I just started developing online retail for Studio Moderna at the end of 2004.
It was an amazing time. A much simpler world when it was easy being a marketing hero just using traditional direct response metrics.
We lived and died by conversion rates, cost per order and direct campaign profitability. Sure, we used other metrics as well, but these were the Holy Grail.
But then the world started changing.
What we’ve been taught about direct response measurement online “suddenly” became outdated, and is today even dangerous if used as our primary compass.
1. Reliable Direct Response Measurement is a Wet Dream
Direct marketers like to see the world as a simple funnel. And we all love the safety we get from making decisions based on the funnel.
Is a campaign generating a low conversion rate or too high cost per order?
No problem. Kill it. That’s the safe thing to do.
Why keep investing in something that evidently isn’t generating direct sales?
Unfortunately, life is not that simple, and the “funnel model” is practically dead.
The Scary New World and Google’s Zero Moment of Truth Model
As shown by Google’s ZMOT Study, the consumer path to purchase has become increasingly complex:
- In 2011 consumers used an average of 10.4 info sources before buying. And that’s twice as much as in 2010.
- Purchase research for most product categories peaks at 2 – 3 months before making the final decision.
- Pre-shopping research (zero moment of truth) now has more impact than stimulus and the in-store purchase decision (first moment of truth).
And there’s more: customers buying online do 9.2 searches on average before buying, and those buying offline do 7.6 searches (source).
It’s Not Really a Funnel …
The decision making process isn’t even close to being a funnel.
It’s actually closer to this, just incredibly more complex in real life:
We can’t really easily measure the direct impact of offline media, so let’s skip that for now.
But even just looking at the online channels we have a problem …
The Danger of Relying Only on Google Analytics Conversion Rates
Tools like Google Analytics will reliably report the “last click” source that brought visitors to your website just before their “first moment of truth” purchase decision.
The problem is, they only tell you which sources brought you customers that were at the very end of their purchase cycle.
They don’t and can’t reliably tell you what stimulated the customer to start researching the purchase and how he or she arrived to that final moment of purchase.
The result? Making decisions just based on “last click” metrics means you’re only investing in grabbing late-stage customers, but are actually ignoring early-stage potential customers who haven’t yet decided to buy from you.
And that means ignoring your greatest revenue growth potential …
2. The False Security of Online Multi-Channel Attribution Models
But wait, hasn’t this problem already been solved with online multi-channel attribution models?
Multi-channel attribution models try to give us this:
Just one problem: the current technical limitations (not to mention the upcoming privacy troubles) make reliable multi-channel attribution impossible.
They certainly tell us part of the story, and that’s a hell of a lot better than nothing. But still, only part of the story.
Can’t Measure Cross-Browser and Cross-Computer Behavior
You start researching a purchase at work, and then buy in the evening at home. Or even research in one browser, and buy through another.
The end result – Google Analytics, or any other online analytics tool, can’t measure this behavior (unless you’ve somehow connected “both” users, perhaps via login).
Imagine this scenario:
- John clicks on your ad at work, generating a cost for you. He looks at the product, does some more research, and decides to by later.
- In the evening he enters your URL directly and buys on the spot, but with no ad cost to you.
He wouldn’t have bought if he hadn’t clicked on the ad at work. And yet that ad gets 0 attribution for the sale, and the direct URL entry gets the full attribution.
Looking just at direct results you might make the decision that the ad campaign isn’t working. Of course it’s not working, if it’s not generating direct sales!
But is it really not?
Can’t Measure Cross-Device Behavior
Now add different devices into the mix.
Can’t Measure Third-Party Services
And finally, just to make things even more interesting, don’t forget that third-party services people use to research your product also have an impact.
Reading your product review on Yelp, Amazon, Zagat or other services. Seeing your posts on Facebook, but not clicking. And more.
You can’t measure the impact of any of this using traditional online analytics tools … and yet, you can be certain there is an impact.
Is It Google Analytics’ Fault?
Absolutely not. And neither is it Omniture’s, KISSmetrics’ etc.
They’re all amazing tools. It’s just how the internet works that’s the problem.
Don’t Worry. It’s OK!
We have more data than ever before. But, we don’t know everything.
Our challenge aren’t the tools. We can’t make data perfect.
Our challenge is changing our mental model.
- Don’t rely only on direct response metrics.
- Don’t judge campaigns and traffic sources purely on their direct response results.
- Don’t focus just on late purchase stage customers and traffic sources, unless you want guaranteed low growth.
- Develop new online analytics models and frameworks to measure the entire purchase cycle.
- Evangelize a different approach to online measurement within your organization.
And above all, admit to yourself that you just can’t measure everything. Plus, you can get started with the follow-up post
To drive the point home, here’s an amazing quote from Jim Novo, from a conversation we had on Avinash’s blog (BTW – this post is a must-read for understanding the challenges of multi-channel attribution):
There are simply limits on what can be “proven” given various constraints, and that’s where experience and a certain amount of gut feel based on knowledge of customer kick in.
If you can’t measure it properly, just say so. So much damage has been done in this area by creating false confidence, especially around the value of sequential attribution models where people sit around and assign gut values to the steps.
Acting on faulty models is worse than having no information at all.