Article

Third Party Attribution Is BS

Analytics Attribution Consumer Behavior

August 28, 2023

I figured I’d spice things up with a fun, usually-controversial topic: Attribution.

I’ve had dozens of conversations with brands, performance marketers & investors, all squarely focused on answering the question: “what channel(s) are causing sales?” Many of these conversations were driven by broader economic uncertainties, with the people involved diligently trying their best to put their organizations in positions to (at the very least) survive any economic downturn, and at best, to emerge stronger.

That question has led virtually all of them to adopt a third-party attribution solution – whether it’s Triple Whale, Northbeam, Hyros, Rockerbox, Polar Analytics or Marketo Measure – and none of them are closer to answering the initial question correctly.

This is a statement I would not have made 5 years ago, back when I believed attribution was a solvable problem.

After years of testing, advising and using multi-touch attribution tools, I’m more confident than ever that attribution is, at best, redundant and at worst, a farce.

To answer why, let’s take a step back:

When a brand goes down the attribution rabbit hole, they’re almost always trying to answer one big question:

The big question: what is driving sales/leads/[insert thing we care about here]?

And, like all big questions, this one has some follow-ons:

Where* are we wasting resources due to inefficiency or overall ineffectiveness?
Where* could we realize incremental positive outcomes, if we allocated additional investment?
How much cannibalization is occurring between new + existing customers?

* = channel, platform, tactic, creative, whatever.

In a vacuum, those are crucial questions to ask. And they so happen to be the exact questions multi-touch attribution platforms (“MTA”) claim to solve.

If you’re unfamiliar, a typical MTA has three core functions (and a lot of supporting bells & whistles):

Reporting – these purport to show marketers/operators the 30,000 foot view of their investment allocation AND the customer journey (similar to Google Analytics 4), vs. the channel-specific view you get in an ad platform.
Attribution – each platform has their own “secret recipe”, but they all do the same thing: attempt to quantify how much each touchpoint contributed to the end outcome you want to measure (leads, sales, demos, whatever).
Optimization – the final piece of the puzzle synthesizes (1) and (2) into a set of optimization recommendations (read: shifting budgets around based on their model + your current spend per platform).

In theory, this sounds great – which may be why there are so many of these platforms out there. Unfortunately, the reality of what these platforms deliver tends to be radically different than what is promised.

Take, for example, the following customer journey for a jewelry brand I’ve recently audited:

User learned of the brand via an influencer collection announcement.
User searched for the brand directly on Instagram + visited profile.
User took no further action that day
The next morning, the user was served an IG ad for the brand (no mention of influencer or collection)
User clicked on ad, visited website, did NOT purchase, left website
User discussed brand with two friends during lunch, both of whom visited website + IG
That evening, user visited website via branded search + booked a consultation for custom jewelry

In this case, you’re likely to see 4 conversions recorded if you sum across all platforms: (1) the Influencer; (2) Organic Social; (3) Meta Ad; (4) Google Branded Search Ad. But you didn’t sell 4 necklaces, you sold 1. Fundamentally, this is the problem MTAs purport to solve. By assigning credit to each touchpoint in the journey, and having the credit sum to the number of sales, attribution platforms claim to make sense of the above journey, thereby helping brands make better investments & decisions.

So, put this to the test. While it may not appear to be the case, the above is a delightfully simple customer journey – and as such, it’s one a MTA tool should be able to get close to right.

Common sense probably says the following:

The Influencer played a significant role in discovery – absent this, the user likely doesn’t learn about the brand.
The friends AND Instagram played a significant role in consideration + deal acceleration
Branded search played a relatively small role, primarily in expediting the transaction + deal insurance (e.g. preventing competing brands from trying to wedge into the consideration set – which is valuable!)

Looking through this brand’s history, I can see the above (or a close variant thereof) is a relatively common path to purchase. This is likely true for many higher-end and/or higher-consideration products/services that use a combination of Meta, Google & Influencers.

Yet, when you look at the MTA “credit” allocation for the above conversion:

0% Influencer

0% Organic Social

0% Friends

34.5% Meta Ads

65.5% Branded Search

Whoa. Pretty much the polar opposite of what most marketer’s intuition (and common sense) would say. But this isn’t a surprising result to anyone who has dealt with MTAs – for the following reasons:

MTAs don’t have a way around walled gardens – for instance, the organic view of the influencer’s content, and the profile visit.
Only data that can be added to the view is considered – we are nowhere close to the point where a lunch with friends is observed, recorded + uploaded to MTA platforms (and honestly, the world is better for it, dashboards be damned).
Parsing intents + dynamically adjusting attribution models based on campaign types is a work-in-progress – and always will be. Any good (and honest) digital marketer will tell you (1) not all branded search is the same and (2) basic branded search is 90% insurance + 10% opportunism. No one is divinely inspired to search for your brand. They heard about it somewhere and decided to google it. The fact that you’re there (and boxing out the competition) is good, but branded search deserves relatively little credit.
MTAs don’t have a magic key to unlocking more data – while many platforms have rolled out their own “pixel” or “tracking tag” or whatever, here’s the reality: none of them can get access to data that you can’t get via Google Analytics 4 + some ingenuity. The *only* advantage MTAs bring is that they centralize otherwise-available data from multiple platforms (which you can do with Funnel or Supermetrics or any number of other platforms!). There. I said what I said.

And, as a brand owner or marketer, if you were to act on the above, you’d likely: (a) not renew that influencer; (b) pull money from Meta Ads & (c) throw more into Google Branded Search – which is pretty much the exact wrong thing to do. The most likely outcomes are either (i) juice sales in the short-run, as you squeeze the proverbial water from your now-shrinking sponge and/or (ii) actually kill your scaling altogether, as the MTA effectively recommended to shut off the entire demand creation engine driving this brand.

The most insidious part of the above is that the results aren’t going to be obvious in the short-run; this is the death-by-slow-boil. The immediate results of following the MTA optimization recommendations in this case are likely to be a negligible decline in sales and a rapid spike in efficiency. The mid-to-long term results? Cratering demand as a progressively larger share of your revenue is coming from a continually-shrinking segment of your existing customers (your brand loyalists). But even loyalists have a limit – both emotionally and financially. Once you hit that, it’s all downhill, and fast.

While this particular example is hypothetical, I’ve personally watched this scenario unfold far too many times. And in each instance, I’ve heard the urging of MTA providers that we’re saving money and doing better – all while observing a progressively larger share of sales coming from a perpetually-shrinking core base of customers. That’s how brands die.

If you’ve stayed with me this far, you’re probably wondering, “So MTAs aren’t great. What’s the alternative?”

We use a combination of three things: (1) our actual sales / lead data from our website/CRM; (2) Google Analytics 4 & (3) a data ferry to import cost data from ad platforms to GA4. For most brands, GA4’s Data-Driven Attribution (“DDA”) is every bit as accurate as anything you’ll get from a third-party MTA provider, it’s FREE, and it integrates directly with Google Ads.

That last part is critical, because it makes determining the contribution from your other platforms delightfully simple. And while DDA isn’t transparent, neither are any MTAs. At least with Google Analytics 4, we can export hit-level data to BigQuery to run our own analyses.

Once you have those three things, determining your Marketing Efficiency Ratio (MER), along with your Acquisition MER (aMER) and Retention MER (rMER) is relatively simple. So too is setting your in-platform cost caps / target CPAs / bid caps / max CPCs. Curious about how to do that? I outlined the process we use here.

The added benefit of the above approach is that it avoids adding non-interactive/synthetic layers into your overall optimization process. When you introduce an outside data point into your allocation + optimization decisions (for instance, a recommendation from a MTA platform), it forces you to compare your in-platform data vs. your MTA tool vs. your actual business data vs. your web analytics data – which is enough to make even a data scientist’s head spin.

This is further complicated by the fact that all major ad platforms (Meta, Google, Amazon) only optimize off their own data signals (for good reason: if they considered outside/3P signals, it’d be pretty easy to game the system).

But what about impulse-driven brands?

Most of what I’ve said above applies to brands with longer consideration windows, somewhat-to-very complex journeys, and relatively high price points. So, what about brands that focus primarily on high-impulse/low consideration purchases?

Well, to be blunt, if your brand is a low-consideration/high-impulse purchase, what are you gleaning from MTA that you wouldn’t from DDA or Last-Click in GA4? Nothing? OK. So take that money you were going to go pay a MTA and spend it acquiring more customers – at least then it’ll add some value to your business.

The Bottom Line:

Attribution is re-arranging the food on your plate and claiming you’ve eaten a meal. You can certainly do it, but don’t expect to leave the table feeling full + ready to take on the day.

At the end of the day, you don’t need a fancy third-party tool to tell you where to invest your money or which channels deserve “credit” for which sales. You need to use the data you already have for free (CRM/CDP, web analytics + ad platform), alongside common-sense, intentional, focused decision making.

And if you really want to spice things up, go either (a) audit your customer journeys for recent buyers (yay User Explorations in GA4) or (b) talk to a few new customers + ask them how they came to buy/submit a lead). I promise you that the above will give you more insight and clarity than any MTA – and you can take that $1,000+ per month you’d spend on the tool, and use it to grow your business.

Related Insights