Advertising Attribution Platforms are Horoscopes for Marketers. Triple Whale vs. Northbeam vs. Hydros

Which attribution platform is the "real" psychic?

Privacy changes surrounding cookies have made tracking much more difficult than it was in the past. For many performance marketers, 30 – 80% of their users can no longer reliably be tracked. If users can’t be tracked the performance marketing system based on allocating marketing dollars to the channels with the best return on ad spend falls apart. There are a few ways to mitigate this data loss. Facebook and Google have server-side tracking protocols that can claw back a small sliver of the data, from what we’ve seen, less than 10% of what’s been lost. But even with that small add-back some things are now completely invisible.

Audiences: Dying
Impression tracking: Nearly dead
Geo-targeting: Weakened
Retargeting: Mortally wounded
Multi-touch user journeys: out the window

Marketers have long been addicted to this tracking data. Transitioning from data-based decisions to unproven decisions, that are “data informed” but mostly gut level, is terrifying.

This is where the miracle of the Data Attribution Platforms comes in. Northbeam, Triple Whale, Hyros, Wicked Reports, and more all promise to restore lost data and attribute it to the proper ad channel so you can still make data-based decisions. These tools are amazing at data visualization but are they any good at actually determining attribution? In a word, no.

Let’s talk through how these tools work and the incentive structures behind them. Every ad platform has an incentive to claim the most credit for conversions. Google wants more money for Google, Facebook wants more money for Facebook. The ad platform tracking tools are designed to attribute tracking to the ad platform. Third-party trackers like Adobe Analytics, Google Analytics (serious conflict of interest), and Matomo are supposed to be ad platform agnostic.

There are four basic ways for a measurement platform to take credit for a conversion (attribution).

Cookie-based: The user allows cookies in their browser, either via opting in or not opting out, and the browser allows sufficient duration of the cookie so that the tracking platform has visibility that an ad impression or click was part of the user’s journey.
Server-side: The user is authenticated either by the platform or the site as a logged-in user. Their session data is stored in log files and stitched together to create the journey.
Email add-backs: The user’s email is registered at the time of an impression or click. The eCommerce platform sends the ad platform email data about who purchased. Those conversions are added back into the ad platform when the data is reconciled.
Modeled conversions: This is a simple multiplier applied to the channel. “We think you have 15% more conversions than we can see through any of the other methods.” Channel x 1.15. This quickly devolves to playing favorites for channels.

Modeled conversions are where we lose our tether to reality. It’s where the platform takes credit for conversions that are in the direct or organic buckets. There are some obvious limits to how much credit a platform can take. It can’t take credit for double the actual conversions you had in your Shopify dashboard without losing credibility, but otherwise, this is where the real magical thinking begins.

Attribution platforms do not have a feedback mechanism to tune towards reality. They have feedback mechanisms to tune toward marketers’ feelings.

Tuning to results isn’t really possible. The algorithms aren’t transparent. There is no feedback loop that informs the algorithm how the business is doing, or about the myriad of variables in the creative offer or the marketplace.

Even the tried and true methods for brand measurement won’t work. Brands often conduct surveys to see if people remembered the content of an ad and how they feel about the brand. Lift in sentiment or awareness is a signal that can be used to inform the media mix. If there is no cookie, there is no way to reach people exposed to the ad to survey them. To my knowledge none of the popular attribution platforms are doing this.

There is a feedback loop for feels. Marketers right now are head to head testing attribution platforms and determining winners based on what feels right to them.

If you like TikTok, you’ll want an attribution platform that has a bigger TikTok multiplier. By the way, I love this follow up exchange:

Machine learning is right. Somebody typed a multiplier into the computer machine and learned that marketers who favor that channel liked it. This tuning to the feels is a time-honored practice of psychics and astrologers. If this is machine learning then Miss Cleo was an icon of machine learning.

Ultimately there might be some good reasons for using these platforms even if you know that the results are mystical. For one, you have to “model” your conversions. There are going to be some marketing efforts that certainly have some impact on conversions that is not easily measurable. You’ll need to extrapolate the results and outright guess at others. An established platform can give you a framework for those guesses based on their guesses.

These platforms make visualizations that can be presented to whoever is funding the marketing efforts. The visualizations can be key in justifying budgets.

Ultimately, we’ll all be better off not relying on these platforms. The best thing to tune is the marketer’s judgement. Hard earned experiential learning can beat mystical “machine learning”.