The Top 3 Things We’ve Learned From Our Data and MarTech System Audits

After years of helping companies build the best data and Martech stacks we've found a simple truth: when it comes to data, all the happiest companies are the same...

… but, just like how Tolstoy said in his opening of Ana Karenina, “All happy families are alike; each unhappy family is unhappy in its own way,” and in that same vein, every unhappy company has their own unique history and challenges that resulted in data systems that make some large part of their internal users unhappy.

At Pickaxe we’ve helped companies connect thousands of data sources, build massive data warehouses, custom CDPs, and ETL solutions. We’ve also had to “eat our own dog food” when building our own SaaS platform for automating data analysis and insights. We’ve seen the good and the bad, and we’ve noticed a few common trends, and most importantly, we’ve seen that the happiest companies with the best data systems all share many of the same core fundamentals.

Very rarely does anyone get it right from the start, but in the end, these are some of the most important lessons learned.

1.
A carefully designed data structure and schema that can be understood by non-technical users goes a long way.

A good schema shouldn’t reflect the organization of the company. Nor should it reflect the source data systems where data is coming from. A good schema reflects the needs of the business in a way that makes the data business ready, and actionable. 

When the schema is good, you know the company is both well organized and understands their core business problems. You’d be surprised how many times we’ve found even the largest, most successful companies struggling because their schema is thought of as more of a technical problem than a business solution.

When your data architecture is done well, the analysts, marketers, editors, and business users looking at the numbers relevant to their job know what actions users took on the site (or offline) to trigger those metrics: a Subscription Start is when a user clicked on a particular button and submitted the form, a Winback is when the same thing happened and the user previously had a subscription that lapsed, etc. In companies with strong data practices, the schema is carefully defined and shared so that a non-technical users can look at the numbers in a report and know intuitively what they mean.

2.
By "carefully designed" we also mean easy to read with consistent and clear naming.

And, as above, a good schema also means data is clear and actionable, and users shouldn’t have to dig through tons of documentation in order to understand what field names correspond to, or how to find the data they want to find.

A good schema should read like a plain language guide – metrics have labels that define what they mean, for example “New Customer” is easy to understand by everyone but “nCust” is not. Where possible, metric and dimensions are written as noun-verb pairs to describe actions – examples include “Watched Video”, “Browsed Product”, “Converted from Free Trial”, “Purchased Product”, “Signed Up”, etc. 

This way your analytics queries can directly connect to the core business questions that your team wants answered. “How many Users watched video and then Subscribed?” or “How many Returning Customers initially acquired from the Summer 2020 Campaign Purchased a new Item in 2021?”

This is also why Pickaxe’s Universal Schema is designed the way it is, and why doing Queries in Pickaxe can feel magically simple and intuitive, because we clean up all the messy data that’s fed in and turn it into a schema that everyone can easily understand. 

3.
Metrics are always thought of as being part of a group, or inside of a larger funnel.

Companies with unhealthy data practices tend to focus too much on single metrics. They’ll ask questions such as: How many users saw an ad and installed the app? How many users signed up for a free trial? How many purchased a product?

While these are certainly all important questions, they risk painting an incomplete picture. By themselves those metrics are all valid, but there’s a danger in not considering them as part of a larger funnel that shows how all the data connects, and how they relate to core business KPIs.

Data savvy companies tend to group metrics: app installs & account creations, free trial starts & free trial conversion, CPA & LTV, Product Views & Purchases, Views & View Time, Views & Completions, Reach & Engagement, etc.

This also helps ensure that data is collected and made actionable in the most useful ways. By thinking about how each metrics relates to other metrics, you can cut through the noise of having too much data, and ensure you are focused on the combination of metrics that actually drive your business.

Bonus!
A short list of the most common pitfalls we've seen from even the best companies.

  • Using email as a unique identifier. (Emails change!! We get it, sometimes you just don’t have a better option, but usually there’s a better way!)
  • Not properly “identifying” users on the front-end, or having any kind of inaccurate identity-graph. (If I’m in your system four times with four different attributions, which one gets credit?)
  • GTM events have grown haphazardly, and no one is sure exactly what each trigger does. Which trigger is a purchase? Which fires the exclusion list? And are all the properties needed included in the tags? And were they well-named?
  • Not having (or worse, having and not properly using) an attribution tool to properly identify which users came from which campaigns, or generally lacking in governance on campaign naming and UTMs.
  • Not carefully monitor attrition – how many existing users have gone dormant, haven’t viewed content, haven’t returned to the store, aren’t opening emails, etc. If you are focused on user totals without considering which users are truly active (or completing key goals), you are either fooling someone else, or fooling yourself
  • Paying to deliver marketing to customers who just bought! Exclusion lists aren’t automatically updated (across platforms, with offline events included, and – in the case of subscription services – refreshed with users whose tenure is longer than the social audience six month expirations). We’ve all seen ads for products we’ve just purchased, and sometimes that’s a good strategy, but often it isn’t. If you’re selling anything #preventInefficientMarketing to customers who just bought.
  • Creating lookalikes with the wrong customers. If you think you have some customers who are more valuable than others, then you shouldn’t be trying to find lookalikes for your least valuable users!
  • Forgetting that the funnel doesn’t end when a new customer converts, it keeps going until they start referring their friends!

If you want to see how we can help you give your MarTech and Data Systems superpowers, contact us!

Over the years we've worked with dozens of companies to evaluate, re-instrument, and improve their data analytics and marketing technology platforms, and through all of those experiences we've learned some interesting things.

learn more

Let us show you how we can help you make the most of your data.