Good morning,
Weâre back again with another DTC Summer Series newsletter takeover!
For those of you that are new here, each week through the summer, weâre delivering you learnings directly from players in the DTC space.
This weekâs newsletter takeover is written by Margaret Fortner, VP of Marketing at Glamnetic.
Grab a coffee, a snack, get comfy, and letâs dive in.
â
Hereâs what youâll find in todayâs DTC:
âď¸ We dive into the world of attribution and how brands can remain profitable by harnessing data tracking.
đ° Learn how the iOS 4 disruption impacted advertisers and solutions going forward for multi-touch attribution.
đ AndâŚwant to hear more from Margaret? Hit play on this throwback podcast episode.
Youâre reading this newsletter along with new subscribers from: Imperial Bottle Shop, Streamline Brands, and Mez Foods. đ
đ How to Succeed in Attribution by Trying Really Hard â Part Two
đĽ iOS 14 Disruption
At a high level, iOS 14 turned user data sharing from an opt-out to an opt-in.
As anyone who once had double opt-in turned on for Klaviyo lists knows, opt-in rates will substantially drop regardless of what youâre asking a user to do. The broader rollout of iOS 14 also increased user awareness of data tracking, further lowering the likelihood that someone would willingly turn over their personal information and activity.
â
This basically placed a blindfold on ad platforms and told them to drive in the Monaco Grand Prix.
Meta and co. had increasingly relied on this data for optimized campaign operation. With the system fully disrupted, there was a drastic drop in reported conversions. Even more crucially, the actual performance of the campaigns dropped as the audience modeling based on user behavior and activity was impacted.
These two factors combined to remove even the basic directionality of platform attribution. In many cases, we went from over-attribution to under-attribution. Meta responded by introducing modeled conversions to questionable results.
The outcome is well-documented at this point: brands cut investment, overall profitability plummeted for many, and many of us had no idea what to do.
When the relationship between reported actions and business results is muddied, making the right decision becomes substantially more difficult.
â
â° Where We Are Now
Just like teenagers are rocking JNCOs, everything old is new again - or everything that was old and consistent to certain brands is now more readily available and accessible for everyone.
Legacy MTA platforms commonly went hand-in-hand with the other cost and manpower-intensive large brand advertising tools in ad servers and ad trafficking teams.
Our modern MTA platforms, like Northbeam and TripleWhale, are Shopify-optimized, easy to implement, and present with a UX that is very familiar to those accustomed to spending their time on ad platforms.
These tools leverage the combination of existing UTM tags, often with a tracking string added and a unified platform-agnostic pixel.
While there are a wide variety of attribution models available within most of these tools, the maximum incrementality over Google Analytics comes from options that include view modeling.
The dominance of video-on-top advertising platforms has increased the frequency of conversion paths with clickless interactionsâthese would be completely lost with sole reliance on click-based UTMs.
At a high level (and itâs all a black box at the end of the day), these clicks+views models rely on a combination of UTM data, modeling based on platform-level session volume, and other inputs such as post-purchase, âHow Did You Hear About Usâ surveying, and platform data to create predictive values of those non-click touchpoints.
Comparing a clicks-based model to these clicks+views models, youâll most commonly see conversions shift from bottom-of-the-funnel click-eaters like email and Google Ads/Organic to our mid- and upper-funnel platforms like Meta and TikTok.
In our experience at Glamnetic, Northbeamâs clicks+views model even showed higher conversion value attributed to TikTok Ads than the platform itself did. This allowed us to take advantage of a low CPM opportunity soon after iOS 14 when other brands more reliant on platform data had pulled back their investment.
In the three years following this seismic shift, Meta has aimed to rebuild its data set largely by promoting the aggregation of campaigns and actions. Whereas in 2015 you might have separate ad sets for desktop Right Hand Rail and Mobile Facebook Audience Network, Metaâs best practice has become throwing a bunch of creatives into a single, broad audience-targeted campaign reaching across website and shop platforms.
Both Meta and TikTok have launched their own Shop platforms to further consolidate user data and actions and, in TikTokâs case, to own outright vs. sharing it with advertisers.
At the same time, many brands responded to the initial drop in Meta effectiveness by diversifying their media mix. Advertisers who might have once spent 80% of their monthly budget on Meta are now spread between TikTok, Snap, Pinterest, TV, podcasts, YouTube, and, of course, Meta.
While platform effectiveness may be trending back up, the increased complexity of advertising mixes limits even the directional validity of the platform-specific reported results.
Beyond that, isnât more data always more fun?
đ How to Succeed in Attribution by Trying Really Hard â Part Three
đ Yes, More Data is Always More Fun
We first implemented Northbeam as our MTA partner at Glamnetic in January 2022 and have yet to consider switching. We leverage Northbeam not just as a platform but also as a connector to our Business Intelligence suite in Looker.
We look at Northbeam data daily across our ad buying and creative teams. Our web team frequently uses it for page and site optimization, and we also use it regularly enough on an ad hoc basis due to curiosity that our co-founders both have logins.
That being said, I would never take Northbeam at face value by itself. Even the most sophisticated AI-ML-NLP-driven attribution model in the world cannot follow every customer around to note their actions and what they interacted with (if you have one, please share).
At their best, these attribution models are highly accurate directional tools. If your MTA tells you that a specific Meta campaign is so amazing that you 5x its budget, does that number really matter if your MER takes a hit?
Anchor metrics are crucial to getting the most out of any attribution tool or model.
An anchor metric is a core business metric that is unaffected by any sort of attribution modeling. MER is one example.
At Glamnetic, we look at spend:net and spend:net new. This is simply all spend over a given period divided by all net revenue over that same period (for spend:net) or all first-time customer net revenue (for spend:net new) as defined by Shopify over that same period.
Spend:net will remain the same regardless of whether Iâm looking at a 7dc+1dv attribution model in Northbeam or Google Analyticsâ most basic last-click model. This serves as an anchor against which to compare your chosen attribution tools over a given period of time.
When we first tested out Northbeam, I tracked our spend:net and spend:net new against total reported paid platform ROAS and individual paid platform ROAS daily over the trial period.
We saw a clear and close directional relationship where, when Northbeam ROAS was higher, spend:net was lower (which is good for those less familiar), and gave us the confidence to move forward in this partnership.
Two years later, I still run through this exercise on a bi-weekly basis looking at the previous two weeks, as well as a core part of any ad hoc performance analyses.
While last click attribution is a method that should never be used by itself, we use it as another piece of the triangulation puzzle, particularly when looking at comparison periods.
Any attribution methods that leverage view modeling involve some form of statistical projection, and platforms can update these models over time. Google Analytics (or Shopify) Last Click serves as an apples-to-apples comparison toolâyou wonât see any major changes in how it works unless your team gets sloppy with UTMs or was sloppy in the past.
The final core piece we leverage is post-purchase How Did You Hear About Us (HDYHAU) surveying, often combined with UTM click-based attribution.
This is a flawed method by itself as well, as itâs fully reliant on the customersâ memory and their own internal attribution. A customer may say that they heard of us through the branded Google ad they clicked on despite having seen 10 other ads across Facebook, Instagram, and TikTok that gave them that awareness to make the branded search.
Weâve also found that our HDYHAU results skew toward platforms where we invest more organic social efforts, and those platforms where we may have less of a presence are underreported even if we invest heavily on ads.
With all of this in mind, you might ask where the value for MTA platforms comes in. If youâre not asking that, and itâs obvious to you, youâre probably an ad buyer or someone whoâs done enough ad buying at some point in your life.
Different levels of attribution matter to different members of an organization. Our co-founders donât need to know that Ad A in Ad Set B in Campaign C on platform X has a 1.23 ROAS, but my creative and buying teams do.
Having a handle on the cadence and level at which to use these methods is important. Using Glamnetic as an example, I look at our ad-related performance via Northbeam and our spend:net / spend:net new daily.
Every week, Iâm also adding on a review of our cohort profitability metrics. Monthly adds a further layer of macro comparison metrics, business-level profitability data, and retroactive LTV analyses of historical cohorts. None of these views or KPIs are necessarily more important than others. If everything is set up correctly, Ad A with the 1.23 ROAS getting spend and improving in efficiency should translate to better business-level EBITDA, but I canât really know that on the day where weâre making the decision to turn that ad on or off.
Many other brands out there have done incredible work developing in-house attribution models, often in combination with many of these tools. If anyone has worked on something like this, Iâd love to learn more.
We view what we have in place as something easily achievable with a data-oriented ecommerce team and an ability to zoom in and out in levels of detail, as well as to weigh all of these data signals as core ingredients to making a great soup.
đ How to Succeed in Attribution by Trying Really Hard
Attribution â Itâs more than just a fun buzzword guaranteed to boost organic hits on your latest and greatest podcast episode.
The implementation of iOS 14 and broader shifting economic trends brought a renewed focus on profitability within the DTC world, which extended to the marketing space.
â
Additional scrutiny is placed on every dollar spent, and with that comes an expectation of confidence that what we think is working actually is.
I was lucky enough to have early exposure in my career to the multi-touch attribution (MTA) precursors to platforms like Northbeam and TripleWhale, which are more commonly used today.
Soon after that, the strength of the Facebook Pixel and the accessibility of its associated advertising tools pushed those platforms out of style, and many of us were reporting and making decisions based on platform ROAS without thinking twice. And at that point, it worked.
The point of attribution is not to have the coolest, most complex in-house statistical model or the flashiest machine learning-driven shiny toy â itâs to empower you with the right data to make the right decisions more often.
The perfect attribution model or platform doesnât exist, and, more often than not, the best approach is to triangulate between multiple available sources of truth, with a healthy dose of skepticism, to find the right mix for you and your needs.
â
đď¸ Marketing Attribution Pre-2021
At a high level, it all starts with UTMs and pixels. UTMs were primarily used for last-click attribution methods, which still stands as the default in most Google Analytics instances or your standard Shopify sessions report. These should be considered platform-agnostic tracking tools that are set up and owned by marketers.
Pixels are code snippets implemented on websites for user tracking and data gathering.
In the marketing realm, we interact with them most as platform-specific pixels. Everyone reading this has probably implemented a Meta Pixel at some point. As user traffic has shifted substantially towards mobile, pixels work closely with SDKs and other in-app tracking tools to create a more complete picture of user behavior. These mobile tracking tools are also largely platform-specific.
â
Platform attribution validity greatly increased over time with:
â
It should be noted that, during this time, MTA platforms existed and were leveraged largely by more sophisticated and/or more funded brands. Many brands also built their own data-driven attribution models that support decision-making.
For the majority of DTC brands, this pixel and UTM-driven tracking toolkit was not only easily accessible and understandable but also, at the very least, directionally accurate.
While you might get more juice with these additive tools even prior to 2021, the ease of implementation, cost, and broader awareness made standard platform attribution more attractive.
â
One major fault surrounding platform-specific tracking is that it naturally exists in a vacuum.
â
While platforms like Google Analytics intake the UTM data of all sources landing on your site, Meta only sees its own pixel. If a user sees ads across five platforms within the applicable attribution window, all five platforms claim that conversion.
With many brands spending 75%+ of budget on Meta at the time, it was highly likely that Meta had been involved in nearly all conversion paths, allowing platform tracking to be directionally relevant in most cases.
đ IN THE SOCIAL SPHERE
We had the pleasure of speaking with Margaret on the DTC podcast so listen to the episode here. We discuss why Glamnetic decided to invest in podcast ads and the most common mistake when it comes to A/B testing!
â
đĽ Got a B2B Biz?
Join dozens of B2B companies finding demand-gen success through our niche community of 160k brand leaders and founders this year. Talk to our team to learn more.
â
Have you heard our latest podcasts?
â
Donât forget to rate the DTC Podcast on Apple (âď¸âď¸âď¸âď¸âď¸)
â
DTC Newsletter is written by Rebecca Knight and Frances Du. Edited by Eric Dyck.
Was this email forwarded to you? Sign up here.
Please note that items in this newsletter marked with * contain sponsored content.
.svg)
.svg)
.svg)
.svg)
.svg)
.svg)
