I’m writing another post about media attribution because y’all keep arguing about media attribution. Here’s how to do it at every size.
This content was originally published in the No Best Practices newsletter on 07.28.2024.
Another attribution post? Yeah. Because there are still a ton of outstanding questions on this topic. In-platform metrics vs MMM models vs holdout testing–brands want to know what solution is right for them. Of course, the answer is “it depends”.
There are four “levels” when it comes to media mix complexity. Each “level” has different questions it needs to answer and different attribution needs. Hopefully this post will help you answer the attribution question once and for all so you can stop wasting time and money.
What Is The Point Of Attribution?
When you’re managing a paid media budget, you’re always trying to answer two questions:
- Does total spend need to go up or down?
- If I need to increase or decrease spend, which channel(s) and campaign(s) should I adjust?
There are several tools available to answer these questions:
In-Platform Reporting: the metrics provided within the Meta, Google, etc. ads manager.
Business-Level Tracking: using a spreadsheet like this one to track the correlation between new customer sales and changes in ad channel budgets.
Plug & Play Models: these are tools like Triple Whale and Northbeam that ingest all of your data and assign a modeled ROAS to various channels and campaigns based on their own statistical models.
Custom Models: you hire a data scientist specializing in marketing attribution to assess your unique situation and build you a custom attribution model to assign a modeled ROAS to various channels and campaigns.
Experimentation: you run statistically sound experiments (usually geographic holdout tests) to model the lift in sales of a specific channel or campaign. You can partner with a service like Measured or Haus to run these tests or DIY.
Yeet It: go with what you feel, or decide that certain activities have a value that goes beyond immediate sales impact. This method is beloved by many high profile CMOs and founders, so I might as well include it, but YMMV.
The Four Levels Of Complexity
When it comes to a brand’s media mix, maturity and level of awareness, there are four levels of complexity:
Level 0: Pre-PMF
At this stage the brand is not able to drive new customer orders profitably and predictably.
The most common scenario here is a brand using Meta ads as its go to market. The brand can’t find consistent winning ads. The brand misses its CAC:AOV target more days than it achieves target. Ads might “work” for a few days, but they fatigue quickly.
Brands at this stage shouldn’t be worrying about attribution…but that doesn’t stop them from trying!
If you have one major acquisition channel (Meta) and your business-level results aren’t meeting your goals, an attribution tool isn’t going to help you. It doesn’t matter which ad is truly the best performer if none of them are hitting your goals consistently.
At this stage, brands should compare 7 day click Meta reporting to brand-level new customer reporting to see how closely the data sets match. There is a chance that tracking issues (or some other issue) is causing a data loss on Meta or new customers are discovering your brand in some other way or Meta is reporting on conversions from new and repeat customers.
If you think your in-platform reporting is “wrong”, take your best performing ad and put enough budget against it to increase your average daily order count by 30%.
Example: over the past 30 days you drove an average of 10 orders per day. Your AOV is $50. Your in-platform ROAS is 2x. Set a daily budget of ((10 x 1.3) x 50 x 0.5) = $325/day.
Run this single-ad campaign for a week and track business-level new customer sales. Compare business-level CAC:AOV to in-platform CPA:AOV.
Did your sales go up? If the answer is no, your “best” ad isn’t performing. You have a problem with creative, offer or factors outside the ad account. This isn’t an “attribution” problem.
Is business-level CAC:AOV worse than in-platform CPA:AOV? Set a more aggressive in-platform goal for your ad spend and sharpen your approach to achieve it.
Final note: at this stage, when an ad is working in Meta, you’ll know it. You’ll be able to double and triple spend against the ad without impacting performance. And you’ll see your sales go up at the business level.
If you have doubts at this stage, stop dicking around with attribution and focus on testing more ads.
Questions You’re Trying To Answer:
- How does my in-platform reporting correlate with my business-level results?
- Am I spending enough in my key acquisition channel?
- Is my best performing ad good enough to scale?
Relevant tools: In-Platform Reporting & Business-Level Tracking
Level 1: One Major Acquisition Channel
At this stage the brand has one advertising channel that serves as the primary source of new customer acquisition. When the brand scales spend in that channel, the impact on new customer count and spend can be seen almost immediately.
For most of you reading this, that channel is Meta. The brand might be running one or two other paid channels like Google ads, but those channels don’t push new customer demand into the business at scale. I.e. the brand could not double Google budgets and see a noticeable bump in new customer sales.
If Meta ads (or whatever the main acquisition channel is) were turned off, new customer sales would decline by 70-90% almost immediately.
At this stage, brands want to know if and when they should increase daily spend. You can get a yes/no answer via business-level tracking. I outline my process for answering the question here.
Once you have a yes/no, the next question is: what channels or campaigns should I adjust. You can use logic and in-platform reporting to answer this question. Logic: increase spend in your most incremental activities, decrease spend in support activities or under-performing campaigns.
You can use in-platform metrics to identify under-performing campaigns. The step-by-step process for doing this is going to depend on your account structure and relative level of consolidation.
If you have one big scaling CBO, you simply reduce campaign-level budget. If you’re using cost caps, hypothetically you never have to make these decisions (don’t know if I believe that).
Questions You’re Trying To Answer:
- How does my in-platform reporting correlate with my business-level results?
- Should I increase or decrease my daily budget in the key channel?
- If so, which campaign(s) should I adjust?
- Should I increase or decrease the budget in any support channels I might be running?
Relevant tools: In-Platform Reporting & Business-Level Tracking
Level 2: Two+ Major Acquisition Channels
These brands pair their main acquisition channel (usually Meta) with at least one other channel that brings in qualified “upper funnel” customers in a scalable, consistent manner.
Popular support channels include TikTok (paid and/or organic), influencer seeding/affiliate, Youtube and even Meta “top of funnel” campaigns.
The support channel allows the brand to increase the budgets of its Meta conversion campaigns while maintaining its target CAC:AOV ratio. If the support channel was turned off, the Meta conversion campaigns would become unprofitable at their current spend level.
If a Level 2 brand turned off its two key digital channels, sales would decline by 70-90% within a month. The majority of awareness and demand is being driven by paid media. This is a brand like HIMS, not a brand like Nike.
The essential question you’re trying to answer here is the same: should I change my daily spend? If so, how do I allocate it across channels to yield the best result?
Let me drop a truth bomb on you: you’re never going to max out the potential of your ad budget every single day. There are too many variables to model, and many of those variables change frequently. Accept this truth and preserve your sanity.
Business-level tracking is still your source of truth here re: should I increase or decrease budget. You can then pick ONE attribution methodology to help you make that decision (plug & play model, custom model or experiments).
Before you pick your ONE methodology, you might want to test a few candidates from each methodology and determine which one works best for you. But once you pick ONE methodology, stick with it for at least six to twelve months.
The amount of time you invest in this decision should be proportionate to your ad budget. At the end of day, common sense is still going to guide a lot of your decision making here.
You should vet new channels or longer lead channels for their ability to drive incremental sales–usually with geo holdout testing.
Once that has been established, you should re-test periodically. If you need to cut budget, cut from the least incremental/longest lead channels first. If you’re hitting a wall, try reducing the budget in Meta/Google and investing in building TOF.
Questions You’re Trying To Answer:
- Should I increase or decrease my daily budget?
- Which channel(s) should I increase or decrease?
- Within each channel, which campaign(s) should I increase or decrease?
Relevant tools: In-Platform Reporting, Business-Level Tracking, Geo Holdouts and (maybe) ONE other methodology
Level 3: Brand Awareness Ecosystem
These brands aren’t dependent on any single advertising channel, or any single channel pairing. They are truly diversified. Getting here takes a combination of media investment, physical distribution and time: the recipe for mental availability.
Think about a brand like Nike, Coach or Chanel. If one of these brands turned off its Meta ads spend, how would that impact sales in the eCom channel? For the brand as a whole?
A Level 3 brand is rarely a DTC pure play. It typically has distribution in the physical world via its own stores, wholesale partners or both.
If you think you’re in this category, turn off your Meta or Google ads for four weeks. If average daily eCom sales drop by more than 20-30%, you’re a Level 2 or Level 1 brand.
There are several factors that make media investment decisions very complicated for Level 3 brands:
You can’t assume your Meta prospecting campaigns are incremental. Even if you get in front of folks who never purchased from you before, you can’t assume the Meta ad influenced the purchase. There are too many other touchpoints where a random person may have heard about your brand.
You’re fighting for market share vs building awareness. Imagine you’re Victoria’s Secret–you own slightly less than a third of the lingerie market by sales volume. Conservatively, one in five US women 18+ has purchased something from your brand. Growing sales isn’t just about capturing in-market demand. It’s about winning back lapsed shoppers, growing share of wallet and acquiring new customers.
You’re (probably) running longer lead marketing campaigns. OOH ads, print media, events, celebrity activations–you have to book these months in advance. You can’t flex budgets up or down in the same way a Level 1 brand might.
Office politics & “brand guidelines”. There are a bunch of factors plaguing “complex matrixed organizations” that prevent brands from maximizing their return on marketing investment. But the outcome is that many brands iterate on “what we did last year” vs building a customized plan for today’s landscape and goals.
Level 3 brands probably should pick ONE attribution methodology and stick to it, but that rarely happens, especially if the brand’s founding pre-dates the mass adoption of digital marketing.
Questions You’re Trying To Answer:
- Are my “performance acquisition” channels like Meta and Google truly incremental?
- Are my awareness campaigns doing what I intended them to do?
- Am I maxing the efficiency of my retention spend? Has it reached a point of diminishing returns?
- Should I flex budgets up or down to the extent that is possible?
Relevant tools: Whatever the key decision maker thinks is right