I’ve had the term ‘data-driven’ in my LinkedIn profile for 10 years – and I’m dreaming of a world without cookie data.
I know more than most about how data gets used and abused in marketing – I came into the industry from an MSc in Economics, and have since been responsible for: $50m+ spend on Google and Facebook, founding a 50-person growth agency, teaching people data-driven marketing techniques... and I even learned to code so I could better work with big data.
Now that the marketing world is reeling from the impact of Apple’s privacy updates in iOS 14, and the wider privacy backlash from consumers, I have been thinking a lot about how my job would change if all user data went away.
Would it bring about the marketing equivalent of the dark ages? Is this the death of marketing? Will we all lose our jobs? No, that's unlikely… advertising has been consistently 1% of GDP since the 1920s, so it couldn’t get that bad. How bad would it be? Does it have to be bad? Could it be… better?
Maybe we could just do whatever we wanted without data to tell us otherwise? Perhaps we’d go back to the Golden Age of Advertising and 3-Martini lunches? Yeah right… we’re Math Men (and Women) now, not Mad Men: to make decisions we need data, and in this day and age, that means explicit informed consent.
Only using data your users have explicitly consented to is apparently called ‘Zero-Party Data’. As opposed to First-Party Data (assumed consent), Second-Party Data (from a partner), and Third-Party (traded online). I heard it first from Matt Bahr who runs a Post Purchase Survey Tool on Shopify, but I’m now seeing it pop up everywhere as consultants / thought leaders see an opportunity to cash in.
As an operator in this space, I thought it’d be more fun to do a thought experiment: if I was put in charge of growth for an Ecommerce company, how would I run marketing without user data? On day one of the new job, the UN declares all user-level tracking is banned, under the Universal Data Protection Regulation (UDPR). Only aggregated data (minimum cohort size: 30 people) is allowed, or you get fined 10% of revenue. How would I still do my job?
This would be my playbook in the first 30 days:
I’d rip out Google Analytics, the Facebook Pixel and every other unnecessary tracker. My dev team would thank me for shaving a few milliseconds off the page load speed, and conversion rate would go up a few percentage points in low bandwidth areas. I install a Privacy Friendly Analytics solution (or hack GA) that just counts Aggregated Data like pageviews and conversion events alongside meta-data on marketing channels.
We’d need to hack together a solution to track what ad campaigns drove conversions, so I’d grab UTM parameters from the URL and store it in Local Storage (yes it’s like cookies, but there’s no Personally Identifiable Information (PID): just what channel they came from). When a user adds to cart or purchases I’d fire an event, passing the utm source, medium and campaign.
Then the real work begins! The first thing I’d do is start sending out post-purchase User Surveys to the ‘Thank You’ page and the confirmation email (a purchase is consent!), that asks 3 questions:
- How did you hear about us?
- How would you feel if you could no longer use our product?
- How likely are you to recommend us on a scale from 0 to 10?
I’d set the responses to come into a shared inbox or slack channel, and make sure they know they can hit reply to contact me directly (not one of those ‘no-reply@’ addresses. Anyone that responded negatively I’d auto-escalate to customer support, and those that responded positively would get (optional) follow up questions so I could learn more about them.
Then I’d run a few SQL queries to do an RFM Analysis, getting a list of customers by Recency (when did they last purchase), Frequency (how often did they purchase) and Monetary value (how much did they spend). I’d shoot off an email to the top 100 customers across all three dimensions introducing myself and offering a 15 minute call if they would mind giving feedback.
I have started my mission of actually Talking to Customers by taking calls and responding to emails. As I do so, I take note of the words they use to describe their problems, objections and how our products make them feel. I then start using our customers’ words wherever possible in the adcopy of our ads, on the blog and website, and in our emails. This will Optimize Conversion Rate sitewide, which needs to happen to make up for the loss of personalized targeting.
At the end of each customer conversation I ask if I could send them a small gift to say thanks, and get their address. That goes into an Airtable which pings an API to automatically mail them some company SWAG and 3 x 20% off Discount Vouchers: 1 for them and 2 for their friends.
Now I’ve talked to a couple of customers, and have seen the early post-purchase surveys, we can start thinking about our ad campaigns. It’s been a week since the UN banned user-level tracking, so let’s turn off any Non Converting Campaigns and cut spending for any that aren’t under our CAC target (as reported by our hacked together solution).
Now let’s think about Contextual Targeting: based on what we know from our customer calls, where do similar people hang out online (or offline)? Are there ways we can get in front of them? Find creative ways to advertise anywhere that over-indexes on your audience. For example if you hear the name of a blog 3 times just cold email or tweet at the blog owner and pay them $500 to sponsor the blog: no tracking or affiliate links, just a simple banner and a few naturally placed mentions.
At this point I’d also start Deprivation Testing. Take something that nobody very senior is passionate about, that costs us a lot of money, but might have questionable incrementality, and just… turn it off. See if anyone complains or if the numbers dip. If you’re risk-averse try just turning it off in one country or region. Good candidates for this type of test are Brand Keyword Ads, or Retargeting Ads or a Voucher code affiliate partner: chances are you get an easy win!
The honeymoon period is probably over by now. At this point I’d expect to have had at least a few negative customer calls and the cuts to advertising would have probably slowed top line growth. I’d still have some breathing room because of how unprecedented the UN’s legislation was, and the CEO’s drinking buddies would be confirming that marketing performance is down across the board, but that’s no excuse not to get things moving up and to the right.
Now to really move the needle with our secret weapon: Creative Strategy. I’d have a good enough grounding in why people buy from my interviews and surveys to form some hypotheses about how we can get them to buy more. I’d write these up in the GET / WHO / TO / BY framework, aligned to the different user personas I’ve noticed in my research. There isn’t time to commission an agency or team of freelancers, so I’d go it alone with Canva, Copy.ai and a bottle of Rum (write drunk, edit sober).
I’d turn off all existing ads except one: the current best performer. Then I’d replace the old ads I turned off with the best new ad concepts I came up with. I’d go for bigger bolder concept level changes, rather than smaller variations, and I’d just stick them in the same ad sets rather than attempting a split test, just for speed – we’re looking for big wins here not scientific proof.
Finally I’d feel ready to attempt a win back campaign. I’d look at any customers from my RFM analysis that had high monetary value and frequency, but low recently: i.e. they used to be a good customer but now they’re slipping away. They’ll be especially unlikely to respond, so you’d have to try something big: offer an item completely free (just pay for shipping), insider access to test a new product line, or a chance to win a generous prize (make it something only your ideal persona would want, nothing generic like an iPad).
This is where everything comes together. You have the infrastructure you need: you’ve got aggregate campaign level data coming from your hacked together analytics setup. You have rich survey data to calibrate the attribution from analytics, and you continue to personally take customer calls to keep your ear to the ground and train your gut intuition on what they want.
I’d do one more round of ad optimization: turn off or cut your worst campaigns, brainstorm new contextual placements to replace them and scale budgets on the channels that are working. Cut your worst performing ad concepts, and bring in freelancers to create better variants of your best performers, while adding new concepts and themes into the mix.
Finally I’d do some rudimentary Marketing Mix Modeling. In about 4-8 hours you can build a basic Econometric model in GSheets that will backup and calibrate your existing insights from customer interviews, privacy-friendly tracking and deprivation testing. Combining these methods you could triangulate a single truth about what’s working (or not) in your marketing, and elevate your decision-making above guesswork (all without using any non-consensual user data!). All of this would go in a nice presentation for the boss – fingers crossed!
Assuming I kept the job (for now, marketers have the shortest tenure), this is what I’d do to consolidate my wins.
- Talking to Customers: It sometimes feels like a distraction, but there’s nothing better for building intuition on what will work in marketing: quantitative is precise, but qualitative is fast, and it’s speed you need when you’re making most of your decisions.
- Aggregated Data: This is the new normal: don’t expect to be able to drill into a user profile until they’ve given permission. The truth is most user-level data sits in data warehouses and never gets looked at: you aren’t missing much.
- User Surveys: People lie on surveys, but their cookie data lies too, just in different ways (seriously check out some of the things Google ‘knows’ about you). The right questions properly calibrated with other techniques, can be remarkably accurate for attribution.
- Contextual Targeting: To be honest, the best ads were always contextual and retargeting was never that incremental. Google’s not worried about cookies going away, because it doesn’t take a cookie to serve printer ads to someone searching for printers.
- Deprivation Testing: For the really big decisions, there’s no substitute for A/B testing. For channels where you can’t easily A/B test, find ways to turn it off for a randomly assigned group of people: make a change on 50% of your URLs or just in one country.
- Creative Strategy: you can never get enough of this. The best creative ideas work on any medium and can pay dividends for decades. “Never stop testing and your advertising will never stop improving” - no it wasn’t a growth hacker that said that, it was David Ogily back in the 60s (pre-internet, for those paying attention)
- Marketing Mix Modeling: I’d work with a vendor or hire an internal team to do MMM properly, using Python code or R to make it more scalable. This would be the primary source of truth for budget allocation once per quarter, though I’d do small weekly or biweekly updates to track the accuracy of predicted vs actuals.
I hope that was a useful exercise: now back to reality. You should probably do all of that stuff even if you don’t expect tracking to go away anytime soon! It’s just good marketing. The fundamentals. Basic stuff. It’s actually surprisingly easy to think of ways to improve marketing effectiveness without invading user privacy. In fact, when you think about it, how often do you even use all that data? I’m not worried about a post-data world: in fact I think it might even lead to better outcomes!
Oh and if you’re reading this and thinking “we should just hire Mike to do this for us”, I’m working full time on my own business after leaving my other business and I really promised myself I wouldn’t take another job. So unless you’re offering some ludicrous amount of money, or are working to make the human race interplanetary, offer the opportunity to some ambitious young intern instead, and hand them this guide.