So You Want to be Data-Driven?
Dear Executives,
My name is Adam Greco, and I have been in the digital analytics
field for quite a while. Perhaps we follow one another, or maybe someone on your team has sent this post to you. I hope that you will read this entire note so you can help your team to help you. I get a little Jerry Maguire here because this is a topic I feel so passionately about. I want to help everyone get some things straight about what it takes to be data-driven when it comes to digital analytics.
Everyone in the digital analytics field wants to be data-driven. The world is going digital, and the recent pandemic accelerated that by at least several years. But being data-driven in digital analytics requires more than lip service.
In my experiences at many conferences and consulting with many organizations, I have heard executives wax poetic about their company’s data-driven excellence and their esteemed analytics program’s greatness. I have even seen cool, polished videos where you executives are interviewed about how effectively you use data.
But in most cases, claims to be data-driven are total bull#$%t! Are you really using your data to make all your business decisions? How much do you really know about the foundations your dashboards are built on? How often do you think about data layers? JSON objects? Schemas? Query Strings? Do you know how these inform your dashboards? Does data quality keep you up at night? Do you routinely reflect on the enormous amount of work that has to be done, in concert, to achieve value in data?
To have your digital analytics implementation be effective, you need to deal with the following:
- Business objectives/questions
- Solution Designs
- Data Layers
- Tagging Specifications
- Tag Management Configuration (Data Elements, Rules, Extensions)
- Analytics Tool Configuration
- Quality Assurance
- Dashboards/Reports/Conversion Metrics/Segments
- Analysis
- Testing & Optimization
All of these items are interrelated and build upon each other. Unfortunately, today, most of these steps are done manually and managed haphazardly via various disparate documents and/or tools. The result is that most organizations spend too much time implementing and not enough time taking action on their data.
As my old mentor and boss Eric Peterson wrote (back in 2008!), Web Analytics is hard! So, I will detail what your team has to go through daily just to get your company to the point where it’s possible to begin to be data-driven. I promise that your team will appreciate you reading this.
The devil is in the details...and all the work it takes to get them right!
Being data-driven in digital analytics begins by having everyone agree on the organizational business objectives and how those translate to your digital properties. For example, why do you have a website or mobile app to begin with? As an executive, you need to clearly explain to your analytics team why you have chosen to spend $$$$ on your digital properties and the actions you want visitors/users to take. This will ensure that they are not wasting time working on things that aren’t critical to the business.
Once you have your business objectives defined, your analytics team will have to identify the business questions they think they should answer to help you know how you are doing with your business objectives. You should take the time to help them identify these questions or, at the very least, provide feedback on the ones they put forward before they spend a ton of time implementing them because, as you will see, implementing new things is hard and expensive!
Once your team identifies the questions they want to answer, they must determine how they will collect data to answer each business question. This is often called a solution design. To create this, your analytics team has to think about a lot of stuff! Should the data be collected as a metric or a dimension? Should the data be stored for a short time or kept until a conversion occurs? What is the best way to model this data in your chosen analytics product (this is where you may hear them talk about crazy things like eVars or sProps!). Years of expertise or money to pay outside consultants may be necessary to help you build solution designs correctly.
Once you have identified your solution design, your analytics team has to create a data layer specification. This tells your technical developers what data they need to pass (via a JSON Object) so that the tag management system (a tool that places JavaScript tags on your site and maps data to analytics tool variables) can get the data it needs.
Creating data layer specifications is more challenging and more time-consuming than you would think because you have many different systems that are all hungry for data! You have tags on your digital properties for analytics, advertising, re-marketing, and possibly even CRM. Your team has to make sure they are working together with many different teams to avoid a Frankenstein-like data layer. Doing that would slow down your digital properties and potentially hurt your conversions!
Once your data layer specifications are created, they have to be shared with developers who oftentimes have very little time available. Your analytics team may have to plead with IT to get just a few hours of their time to push data into the data layer. So, when you’re upset that you can’t see the data you want on your dashboard, the odds are that it is due to a bottleneck in IT. To remedy this, I recommend that you learn more about the tasks I detail here and take your IT folks out to lunch (or dinner!). Remind them how valuable they are: the coding they’re doing is helping to get data to answer business questions that ultimately help the entire company! 
But wait…there’s more! While your analytics team is waiting for the developers to populate the data layer, they have to configure the tag management system I mentioned earlier. Within the tag management system, they have to create data elements and rules for all of the data that will be collected. The data elements define where, in the data layer, each data point can be found. The tag management rules are configured to grab the data from the data layer and pass it into the correct variable in your analytics tool (i.e. Adobe Analytics or Google Analytics):
While in the tag management system, your analytics team may also have to leverage some tag management extensions. These extensions are pre-packaged snippets of code that help streamline data collection. Some come from your analytics vendor, and others are provided by third parties:
When you talk to people about seeing a “360-degree view of the customer” (which you really can’t, by the way!), part of what helps you do that are these extensions that manage IDs for those who use your digital properties.
But we’re not done yet! Your analytics team also has to manually configure all of the variables in your digital analytics tool. They need to provide a name and a series of detailed settings for each data point. Doing this incorrectly can lead to inaccurate data.
By this time, your developers have made good progress on pushing data to the data layer. But guess what? Often, the developers don’t understand all the details that have to be perfect when passing data. They make mistakes that have to be checked for data quality. As code is written and pushed, someone or some technology must verify that the data being pushed matches the desired tagging specifications. If the values are formatted incorrectly, that can break your dashboards. If they mistakenly pass a number as an integer when it is supposed to be an alphanumeric string, your data can be wrong. If they don’t know that an sProp can’t have values that are more than 100 bytes (and why would they know that!), then your data could be truncated, and you might not be able to see your customers’ pathing journeys.
Overall, data quality assurance is very time-consuming and has to be done both at the time of tagging and monitored after code is in production. Most organizations spend very little time and effort on quality assurance because it is seen as a “nice to have” function. But if your data isn’t checked for data quality, all of the pretty dashboards you have might be meaningless. You might even be giving your CEO recommendations that are wrong because they’re based on faulty data! Bad data also leads to painful “data fixing” exercises in which your analysts might spend as much time segmenting/filtering data as actually using the data!
Once your analytics team thinks that everything has been tagged properly and the data quality is good enough, data will start flowing into your product analytics data set. But at this point, it is just data. Your analytics team then has to create reports and dashboards that will help turn the data into answers to business questions. This doesn’t happen automatically or by invisible analytics elves! Your team has to go back to the original business questions and create reports, charts, and dashboards that they feel will help answer those questions.
In addition, your analytics team will have to create conversion metrics to help provide some data context. For example, they can’t tell you that sales have gone up 10% without also explaining that during the same period, visits went up 30% during the same period! Doing so would lead you to the wrong conclusion! In addition, there may be times when you don’t want to look at all of your data but rather just a subset of that data. For example, they may want to see how your website is doing for people who added over $300 to the shopping cart but didn’t purchase. Which products were they looking at? How often had they come? Doing this more detailed dive into the data requires the creation of visitor/visit segments. Building segments correctly is a tough skill in itself!
As you can see, creating reports, conversion metrics, and segments takes time and experience. (By the way, many data people are numbers folks who aren’t always great at making things pretty, so consider cutting them some slack by not focusing on how pretty the reports look!) Try to remember how much work has gone into getting you data as described previously!
Do you want reports or do you want insights?
Of course, the end goal isn’t to simply provide reports. Real analytics teams strive to deliver insights! They want to use the data, reports, metrics, and segments to learn something valuable that can then be communicated to you so that new ideas and strategies can be employed. So don’t simply ask the analytics team for data, let them tell you what they think the data is saying and what ideas they have to improve conversion. That is the fun part of the job, so please let them work with you on that. They’ve gone through a lot of detailed and difficult work to get to this point, so you owe them that (in my opinion!).
After identifying some cool insights, the next step is to do some testing of new ideas. One of the cool things about the digital world is that you can try new things on small cohorts to see if your hypotheses are correct and help boost conversion. Of course, this requires an entire new workstream to identify tests, make new content or creative, deploy them through a testing tool, monitor their results, and make recommendations. But if done correctly and in concert with the steps above, you can turn your digital properties into a virtuous cycle of improvement over time and feel confident when you say that you are data-driven!!
But, keep in mind that everything we have just discussed is only related to digital data. You likely have non-digital sources (i.e., call center, CRM, etc.) that have data about your customers. To be truly data-driven, you need to tie together all of the customer touchpoints. That is why many companies invest in digital analytics tools that feed data into your internal data marts/data lakes or a Customer Data Platform (CDP). Your analytics team may have to send all of the digital data we just discussed to even more systems where they are unified and merged with non-digital data. This allows you to make more macro-level conversion metrics, segments, and dashboards using BI tools.
Oh, and then there are privacy regulations (GDPR, CCPA). Your analytics team has to ensure that they are honoring your customers’ privacy preferences and operating within legal compliance. At the same time, you may have read that browsers are killing off 3rd party cookies, which brings with it even more challenges for your analytics team!
Maintenance
Did I mention that the entire process described above has to be repeated for every new business question? Being data-driven means that your digital analytics team has to be agile enough to keep pace with the ever-changing needs of the business. If your organization requires weeks or months to answer new business questions, then you are not data-driven. You need to help your data team help you by providing the resources they need to perform all of the above activities and do so promptly.
What Have We Learned?
If you have made it this far (and I can tell if you have by using tools like this), I hope that you now have a better appreciation for what your teams go through every day so that you can brag to your other executive friends that you are data-driven!
Now you may understand why your analytics team is always asking for more money, tools, and headcount. They are likely trying their best to help you and the organization, but they are dealing with difficult and inherently detail-oriented tasks. This may be why employee turnover in the analytics industry is much higher than other industries. When your employees leave, a lot of historical knowledge about the process described above leaves with them! I encourage you to think about what I have described and come to terms with how much work is involved. Thank your analytics team, and try to support them as much as you can!
Below is a diagram of many of the items that I described above:
In most organizations, each of the steps outlined are done manually, and information is stored in spreadsheets and text documents. Our team at Search Discovery has created a new software product called Apollo that serves as a platform for all of the information related to an analytics implementation. Apollo leverages a relational database and APIs to streamline and interconnect all aspects of the implementation so your analytics team can get through the processes I described above in half of the time it has traditionally taken. I recommend that you or your team check out our new Apollo product and see how it can help make everything described here easier (your team will thank you)! You can learn more at www.apolloplatform.com.
Wishing you and your organization a more data-driven future!
Sincerely,
Adam Greco
