skills in data analytics and insights is the biggest challenge faced when building a team, despite the relative importance of these skills.
It’s especially challenging to find team members who possess both marketing skills (in terms of concepts and application) as well as analytic skills. See my earlier post that explains why there’s such a gap.
Translating analytics to action faces many challenges. It:
- requires managers first understand the data you’re seeing. At a deep level.
- next, managers must trust the data they’re getting
- managers need to detect anomalies that require better understanding or further investigation
- finally, managers must understand how to translate analytics to action by understanding what the models are telling them they should change.
Translating analytics to action: understand data
I was working with a team of students on an analysis project using Google Analytics data. I quickly saw they had misunderstood what one of the dimensions meant and were using it incorrectly in their analysis.
Obviously, any effort at translating analytics to action requires you create a codebook containing the meaning behind the measures you’re using. Managers need training so they completely understand what each measure captures, where it comes from, and what it means in terms of consumer behavior.
In developing algorithms (predictive models) managers need input into the factors they consider important in building better predictions. This improves both the accuracy of the model and the manager’s understanding of the data.
Translating analytics to action: trust
Make data as clean as possible to ensure managers trust the data. Also, predictive models need sufficient testing, and those test results shared with managers, so they trust the algorithms.
Translating analytics to action: detect anomalies
Sometimes factors go a little wonky, sending your algorithms off and increasing distrust in your models. Managers need sufficient analytic skills to detect such anomalies. I recently worked with a client, and we discovered a conversion rate that just didn’t make sense. He was initially upset that conversion was much lower than industry standards, so, needless to say, I was very motivated to understand why the anomaly arose.
Spending some time pouring through the data, we detected a change in the data coming into our database which pushed subsequent data into later (inappropriate) columns. Hence, we were combining factors in a manner that didn’t make sense. We simply corrected the problem by changing the factors in our algorithm and things went back to normal.
Had we not detected the anomaly early, we might have gone off making bad decisions or spending unnecessary efforts to build a new algorithm.
Translating analytics to action
Sometimes an analysis shows you exactly which actions to take: poor performance of a piece of content, write content similar to those that performed better.
In other cases, it’s not clear how you should go about translating analytics to action.
In that case, you often resort to A/B testing to determine which actions improve market performance based on insights from the analytics. Or, you have to dig beneath the algorithm, into the data itself, to determine what actions are appropriate.
Final thoughts on translating analytics into action
Translating analytics to action isn’t easy and will take time to implement, even if you already have the right people, right skills, and right culture to make it happen. But, the benefits far outweigh the costs.
This article originally appeared in Hausman Marketing Letter.
This article was written by Angela Hausman and PhD from Business2Community and was legally licensed through the NewsCred publisher network.