You’ve heard the marketing analytics spiel before. It goes something like:
Marketing journeys are long, complicated, multi-channel paths. Tracking them is always imperfect, but if you employ an extensive, expensive, difficult-to-configure combination of tools, tracking systems, and deeply talented statistics professionals, you can build a high-quality, predictive attribution modeling structure that approximates reality with relatively minor gaps.
My counterpoint: don’t bother. Just trust your gut.
As a thought experiment, imagine a tracking system capable of capturing buyer journeys like the one below (an example from the very kind Liam Moroney of Notarize):
What would you need to attribute buyer journeys that *seem* like type-in/direct/branded search visits to LinkedIn posts?
- URL parameters or special URLs? Won’t work because the posts intentionally don’t include a link.
- A social analytics tool connected to the LinkedIn API that can show engagements? Still won’t work, because LinkedIn doesn’t allow the identity of engaging-users to be exported or tracked in 3rd party systems.
- Manual analysis of everyone who interacts with every LinkedIn post you publish? Besides being impractical, it would only work if Liam or his coworker actually “liked” or “shared” the post. However, if they did, you could then connect this up to a CRM that tracks email account signups to show that there wasn’t much time lag between the LinkedIn engagement by someone from Notarize and a signup from that company… except maybe they used a Gmail account when signing up? In which case all that effort is wasted.
Maybe you’re like Dan, below:
He’s imagined a system that might be imperfectly able to attribute at least some of the lift in traffic or signups to the LinkedIn post, or maybe just LinkedIn as a whole (with a whole mess of hand-waving assumptions that, in practice, are incredibly difficult). It requires at least two sets of expensive, third-party tools that need to be integrated with a CRM that also offers malleable conversion tracking and segment-able attribution.
The $1M Question: Is the Juice Worth the Squeeze?
Say you knew with absolute certainty that two paying customers signed up as a nearly direct result of a how-to video posted to LinkedIn. What would you do differently in your marketing? More of those videos than you were planning to do anyway? Less of something else that you couldn’t perfectly track?
Now try the other direction. What if your perfect tracking system proved that zero paying customers came from that LinkedIn post? Are you changing something then?
You see where I’m going…
For me, the answer could be zero or fifty and my behavior wouldn’t change one bit. I’m going to keep posting how-to videos about my product not because I can prove they directly lead to signups (or don’t), but because of the positive, high-level, “vanity” metrics that professional analysts so often decry.
I look at how a LinkedIn post did almost exclusively from that screenshot above, i.e.:
- Comment quantity & sentiment
- Views / Impressions
If those look good, I’m gonna keep up the activity with some regularity. If they look poor, I’m still gonna keep it up (maybe I’ll try to improve). If they go absolutely wild, I might do more, and if the activity is zilch for dozens of posts in a row, I may eventually give up.
My marketing actions (and those of our tiny team) aren’t driven by provable attribution, but by a belief that conversions journeys are long, complex, and mostly unworthy of measurement investment.
The Case for “Trust Your Gut” Investments
If you work in digital marketing, you probably have clients, a team, or a boss to whom you need to prove the value of your investments. Ironically, if you can convince them to let you go after channels and tactics that can’t *prove* directly-attributable ROI (or even software measures like time-series “lift”), you’ll probably outperform your “every conversion is attributed” counterparts.
- The effort to create (and then maintain) those high-fidelity tracking systems will cost 10X+ more than the value generated from them. Do you really want to keep up with every minor URL change LinkedIn makes? Every cookie-tracking change Apple makes? Every browser-fingerprinting tactic’s shifting legal and technical modification?
- Non-attributable sources of marketing are almost always lower in competition, and more powerful in impact (because others don’t do them, your potential customers are paying more attention in those places, so your messages have greater reach and resonance)
- Channels with “provable” ROI are almost always claiming credit for channels whose attribution can’t be shown. Branded search in Google, and brand advertising are the biggest culprits, but affiliate links, unbranded search, performance ads, and paid social certainly contribute, too. Marketers who flip the script and shut down those “provable” channels often see an attribution shift, but no loss in conversions.
- Most of the time, you’re not going to do anything different with the information gleaned by these systems. That’s because management is either A) so desperate to show growth they’ll put dollars and people into channels they can’t prove or B) so cost-conscious they’ll only invest in channels with built-in tracking (AKA paid media). Rarely have I found an organization whose marketing investments were truly driven by their attribution modeling systems.
I get that it’s scary to say “I don’t know where our conversions are coming from; I only know that our numbers are going up.” In many corporate environments, that statement is wholly intolerable vs. the more common, “I know that our conversions are down this quarter, and I can show which channels are suffering, and form plausible explanations for why I’ll need more money to make them go back up.“
The Only Catch: You Need a Trustworthy Gut
Saying “I don’t know where we get conversions,” only works if you:
- Have a list of channels and tactics that seem to be working, and that you continually invest in
- Have a secondary list of experimental channels/tactics where you’re investing until you prove a hypothesis (either through “vanity” metrics or time series lift in traffic/conversions)
- Fundamentally understand your customers’ top of funnel behavior (what they read, watch, consume, research, etc. when learning about the problems your product solves)
- And, also, understand their conversion funnel behavior (why they buy your product, which problem(s) they have at purchase time, how they expect it will solve their problem)
- Keep track of enough analytics to be able to spot a disaster, e.g. if conversions fall by 50% in a month, can you reverse engineer whether one channel or another was responsible vs. if market softness, competition, or a technical issue is to blame
Without those, you’re probably too “in the dark” to make reasonable guesses about which channels and tactics to try, what to stick with, which ones are working vs. not.
My broader point is surely controversial. It suggests the work of hundreds of thousands of thoughtful, high quality professionals with immense skill in mathematics, statistics, software solutions, and attribution modeling isn’t worth their time or salary. To make such a suggestion is to imply that executives are so addicted to provable metrics that they’d rather lose money to get more certainty.
And, I am saying that.
But, the devil’s in the details. Because my advice doesn’t apply universally to businesses of all sizes. If you’re at Coca-Cola, it’s probably worthwhile to spend $100M in software and salaries for a slightly better social media brand lift forecasting system, even if that system will be largely useless in 2 years. Enterprise marketers should probably read every sentence Avinash Kaushik writes and adopt the complex, challenging-to-implement, but impressive frameworks he offers for experimentation and testing.
For the 99% of readers whose businesses are <$50M in annual revenue, I don’t think you’ll get a positive return on those investments. I believe you can get better results by trusting your gut and doing the work vs. attempting to measure it all with mathematical precision. I believe you can lift sales by choosing channels that are impossible to prove. I believe that free, vanity metrics are often good enough. And, I believe that most of the best marketing (especially for small and medium businesses) comes from untraceable sources like word of mouth, niche sources of influence, dark social, and complex multi-channel journeys.
Unscientific though it sounds, at smaller revenue numbers, gut instincts (when honed over time) will probably outperform any attribution system you can design.
p.s. For those seeking examples (outside SparkToro itself), let me suggest this talk on Marketing at Balsamiq, this thread by Tim Soulo of Ahrefs (partially screencap’d above), and Jerry Z. Muller’s anecdote-filled book, The Tyranny of Metrics.