Search and Performance Insider Summit, Deer Valley, UT, Dec. 13, 2019 - 01:00 PM: Moving From Reporting to OptimizationGet more value out of your technology investments. Move your reporting from backward facing to insight generating.Presenter: AdamReitelbach, IndustryExpert,MerkleAdam is a long-time performance marketing veteran. He works with clients to define goals, build world class solutions, and exceed expectations. He has significant experience in the retail, travel, and B2B verticals.
Who is Merkle and what are your core services categories? As a data driven CRM agency, we set out to build a digital agency 10 years ago that brought an unmet market need to the agency business – transparency, data driven targeting, and closed loop measurement to create a performance media agency, not just a digital marketing agency.
-spectrum / maturity graph of entry-level single-channel optimization ->> integrated, multi-moment, connected customer journey on the right
-everything in between
-everyone wants to be on the right
-day in, we’re stuck on the left in optimizing channels
-Pop quiz – where are you? Show of hands?
-How to achieve that?
In order to advance Marketing maturity and capabilities, Marketer’s must also adopt and integrate products.
-Adobe, others could be used (and Merkle does!), but showing Google because of low barriers to entry and everyone buys ads on Google
-Left is single channel, right is cross-channel automated
-Looking at Google stack, you can visualize the connection points
-Every-man’s approach that Google makes readily available
-We estimate that >90% are between nascent-emerging (google data)
## you’re not behind
## you have the opportunity to advance
-Multi-moment is realtime.
##In-store data syncs with website with email, etc. Customer journey execution is at a 1:1 basis. (Verizon is an example)
-Need a vision to get there, with less budget than the big players
When we ran the model on our dataset we learnt some interesting trends
What we discovered was that Brochure, Test Drive were low on the list, but what we did see was things like The Number of Sessions, visits to the Lifestyle, finance and locate a dealer page were all key indicators that someone was likely to purchase a vehicle
One of our initial challenges was solving the problem of tying online data to vehicles sales data
One of the challenges that automotive brands face is that traditionally everything that happened online was very separate to what actually happened in a dealership – there was no way for brands to associate the research took place and how that directly influenced the sale of a vehicle on a 1 to 1 basis.
Through online Analytics tools we could analyse a consumers behaviour, we could understand what they do onsite, which channel drove them to the site, when they downloaded a brochure or requested a test drive but this is where it stopped.
Whilst onboarding MINI onto GA360, through our discovery phase of MET-A we identified that MINI send consumers a series of emails after they have purchased a vehicle. By appending a few parameters to links in these emails and with some advanced configuration in GTM and GA we are able to identify consumers who have purchased a vehicle. Once we can get the consumer back on site, ID’s are linked and we understand all the research the customer did prior to entering the dealership and we now also now know exactly which vehicle the customer purchased.
Using GA 360 and the full configurations all this rich data is streamed into Google BigQuery on a daily basis allowing for deeper analysis and discovery of insights.
This chart is called a ‘Confusion Matrix’, it tells us how accurate our model is. It’s divided up into 4 quadrants, those that haven’t yet purchased vs those that have purchased and what our model predicted vs what actually happened.
What we can see is that of those that we predicted would purchase a vehicle, in 90% of cases we were actually correct.
At the time of running the model, we could also see that 70% of the time we were correct in identifying customers who weren’t likely to purchase a car.
So once we understand the likelihood of a consumer purchasing a vehicle we can use this probability, load directly back into Google Analytics 360 for retargeting. We can build audiences of customers with a high, medium or low probability to purchase. We can define strategies, what do we do if we know someone is likely to buy a car? How do we find more of these people. Do we decrease media spend on the customers that we know have a low probability to purchase?
Can we bid more aggressively across generic terms for customers who look like those that are have already purchased?
3 months after loading our data into GA we could then analyse the results. What we found was that of the customers that we predicted had a high probability to purchase, these customers actually had a conversion rate 3.5X than those with ‘medium’. Those with ‘medium’ probability purchased at a rate of 8X more than those with low.
Seeing the results after a period of time across real data was a way for us to accurately validate that our model was working
This presents huge opportunity for us to now embed these learnings and insights into our marketing strategies going forward.