Attribution obsession is ruining your strategy
The real cost of overly-engineered attribution, ironically, is the very thing it’s designed to solve. The solution is actually pretty simple.
If you joined a startup thinking it’d be all ping-pong tables and free lunches, only to find that the VC money had dried up, the ping-pong table got sold, and what’s left was the obsessive data-driven neurosis that every action must be measured and quantified, you’ve probably felt the pain of attribution.
Suddenly you “need” a watertight, multi‑level, CAC‑optimised, board‑ready, AI‑native, cross‑functionally aligned attribution model, and your entire marketing strategy has been handed over to some analyst who hasn’t spoken to another person outside work in a year and thinks marketing is a cost centre.
All of this is made worse by the fact that it doesn’t paint the picture you intuitively know to be true as a talented and, frankly, attractive marketer who is great at parties and speaks to people outside work all the time.
The promised land of attribution was a seductive idea. A perfect dashboard that tells you exactly where to spend your money?
Just wire it all up, Tom, and I’ll take it from here. Shhhh, shhh. There you go.
The real cost of overly-engineered attribution, ironically, is the very thing it’s designed to solve.
The problem with attribution
Attribution is the flattening of a marketing strategy. It takes every micro‑interaction, trust‑building moment, incentive and message and squashes it all down into one thing that gets credited for your sale.
Any attribution method does this to some degree. You can’t avoid flattening completely. But there is a way to do it without wrecking your strategy.
Brands that obsess over perfect attribution end up bending their marketing to fit the model, rather than building a model that gives them enough direction to guide what they already know about how their customers actually buy.
In that world, last click actions will win every time. And while these clicks are useful, they shouldn’t be driving your strategy.
When you focus all your energy on what brought someone to your website, you slowly cut out all the things that help to control acquisition cost, capture mindshare, build trust and drive longer‑term growth. Without those, you have to serve more ads, more often, to convince someone to try your brand. That gets expensive.
Strategy comes from user journeys
Real journeys are messy. Someone hears about you on a podcast. A week later they see a LinkedIn post from a friend. They Google your brand and click an ad because it’s at the top. Two days later they come back via a referral link. A month later, they finally sign up on a laptop they found under a bridge.
When you map a journey properly, you care about where people first hear about you, what finally gets them to the site, what they see on that first visit, and what has to happen before they’re ready to buy. You care about the stages and the jobs: who we’re trying to reach, what they believe now, what has to change, and which moments actually move them forward.
The job of strategy is deciding which parts of that journey you’re going to invest in, and how.
User journeys and research should dictate where you want demand to come from and which bottlenecks you’re trying to relieve. Sensible attribution can nudge you towards or away from those bets, but it shouldn’t get to define them.
Clicks are for operational credit
Click data sits underneath strategy at a tactical level. Shoutout Bitly.
At Yonder we used Bitly as our analytics hub for every campaign – the very first referral program I built at Yonder was built on Bitly. It’s how we managed affiliate placements and new creator partnerships. We used QR codes on printed handouts, or for one‑off landing pages. It gave us one clean view of how those individual assets behaved. But importantly, they didn’t tell us whether to run the campaign or not.
Clicks are tactical. If we’d promised an affiliate they’d get paid per click or per signup from their link, Bitly made it easy to see whether we owed them money. Clicks give operational credit - did this specific partner, link, or asset do what we agreed?
Where teams get into trouble is when they allow that same click data to give strategy credit. Not good. They look at the last link someone clicked and treat it as proof that “this channel created the demand and converted customer”. Which is extremely unlikely.
They move budget based on whichever URL happened to be at the end of a very messy chain. It’s how marketing budgets become skewed towards Paid Social, Paid Search and affiliates – because they’re often the final triggering action or step before a conversion.
Use clicks for operational credit and for debugging your journeys – are people seeing and using the things you’re putting into the world?
How Did You Hear About Us (HDYHAU) is still the best directional attribution
The humble “How did you hear about us?” is absolutely goated.
It’s simple, quick to implement, and still one of the best ways to get a directional sense of where customers are coming from. You’re asking the only person who was actually there.
Granola, WisprFlow and loads of other companies still do this in their onboarding flows and they’ve got millions of customers. You are not too big, too sophisticated or too complex to ask.
HDYHAU still flattens the story. People can misremember. They anchor on the most salient touch, not the first one. There will be noise. No attribution model is perfect.
The difference is how. Click‑based attribution flattens the story to whatever got tagged last. HDYHAU does it around the customer’s own memory – what they believe was important in discovering you. It’s more likely to surface podcasts, communities, friends, and content – the things your dashboards underweight – and less likely to reward whoever happened to be at the end of the chain.
It’s easy to implement. On your checkout or sign-up flow do this:
First, you ask about the medium in broad strokes:
“Where did you first hear about us?”
Social. Podcast. Newsletter or email. Search. An event. A recommendation from a friend or colleague. Website content. An integration or another product.
Then, only once they’ve picked the medium, you ask the specific follow‑up:
“Which one?” or “What website?” etc.
If they say social, you ask which platform. If they say podcast, you show a short list of your main shows plus an “other / don’t remember” option. If they say website content, you name a handful of flagship pieces they might recognise. You’ve separated the type of touch from the specific surface. Basically mapping to Source and Medium from typical UTMs.
This sits alongside your understanding of the journey and either reinforces it (“yes, people do seem to discover us through creators and one specific guide”) or challenges it.
There is no perfect end to this story. HDYHAU is still a best guess. You need to let go of the fantasy that any attribution view – even a good HDYHAU setup – can fully explain why growth is happening. At best, it nudges you towards or away from the bets you’ve already made about your customers and your category.
Keep CAC blended and optimise channels on inputs
Most of the pressure to over‑index on attribution is really pressure about CAC.
If you can’t say “this channel brings in customers at £X”, how do you argue for budget? A best approach is to just treat CAC as a blended goal for your whole marketing strategy instead.
Every month, add up your total sales and marketing spend and divide by the number of new customers (or qualified opportunities) you created. That’s your Blended CAC. Not “paid CAC”, “content CAC” or “field marketing CAC”. Just the cost, on average, of acquiring a customer given the system you’re running right now.
I’m not saying you should never look at channel-level CAC. But just be careful not to treat that channel as if it did all the work. Every other touchpoint subsidised that conversion. Invest in brand and content, and your paid CAC drops. Strip it back, and paid gets more expensive.
Give each activity a clear job in the journey, and judge it on whether it’s doing that job well. A paid social campaign whose job is to introduce the brand to new people should live or die on whether people actually stop scrolling and pay attention. Long‑form content is about engagement and movement into product pages.
This is where Bitly comes in so you can standardise how you track links into your site and see whether people are interacting with the things you publish. Then you can just build a dashboard in their tool and track it all with their Weekly Insights or just chat with their AI-chat Bitly Assist to find out what’s working. You’re using clicks to keep yourself honest on inputs, not as the sole measure of output.
You don’t need a detailed attribution model
You don’t need a Marketing Mix Model. You definitely don’t need to be debating multi-touch vs data-driven vs time decay in a company that’s still trying to get from 200 to 2,000 customers.
Keep it simple. User research sets strategy. HDYHAU tells you where demand is coming from. Blended CAC tells you whether the system is getting better or worse.
Do that consistently and you’ll be much less likely to wake up six months from now staring at a rising CAC chart, wondering why “doing what the data told us” somehow made everything worse.
While I’ve got you…
My name is Tom. I’ve launched and grown products at some of the UK’s most loved consumer brands like Monzo and Wise and was part of the founding team and VP Marketing at Yonder.
If you’re a senior marketer at a startup, this Substack is for you. I write about what actually works in startup marketing (and what definitely doesn’t) for marketers on the verge of breakdown.
Take my short reader survey so I can learn more about who reads this so I can write more useful stuff. I know you ask your customers to take surveys, so one won’t hurt you. Thanks in advance.
Here are some of my most popular reads:
I would add to HDYHAU... why did you take the call today? Because the how is one side/the why is whole other one. If you could tap into both, that's where you can get more deals to come in! Also this is the thing I'm most passionate about-- and one of the reasons why I'm building upside.tech
Hey Tom. I have been following you on Substack.
I am a student and I am researching on substack. Would you be interested to spare 10 min of your time, to improve substack?
https://substack.com/profile/498011815-sriharika-nallagorla/note/c-269909811?r=88i4dj&utm_medium=ios&utm_source=notes-share-action