InfoTrust book notes (Crawl, Walk, Run)

[r002]

Sunday

Ch 7: GCP for Marketing

Pam Castricone (Head of Data Science) & Tyler Blatt (Head of Integrations)

• Pareto Negative Binomial Distribution model
  • Non-contractual transactions which can occur at any time
• Different models
  • CLV modeling
  • "Larger, comprehensive, customer-centric strategy"
  • Propensity modeling
  • Has narrower focus-- improve ROI of remarketing campaigns
  • Much shorter time frame: Normally, 3/7/14/30 days.
  • Eg. Logistic regression
  • Forecasting
  • Eg. Holt-Winters model (auto-regressive integrated moving average; ARIMA)
  • Setting up forecasting with GCP
    1. Data collection process. Eg. Schedule GCFs to gather data at periodic intervals to store somewhere
    2. Pass the data through a model that generates the forecasts. Eg. Use GCFs again (assuming processing <10min)
    3. Visualize the forecasted data; make easily shareable within the org. Eg. With GDS
       

References

• https://mitx-micromasters.zendesk.com/hc/en-us/articles/360035902191-How-much-does-the-MITx-MicroMasters-program-credential-in-SDS-cost-and-is-there-financial-aid-available-

[r002]

Ch 8: Digital Analytics Maturity: News & Media Orgs

Andy Gibson (Head of Vertical: News & Media)

Nascent Publishers - Lack of first-party data

• Advertising revenue still largely based on impressions and clicks
  • High-level KPIs like page views, sessions, users
• Audiences are largely anonymous
• Social media arm is largely divorced from CMS/Analytics arm
• Finally, these publishers often operate in different markets (eg. local TV stations)

Emerging Publishers - Better Personalized Experiences

• Begins with an audit
• Due to privacy regulations, need consent though from consumers after they login to collect data on them

Connected Publishers - Unified first-party data

• Break down silos to integrate all of the various, disparate platforms (eg. GA, social media, offline-media) together to form a single CLV.
• After CLV is created, a point person then needs to actually communicate with the different teams to properly leverage this data
  • Eg. Editorial, research, social media, ad monetization, email marketing, etc

Multi-Moment Publishers - Activating unified customer data

• Analytics 360 breaks down your audience into cohorts (eg. age, gender, geography, custom user data)
• Ad Manager can then directly use those cohorts to granularly serve targeted ads
• Can also use A/B testing to see what personalized experiences works best to deliver conversions
  • Google Optimize 360

Mobile & Over-the-top platforms (OTT)

• Analytics 360 can easily track mobile scroll, which part of video people watched until, scrolling/reading speed, etc

Where to start?

• Use first 60-90 days internally examining your company. What are all of the different teams collecting/using customer data? Audit them to see what data is/isn't being collected
• Use Tag Inspector to consolidate/remove extraneous tags-- improve loading times/performance!

Measuring success

• Avoid super-high-level metrics like "traffic"
• Battle between aggregators like Google/Apple News and providers like NYT/WSJ
  

Challenges moving forward

• Apple's Intelligent Tracking Prevention (ITP) - limits cookies
• Apple/Google News example: Right now there is a very real tension with the aggregators vs the news providers. Aggregators that deliver the last mile get all of that valuable customer data. But news providers, unless you're the NYT or WSJ mostly don't have the muscle to build entire platforms. And even if they do, they're only getting a fraction of the audience.

[r002]

Ch 9: Digital Analytics Maturity: Consumer Packaged Goods

Chris Vaughan

Terms

• Consumer Packaged Goods (CPG)
• Fast-Moving Consumer Goods (FMCG)

The Challenge

• 90% of CPG are still sold offline at retailers like Wal-Mart, Kroger, Target
• Currently, CPG websites that do exist, do three things, none of which generates revenue:
  • Inform customers a new upcoming brand/product
  • Promote a current brand/product campaign
  • Generate customer loyalty/engagement
• These websites currently only exist because "everyone else has them"-- they get minimal attention/investment
• Currently huge international orgs basically just need to "hope that one of their individual markets/brands won't inadvertently track something on a website that triggers a huge privacy fine." 😯

A specific example

Let's make this real. I was thinking through this chapter as I read it and here's a real-life example in the CPG space that, amazingly, in 2021 still actually hasn't been solved.

There's a specific kind of soy sauce that Shaza loves. It's called the Mizkan Ajipon. Whenever we go to Cam's, if there are any available, we'll get an entire armful! (It goes great with white rice and soup dumplings!) But like Vaughan describes, if you visit Mizkan's website, it'll look like this the screen snap below.

Observe: The website is just informational with a basic description and some nutrition info. You can't actually buy the sauce from the Mizkan website itself. (There's a separate "Where to buy" button that then lists vendors and resellers.)

If you are Mizkan in this example, chances are you've at least deployed a single enterprise-level instance of GA on this product page. But high-level metrics like bounce rate is now misleading/meaningless/useless. If people visit this page and immediately leave in droves, what does that tell you? Nothing! Maybe they hate the product. OR. Maybe they're off to amazon.com to buy it! Who knows? 🤷‍♂️ This is a challenge for nascent players in this space.

In fact, Vaughan goes to point out: Many large orgs who do already have some basic GA instance and a centralized digital analytics strategy/core team set up are actually hugely resistant to change. He called them "thought vacuums"-- because the executives are siloed, all of their insights are local and only part-of-the-elephant. I'm reminded here of the old saw:

"Wrong/incomplete info is worse than no info!"

since you're often grasping only part of the picture and very possibly deriving incorrect, misguided, or even harmful conclusions for next steps/strategy/tactics. 🤦‍♂️

Also, tbf: There are hundreds of these kind of product pages globally. Updating websites on this scale, and then testing them to see if the new campaign yielded results, is simply logistically tough. Many potential confounders too.

Thus, currently, many folks rely on gut feeling and even when there are analytics delivered, they're often not actionable because they're lacking context.

Emerging Giants

• Marketing needs to work with Analytics
• Local Analytics Champions (LAC) in local markets. No long a single monolithic, top-down approach
• This is the point where orgs move from Free GA to Enterprise GA
• They also begin actually customizing their GA reports with event tracking
  • Tradeoff: Obviously, requires significant development work
• Still possibly piecemeal in their implement. Eg. Using GA 360 but not using Display & Video 360 and instead using some other proprietary programmatic solution. Thus, still missing holistic CLV.
  • Ie. "Glaring holes" in the standardization of data

Connected Giants

• Everyone trusts the data to use it to make decisions
  • Importantly, even non-analytics people trust the data to make business decisions
  • Data is open and available to everyone
  • What KPIs are important where in what contexts are clear
• LACs drive the campaigns
• Most importantly-- customer is happy and consistently give good qualitative feedback on surveys
• System and all platforms are completely integrated and interconnected
  • It's easy/fast to try new ideas, implement them, get feedback, and iterate quickly
  • A good example is when a software project has good tests! You can move quickly without fear!
  • Accidents and breakages (like accidentally leaking private info) can be fixed fast!
• Privacy is paramount at this level now so much data is being collected
  • InfoTrust Analytics Monitoring Solution "detects irregularities and discrepanciesas well asPII`

Multi-Moment Success

• The holy land: Data Science enters the picture
  • "Either an internal or third-party team within the org who specializes in developing advanced insights".
  • Different forecasting techniques. Eg. CLV, propensity, audience scoring, sales forecasting
  • Audience scoring - looks at past user website behavior to predict which are likely to perform certain future actions
• Payoff at this stage is huge: You've essentially "hacked" the sales funnel!
  

6 Ps: Purpose, People, Platforms, Processes, Pace, Payoff