latent effects modeling for fun & profit

latent effects modeling with AI/ML for fun & profit

We’ve been posting opinions and comments that urge marketers to “get over” artificial intelligence (AI). To dive into AI’s veritable treasure house of research tools and insights, instead of circling it warily, for fear of banging into  a “disruption” moment.

 So, we hereby plant the flag of AI for the regular guy (and gal, of course)  well, at least for the professional who wants to use precise and accurate data to propel insights that simply wouldn’t be perceptible with traditional approaches to analysis.

 Let’s take a geek’s-eye look around our Artificial Intelligence/machine learning workbench.  Oh, here’s a cool tool — Latent Effects Modeling.  Latent variables, as opposed to observable variables, are variables that are not directly observed but are rather inferred (through a mathematical model) from other variables that are observed (directly measured).  In short, Latent Effects Modeling takes what you can see and directly measure and infers (predicts) the variables you can’t see or measure.  For example trends in data that would otherwise be missed.

 Imagine a box top, bottom, each of four sides are datasets, all of them interacting with each other at full gallop.

 Now imagine a company in the fishing industry – Flounder Inc. In this case, one dataset is about weather, another is about Flounder’s fishing- based revenues, another about how big ships are traveling through their fishing grounds and affecting fisheries – and therefore revenues, or not … toss your own variables and push the button to start the pattern matching and factoring in the variables.

Complex? Yes, but the model crunches away … and comes up with trends and predictions that reveal the unseen trends and variables and how they may play out if the data remains the same.

Latent Effect Modeling is in play as we speak, generating some of the most widely used predictive models of the COVID-19 spread, as well as looking for trends and patterns in marketing data.

people are using it for all sorts of predictive modeling. things like:

  1.  Which of several digital campaigns is most likely to produce positive long-term memory of your brand?
  2. Which presidential candidate produces the least revulsion/disgust in voters?
  3. What range of prices is most likely to produce the greatest sales without losing you money?
  4. Which of several email messages is most likely to produce the action you wanted from consumers (e.g., try a new product)?       

disruption isn’t A Four-letter word

In his “disruptive” book, The Innovators Dilemma, Harvard Business School Professor Clayton Christensen talks about “Jobs to be done”.  The Job that most companies have been asking AI to do thus far has been to either reduce costs or measure efficiency – with the by-product being increased profit.

 While these less-advanced organizations focus their AI initiatives on cost reduction, more-advanced companies see revenue increases from AI, indicating a shift to more strategic — and customer-centric — AI deployments. It’s time to think of AI as a way to directly increase revenue.

 Two quick real-world examples of AI techniques, like LATENT EFFECTS MODELING, are making their way into the marketing world and healthcare worlds. 

Utilizing that mound of data that is created every time you launch a campaign to ultimately predict trends in brand loyalty, eCommerce transactions and lead generation. Companies are seeing increases in these types of transactions by using Latent Effects Modeling to understand the trends & patterns in their data.  Trends that no human analysis alone could find.

 The current COVID 19 pandemic is probably the best example of data and patterns that are crucial to understand – but have several crucial variables.  The data and the datasets are so big that looking for trends or patterns and factoring in different variables becomes humanly impossible.  Dr. Christopher Murray at the University of Washington’s Institute of Health Metrics Evaluation is using Latent Effects Modeling to help the world understand the impacts of many trends and scenarios. 

 Large enterprise marketing organizations can have many relationships that influence revenue. Whether that’s a network of partners, campaigns, etc. the ability to analyze long term impact of variables can be challenging. The application of machine learning methodologies can reveal insights buried in the hidden and layered interactions of variables. The individual impact is often imperceptible but when revealed through an advanced analytical approach can result in eureka moments that influence future resource allocation and planning.

 The  best way to really get your mind wrapped around Latent Effects Modeling and other AI/ML power tools is to consult with those of us who are using them.  They work for us and our clients, as we write. Get in touch. We’re always ready, willing and able to talk.



The Bay Area


(206) 420-6121


Mon - Fri : 8am–6pm PST

Sat - Sun: Closed

applying AI: packaging’s impact on sales

llWhen we look at the fast moving consumer goods (FMCG) landscape, we see nothing short of incessant shelf warfare. And few weapons are as critical to victory as package design. What catches whose eye? Is the reaction positive, negative, indifferent? 

In the olden days, we’d wander alongside a typical shopper in real retail aisles, in real time, and try to tease out answers. Along with these shop-alongs were consumer purchase diaries and facility interviews and questionnaires and similar so on and so on.

For any experienced marketer, It’s obvious and frustrating that these just can’t be too reliable. Especially when one-tenth of one percent in market share can be the difference between taking the hill and taking heavy brand casualties. 

One yearns for more precision.  

Lock and load our partner’s proprietary AI methodology: BioNimbus, weaponized in this case in the form of virtual reality (VR). 

Don your VR headgear.

how AI gets inside your subjects’ heads. literally.

Just as gamers navigate from choice to choice while playing, VR subjects wander in the virtual environment. One of our successes, for example, took them to the cereal aisle, with its cartoon characters, famous athletes, organic or not, less sugar/carbs/protein or not, NEW! or not, different color palettes for the same package. The variants are infinite.  

Our VR headgear tracked their actions, both voluntary and involuntary.  We saw them linger over certain packages, over certain parts of the package, and for how long. We saw them take some packages off the shelf, turn them over for more information, return them to the shelf–or buy them.

Borrowing from the gamer model again, our VR tactics can include incentives for “shoppers” to try or buy, to earn points to redeem for prizes or discounts later–many ways to measure product appeal by subjects’ agreement or refusal to accept incentives.

Compared to old-school methods that often drive (or bore) away participants, VR is a virtual walk in an incredibly data-rich park. Fun for the subject, potentially brand-making-or-breaking for the marketer.

So you say, “OK, so far, most learned AI practitioner, convince me more?” 

Well, why don’t we look at results? They’re beyond proofs of concept. They’re proof of AI in practice. 

In the supermarket exercise, we tracked  every virtual package removed, studied, or ignored, every eyeball flick and linger, every brainwave and skin response stimulated.  

The learning?

Masses of incontrovertible data showed that consumers preferred the packaging they already had. Talk about saving boatloads of money on new packaging–that could have hurt, not enhanced, the brand and the bottom line! 

virtually limitless

Our BioNimbus tool chest is packed and ready to go. It offers marketers virtually limitless customization options. We can tee up an auto dealership or any physical space that you can study for brand messaging, impact of interior/exterior architecture, lighting, background music, sales force wardrobe … or a cell phone or tablet, a web page, a golf cart, an exotic fish, a face or body, an ad, a poster, a car, an interior of a home or office — whatever you want.  

Speaking of proof that we’re on to something big, our client list offers plenty. Talk about intelligence–nothing artificial about our relationships with leading brands.  



The Bay Area


(206) 420-6121


Mon - Fri : 8am–6pm PST

Sat - Sun: Closed

applied AI in marketing: healthcare product management

Here’s a case where AI rolled into the ongoing rollout of the Affordable Care Act (ACA). To avoid getting lost in the weeds, let’s just recall that there were big-money battles involving healthcare providers, the Federal government, and the confused, newly ACA-insured and wanted-to-be insured. 

Problem: Two very large healthcare providers urgently needed to learn about the new members who came in through the ACA. 

Like today, millions of Americans had billions in medical bills they were unable to pay.  Big money was at stake for providers who would potentially have to take on those liabilities. 

So who were the new ACA-driven guys? For their own readiness to serve, providers needed to know. What were their likely future care needs, their payment resources or shortcomings?  What if they were uninsured and had no health history?

One organization saw itself already losing millions. They called on an ACA provision that allowed them to sue the Federal government for repayment of losses. They needed accurate data to support their huge claims of loss.

The other organization tried a different approach: remediation, i.e., identifying high-risk customers and negotiating agreements to reduce likely losses. Put another way, offer these folks plans with get healthy/stay healthy incentives like diet and exercise programs.

But how to identify those customers? No human hands could possibly manage all of the data pouring in. 

How AI machine learning managed a flood of data 

Imagine a never-ending downpour. Trillions of raindrops, some useful, some toxic, each one interacting with others, changing as it falls depending on what it encounters. 

Running this awesome show is our immune system, tasked with identifying each drop in a nanosecond and deciding its fate. Kill? Alter? Allow to pass untouched? Send elsewhere for further analysis?

You may say, “Nice analogy, Mr. AI Guru! And so?”

So the rain is terabytes of data. For data management, AI is our immune system. 

And the machine learning (ML) model we devised paid off handsomely.

Insight 1: We were able to identify those few among the raindrops that were new recipients:
  • Who were and would be active consumers of provider services and 
  • Who brought high net value to the provider, as defined by dollars paid in premiums minus the cost of providing those services. 

But the learning didn’t stop at finding those profitable sweet-spot clients. 

Insight 2: Among the millions of new clients eligible for remediation, the model found those whose cases promised the most immediate loss-mitigating results–and put them at the head of the queue.  

Result? Millions saved. 

Talk about intelligence–nothing artificial about that bottom line.

There’s a lot more where that came from. Service delivery strategies, product development, optimizing program results?   

Yes, let’s talk about intelligent modeling in AI and how it can uncover unique solutions.  



The Bay Area


(206) 420-6121


Mon - Fri : 8am–6pm PST

Sat - Sun: Closed