Yes, there are numbers in marketing, too
During week 5 of CXL Institute’s Growth Marketing Minidegree, I wrapped up the data & analytics module with courses on Google Tag Manager (GTM), attribution, and Excel for marketers.
Google Tag Manager
To review from last week, Google Analytics (GA) collects, stores, and reports the behaviors of a website’s visitors, but only excels at the storing part. Google Tag Manager (GTM) is a separate Google Marketing product that focuses on collecting data that it can send to GA and any other site that collects data (Facebook, Google Ad Manager, etc). GTM is made up of tags, triggers, variables, and data layers.
Tags are code snippets for GTM to send your website data to third parties to store. Essentially: what you want GTM to do.
Triggers tell GMT when to fire tags and fall into four categories: pageviews, clicks, engagement, and custom triggers. Within these categories you can find actions such as playing a video, scrolling past a certain page depth, or clicking on a link.
Variables are the information GTM needs to do its job. GTM has built-in and user-defined variables; for example, which page was viewed or how long a video was watched.
GTM temporarily stores data into data layers, which function like a file drawer. Data is organized into key and value pairs and then sent off for more permanent storage to the third party tag-receivers.
Google Tag Manager Takeaways
I took the GTM section at a slower pace since 1) I was not previously familiar with the GTM interface and 2) fewer modules involved a hands-on component. There’s no GTM demo-account, and since most of the modules involved e-commerce and conversion tracking, there were fewer opportunities to apply my learnings to the Google accounts I use to track activity on my personal website. But I did set up the scroll depth tags and triggers, which I’m excited to see and learn from once future traffic fills out the event in my GA.
The attribution module broke up the two platform-heavy GTM and Excel classes with an interesting course that merged the theoretical and analytical components of growth marketing.
Attribution as it relates to marketing is the act of understanding the context behind the data you can access. Half-definitions usually focus on the distribution of value and finding a number you can plug into the ROI column of a spreadsheet. Although both can be parts of attribution, the field involves understanding the consumer. Not just customers, but all consumers.
Less than 1/3 of businesses are using defined attribution models, which is why 43% of companies say measuring ROI is their top marketing challenge. You can look at the cost of marketing efforts and the direct sales they caused, but it’s not that simple. First, how do you know how much revenue was a direct result of marketing efforts? And if marketing contributed a portion, how much? Marketing also has offline channels (TV, print, display ads) — how do you measure reach and impact for non-digital parts of the customer journey?
These questions sound similar to conversion optimization (CRO). However, CRO focuses on understanding an individual’s actions on one page, whereas attribution seeks to understand an individual’s interactions throughout their entire journey (every time they engage with a brand throughout their life).
The reason few businesses leverage attribution is because it’s hard, complicated, and oftentimes expensive. There are several models that serve as shortcuts to help marketers estimate the true attribution of marketing efforts.
When a user makes a purchase, the last click model gives all the credit (the dollar value of the sale) to the last click. If someone saw a brand’s Facebook ad, read their affiliate content, searched for the brand on Google, and then eventually clicked on their marketing email and made a $50 purchase, email would receive 100% of the $50 credit.
In the example, though, the customer had 4 touchpoints (that we know of) before they made a purchase. Had they not seen the ad, read the content, or researched organically on Google, they may not have purchased after receiving the brand’s email (or even signed up for their email list and received the email at all). So giving email all the credit doesn’t tell the full story.
The linear attribution distributes value through every interaction equally. So each of the 4 touchpoints (Facebook ads, affiliate content, SEO, email) would receive $12.50 (50/4).
Yet linear attribution isn’t all that accurate either, since it doesn’t show what had more impact and why the consumer interacted with each channel. Surely not everything marketing does is equally as valuable or effective. The time decay model distributes the purchase value through every interaction but with more emphasis on the most recent touch points. The percentage each channel receives is either static (fixed, decreasing percentages based on the number of interactions going back in time) or variable (takes into account the amount of time between each interaction to distribute the value back).
But time decay also assumes equal value — it just depreciates it over time. The position-based model gives 40% of the credit to the first and last touchpoints, with the remaining 20% equally distributed in middle.
This model still carries the issue that the marketer (or model) decides what’s most important.
If a company takes attribution seriously, they use a custom model. Custom models are based on an algorithm and use machine learning to find out which touchpoints are the most important and by how much. Custom models are time-intensive and expensive, but are the future for every company with the means to pursue true attribution.
Beyond how to determine attribution, the theoretical and strategic vision here is to truly be able to measure marketing ROI, make informed customer acquisition investments, and understand customer lifetime value. One strategic implication I found particularly interesting is propensity scoring. Attribution helps determine how likely someone is to convert and what is the value of their conversion. By using historical data to create look-alike segments, companies can start to sort users who have not yet converted into segments and customize their experience accordingly. Companies can tailor their marketing messages, provide appropriate content recommendations, and adjust the user experience depending on predicted likeliness to convert.
The Excel course reminded me of the 5 Excel-based OPIM (operations and information management) courses in Georgetown’s business core. The overview of advanced formulas, pivot tables, and formatting nicely wrapped up the data & analytics section with the platform we all know and love.