It’s been a while. Partly because my last post from February was a bit of a tough act to follow, partly because I’ve been caught up with a bunch of work projects. The company was acquired by LVMH, I spoke about e-commerce economics at SMU and NUS and got so pale from a lack of sunlight that taxi drivers started assuming I’m Russian. Today it’s going to be short and sweet and a little bit different: My thoughts on performance marketing and the pitfalls of spending large amounts of money on so-called data-driven marketing channels. Let’s go:
1. The attribution lie
All the way up until 2013-2014 it was generally agreed upon out here in Southeast Asia that the correct way to optimize a digital marketing mix was to look at the channel revenue attribution list in Google Analytics and distribute spend from under-performing channels to over-performing channels.
Then we all realized that this probably doesn’t make a lot of sense when some channels are much more likely to be at the end of the user purchase funnel than others (cue Mr. Kaushik), and we then went to town on creating our own cool attribution models inside Google Analytics and look at assisted revenue instead of Satan’s own offspring, last-click attribution.
So far so good, but then we started realizing that user journeys probably aren’t confined to one logged-in device, and that’s when it started to become a bit hairy. Google went all crazy with their Universal Analytics and User-ID’s that didn’t really work, so we all started creating our own expensive data warehouses instead.
And where are we now? I have yet to see a company fully make sense of this humongous pile of data and report on marketing in a way that fully encompasses the new user-centric lifetime value profit view. Facts are that GDN still gets way more scrutiny than AdWords and that Facebook Ads typically gets neglected even though it tends to outperform almost any other paid channel out there.
What to do? It’s time to put our money where our mouths are and report profitability on a user-by-user level. If we know that a typical user only becomes profitable after x months or x years, why don’t we value our companies on that basis instead of looking broadly at all users in aggregate. Perhaps we could even stretch ourselves to making “lifetime profit per user” the new undisputed KPI?
2. The optimization lie
This is probably my biggest current issue with the performance marketing religion: It tends to favor the safe over the experimenting and adventurous. The job of a digital CMO should be to hire channel managers for all meaningful marketing channels and give them a simple monthly target on budget growth and marketing efficiency growth. Instead, it’s become a lot more about providing detailed revenue estimates to CFO’s and CEO’s. Why’s this a problem?
Well, the problem lies in the fact that while the job has turned into that of a traditional large-company CMO, the marketing and growth goals are completely different. If a large company grows 10-15% per year it’s ahead of the pack and everyone is happy, but if most growth companies grow less than 100-200% it’s bad news. Hence, to properly work, we must make substantially different marketing decisions.
In short, we need to fully embrace chaos. We talk a lot about A/B testing and being data-driven, but we forget that when the experiments are highly controlled in an environment that is made for testing (Facebook Ads and AdWords come to mind), the optimization pay-offs are also going to be controlled. That’s all fine and dandy if we’re happy with 10-15%, but to maximize our marketing potential we need to devote much more marketing budget to new channels and completely new experiments on existing channels.
To take AdWords as an example, it’s a channel that from the outside lives up to the requirement of supporting a data-driven mindset AND the screwed-up requirement of being easy to estimate future revenue for. Hence, we all move more and more of our budget to AdWords, driving up the prices for everyone and getting to a point where what we’re really paying for is the insurance premium of being able to forecast and extrapolate future earnings.
I dream at night about chaos-loving online CMO’s who understand that if they just diversify their crazy marketing experiments across enough markets and audiences, they can worry about evaluating and shutting down/boosting entire channels instead of 5-10%-improvement-potential AdWords ad text split-tests. They’d understand that oscillations are inherently good as long as the experiments are spread out enough, and that that’s the only way to beat market gains over time. New Market ETF’s over High-yield legacy stocks, yo.
3. The big data lie
I see it constantly. Clients believe that the way to further improve their marketing efficiency is to use more data. They sign up for enough analytics solutions to make their load speeds intolerable and build data warehouses large enough to store a mid-sized water jet. What this typically ends up with is more analysis and less execution in a landscape where there’s typically lots of smart decisions to be taken without even glancing at the data.
As mentioned above, my ideal marketing structure is one where channel managers have free power to experiment over a cohort much longer than a week – ideally monthly – and are incentivized purely on slowly growing budgets (4-8% per month would be great) while retaining current efficiency levels or maybe even improving them a bit (2-4% should cut it).
What I hope you’d end up with is a place where the chaos is embraced and exploited in the campaigns instead of through the communications and reporting frameworks within the marketing team. Someone once told me that the difference between a good restaurant and a great one is the noise level in the kitchen. I want the same kind of short, structured and clear communications on my email.
So that’s that. I promise it won’t be 4 months till next time.