Ask chocolate lovers if more chocolate is better and you will get an emphatic “YES”!   But from a business perspective, this isn’t always the case.  Too many choices can work against a brand, retailer and even consumer.  This is often called the ‘Paradox of Choice’. 

An iconic global brand was having some chocolate challenges.  Namely, their retail stores were growing overstocked with product options, often making their small, upscale retail footprints sometimes cluttered and confusing.  Merchandising variations were hit or miss and out of stocks, a dreadful occurrence.  Sales of products in 3^rd party retail were also flattening, spread across too many SKUs.   And, the online experience was not much better in the U.S., Canada and the U.K. which ,along with Japan, made up the bulk of the company’s’ global e-commerce sales.

The company was tipping the scales at a hefty 4,600 SKUs.  Inventory management costs were climbing, as were logistics costs for e-commerce required to ensure shipped product arrived at the consumer ‘s door pristine, rather than a soggy or melted mess.  Managing thousands of SKUs, especially fruit-based ones, was a constant headache for the company. Seasonal variations in size, shape, texture, and taste made it nearly impossible to maintain consistent quality.

Conventional Wisdoms Collide with Cognitive Bias

Yet, the company clung to the belief that ‘more is better.’ The fear of losing sales kept them from cutting down on product options, even though there was no data to support this assumption. It had simply become accepted wisdom, despite the lack of evidence.

 Because this was the ingrained belief, it had become the accepted fact.  The problem was that there was no empirical evidence to back this ‘accepted fact’ up. Still, the global CIO had a hunch.  New to the organization, he saw things with fresh eyes.  He believed that the perception and the internally shaped guidance advocating status quo was wrong and was being enhanced to support the conventional wisdom and status quo.  To prove this out and find an optimal and profitable path forward, a project was established using Fingerprinting to rationalize and optimize omni-channel mixes , marketing and merchandising efforts. 

As FELIX analyzed the brand’s product, customer, merchandising and sales data, as well as a voluminous body of external, related data FELIX could tap into,  it began to isolate and amplify some very hidden, yet very important patterns.

One critical insight FELIX uncovered was ‘Product Anchors’—key products that influence customer behavior and drive sales across related items. These anchors act as hubs, linking consumers to a broader range of SKUs through shared characteristics and incentives. DecisionArts’ Fingerprinting™ process revealed these connections, enabling smarter merchandising strategies.

How these Fingerprinting linkages and Product Anchors work is also illustrated in DecisionArts’ Use Case, “The Myth of Brand Loyalty - how data reveals what’s truly driving consumer choice”.

FELIX was able to identify the hidden ‘Strong Bonds’ that exist between the Fingerprints of consumers, incentives, geographies, marketing and merchandising content, CRM content, return data, SKU data, nutrition panel data, packaging and pricing data.  All this information is plotted in FELIX’s knowledge graph. 

Fingerprinting allows FELIX’s AI tools to begin making assessments within the graph between visible and invisible linkages.  In other words, FELIX began to plot a road map of truth which illuminates what will increase customer satisfaction, SKU reduction and growth in revenue and profitability.

FELIX provided SKU-level guidance on:

·      Optimal e-commerce product assortments and merchandising options within the U.S., Canada and U.K.

·      Optimal composites for the branded stores; for third party retailers and for BOPIS (Buy Online, Pick-up in Store), to ensure quality and maximize basket sizes. 

·      Provide the necessary options for each market and location while at the same time narrowing the nearly endless universe of options and variations currently bogging the company down.

FELIX  successfully identified over 30% of SKU’s that made up 66% of the long tail, which fit  the company’s classifications for product sunsetting.   

“What’s the big deal?  SKU Rationalization is relatively easy and straight forward.”

Identifying low performing SKUs seems relatively straight forward.  However, it isn’t always quite as simple as it appears on the surface. 

Let’s look at one such example

Chocolate covered strawberries seem like a great product to sell.  They are iconic; visually appealing and can deliver a variety of pleasing flavors in each bite. 

Seems like a no-brainer product to create and merchandise.  Right?

Turns out, the reality of manufacturing them at scale isn’t quite as straightforward as their small batch or homemade cousins.  Commercially manufactured chocolate covered strawberries are expensive to source consistently, perishable with limited shelf life and e-commerce delivery in hot weather can be an issue.  Plus, for this brand, the product only seemed to sell during certain holidays and seasonal windows.  

 So, selling chocolate covered strawberries is a bad idea.  Kill them.

Hold on…On the surface killing them makes as much sense as the original idea to sell them.  But killing the SKU would have far-reaching and unintended consequences.  Identifying these hidden realities is one of FELIX’s skills.

Finding the hidden-and valuable middle ground

FELIX identified that chocolate covered strawberries were a Product Anchor which had strong bonds with over 18 other products and nearly 50 SKUs. 

Because the product was an Anchor, when it was merchandised with the other strong bonded products and SKUs, the consumer could easily move from the anchor product (the strawberries) to one or several of the linked products for purchase if merchandised correctly.

 Instead of the consumer purchasing the lower margin, difficult to manage strawberries, they were instead seamlessly directed to other higher margin items which were tightly aligned with the consumer’s characteristics, attributes and criteria. 

 With this change, conversions of chocolate covered strawberries decreased 40% online, however, profitability on the product, as well as each sale including the product went up. 

 Product sales which began with chocolate covered strawberry search but did not include chocolate covered strawberries in the purchase also increased by nearly 20%.

 Furthermore, returns and QA complaints related to the strawberries decreased by nearly 90% AND baskets value associated with chocolate covered strawberries increased by 20% and repurchase rates within a 90-day window increased by 38%. 

 Killing the SKU would have wiped this significant financial gain out entirely, as all the value was hidden, having only been uncovered by FELIX’s Fingerprinting process.

 FELIX revealed that over 30% of the company’s discounting tactics were counterproductive. These promotions often backfired, leading customers to only buy the discounted items in bulk, rather than a variety of products. This strategy hurt the brand by reducing basket diversity and limiting overall sales growth.

 When FELIX mapped out the clusters of strong and weak bonded items visually, the brand’s management was shocked to learn that  the majority of the underperforming products were ‘Clones’ of other better selling products or ‘Orphans’ consuming valuable space and resources, apparently unbeknownst to everyone. 

Let’s look at Clones and Orphans in greater detail:

CLONE:  A SKU that is an extension of another and whose characteristics are substantially alike to the product it was cloning.  Nearly universally, this was unit size or flavor, or flavor-assortment based.  Think of a package of two chocolates rather than four. 

ORPHAN:  A SKU that stands alone as a unique flavor, size, packaging combination.  Think white chocolate/hazelnut clusters.  Orphans seem like a good idea on the surface but are not.  They must be subsidized by other products that are not in the long tail.

In addition to tagging the previously hidden clones and orphans, FELIX was able to successfully identify that the bond strength between products varied online and off, including Orphans.  ‘Bond Strength’ determines whether dissimilar products will be desirable, and likely to be purchased by certain consumers whose Fingerprints match the products and when assortments and merchandising are optimized for this.  An Orphan offline might not be an Orphan online and may easily be paired with another item.

The whole truth is rarely found on the surface.

The result of this initiative had an impact well beyond sales and product marketing.   These improvements also reached the company’s supply chain, distribution and logistics,  as well as customer satisfaction, retention and profitability.  What appears to be true on the surface isn’t always the case. 

 FELIX’s output can be delivered on-demand or as intelligence feeds which can be embedded into other enterprise tools.  In either delivery model, FELIX can effectively  provide decision support and guidance to professionals who seek to understand what exists beyond what is traditionally visible and the implications of available options. 

When we think of brand loyalty, we often picture people sticking with their favorite labels out of some emotional attachment or even a sense of identity. But how much of that is true?

As it turns out, not much. Advances in decision science are showing that consumer choices are more influenced by pattern recognition, availability, and intent rather than pure brand loyalty.

Take Porsche as an example. While the logo might symbolize status or inclusion, it’s unlikely Porsche owners would all move to an electric Taycan because it’s still a Porsche.  

These owners are in love with the 911 because it aligns with who they are and how they see themselves. The brand simply benefits from the alignment of the driver and car's characteristics and attributes.

This reveals a hidden truth: it’s not about the brand itself, but about what the brand’s products represent at a deeper level.

 So, how do brands find out what their customers truly want?

 Enter tools like FELIX, which uses graph-based decision science to reveal hidden connections between products and consumer behavior.

 A real-world example of this is the launch of a super-premium ‘wet’ Mac & Cheese. Data analysis showed that demand existed in unexpected places. Instead of chasing crowded markets, FELIX pinpointed locations and linked products that aligned with the brand's target consumer base. This led to a successful product launch in previously overlooked areas, despite assumptions about market demand.

 Let’s move to more relevant examples we all can appreciate more directly.  Decisions are made in large part based on intent, pattern recognition, friction and an outcome that is aligned with the decision maker’s characteristics, attributes and criteria.

 What does this mean?  What is ‘criteria’?  That seems very…vague.

Not at all.  It is in fact specific to the user.  Criteria can be substituted with the word ‘intent’.  What is the consumer’s intent?  But don’t focus too heavily on this one thing.  Again, decisions are made bases on pattern recognition against intent (what does the person seek as a goal or an outcome), friction (how hard is it to find or identify the thing that aligns with them best), and their unique characteristics and attributes.

Let’s look at a real-world example which illustrates the economic impact of embracing an evolved model.  Before we do, let’s point out that in antiquity maps were largely just ‘directionally’ correct because there was such a significant body if missing data, so the map makers filled it in with what they believed.  And, for what they generally knew, distances were off, by a lot.  These errors were due to the earth being  a sphere instead of being flat and that difference must be part of the math to get an accurate determination of distance.

 Enough history.  Let’s go shopping.

A new super premium ‘wet’ Mac & Cheese product was introduced into several targeted markets in the Midwestern United States. Fingerprinting data from the physical channels was used to inform elements of the digital launch.

Wet Mac & Cheese is commonly associated with ‘Velveeta’.  There are very few ‘wet’ premium Mac & Cheese products on the market.  So, this means that there is not a market for wet super premium Mac & Cheese.  Empty category, no demand.  Crowded category, lots of demand. 

 Follow the demand?

 Not necessarily.  Crowded categories mean Paradox of Choice for the consumer and smaller share for brands at a higher acquisition and retention cost.  Running to the perceived trend is a strategy, just not always the best one. 

 In this case, the customer segment which aligned with this product were already customers who frequently bought other products in a variety of other non-related categories, and had some specific characteristics, attributes and criteria which using normal methods of analysis were totally invisible.

 Decision Arts’ fingerprinting found the commonalities and invisible linkages in the data FELIX ingested and analyzed.  Specifically, which other products the Mac & Cheese p_Key score overlapped with, and which customer’s p_Key scores overlapped with both. 

Fingerprinting also identified where these customers were most prevalent in the retailer’s chain of locations and which ‘anchor products’ had the strongest bond to the new Mac & Cheese.

In this instance, FELIX indicated an unmet, high-volume demand with a 10%+ CAGR that was outside of the demand category that all other brands were competing for.

 The p_Key is a unique, highly precise identifying address in FELIX’s knowledge graph, see it at work in this simple explainer video) and is core to the Fingerprinting process FELIX utilizes.

This is what happened

 When the new Mac & Cheese was merchandised in both stores  within the identified ZIP codes that fingerprinting illuminated as optimal.  Mapping allowed to drill down into Fingerprint and strong bond linkage details (product links, frequency and value scores, target customer density and scores, etc.).   The product was also monetized in locations which FELIX indicated were less than to not optimal. 

 This determination was  based on the strong bond linkages FELIX identified and which strong bonded products linked to the featured product met certain performance goals.  FELIX also provided guidance as to the product performance, by ZIP code indicating optimal product launch and sell through areas, as well as which areas to grow into and to exclude at the outset.

Aisle Performance

 In stores where no Fingerprint merchandising was conducted, the new Mac & Cheese was present in just over 1% of market baskets.

 In stores where Fingerprint merchandising was conducted and where concentrated consumer profiles patronized, the Mac & Cheese stocked in the aisle, in the same configuration as non-ID’ed stores was present in just over 3% of market baskets.

 Merchandising Performance

 A four-shelf refrigerated endcap was stocked with all four Mac & Cheese SKU’s but each SKU was aligned with a series of cross category products the Mac & Cheese had strong bonds with.  These ‘anchor products’ sold well with the consumer segments identified through fingerprinting.  The targeted consumers, the anchor products and the featured products all had overlapping fingerprints.  Fingerprinting also identified certain traits which were linked to messaging receptivity, which in turn drove a specific merchandising message.

 What were the Anchor Product Categories?

·      IPA Beer (specific brands/flavor profiles were ID’ed)

·      Smoked thick cut bacon (specific brands and price points were ID’ed)

·      Two flavors of cheese spread (specific brands and flavor profiles were ID’ed)

·      Artisanal seasoned crackers (brands, specific flavor profiles, ingredients and price points were ID’ed)

·      The Featured Product was consistently included in 43% of the identified consumer group baskets when 1 anchor product was present and selected from end cap on average.

·      The Featured Product was consistently included in 53% of the identified consumer group customer baskets when 2 anchor products were present and selected from end cap on average.

·      The Featured Product was consistently included in 63% of consumer group baskets when 3 or more anchor products were present and selected from end cap on average.

What was required?

·      Retailers were provided an end-cap schematic to follow

·      Merchandising was created and paid for using existing processes and resources

·      Shelf planograms were retained

·      No new slotting allocations/fees were required

·      No specialty redemption methods had to be created

·      No new enterprise level hardware (BTLE, Computer Vision) required

·      No new retailer, distributor or brand inventory management functions required

·      No specialty pricing required, POS or inventory management integration required

·      Maintain traditional collaboration methods between brand team and retail team

·      An appetite to increase revenue and EBIDA (brand and retailer)

·      A willingness to evolve (only slightly)

·      A desire to try something outside of the status quo

·      A willingness to trust hard data

·      An understanding that all AI is not the same, one size does not fit all

·      A believe that the right AI  solution is a tool (like Night Vision Goggles are simply a tool) to help you see something previously invisible, rather than some magic silver bullet that will displace everyone and do everything.

To learn more or to set up a discussion on how FELIX might apply to your business, please reach out at:   hello@decisionarts.io

Click here to visit DecisionArts                                                Click here to learn how FELIX works

As we have said before, AI is a label. Just as 'car' is a label. To make a good decision, you need to know if you are talking about a Toyota Prius or a Dodge Hellcat. If you don't know which is which, for a decision maker, picking one or the other may be a disastrous mistake.

To make better decisions and be able to effectively put the underlying mathematics (which is what AI, machine learning and software is based on) to work, it's important to link the proper tool to the proper application and it needs to be understood and applied in the right context.

Our engineering, data science and decision science team members think in terms of the underlying models and math. That's their job. They need to understand the rocket science, and to push the boundaries. They need to understand and be confident in the 'Art of the Possible' but also be able to learn from other models and disciplines, like neuroscience, psychology, sociology, and quantum physics to name a few.

But, that doesn't mean our customers and those who use our tools need to focus on those things. Instead, they should be focused on the hard work they do as brand and retail professionals...making sure that the right products, promoted in the right way get to the customers who want and need those items most. Delivering on their own 'Art of the Possible'.

The brand and retail professionals we work with should be able to tell simple and powerful brand stories and create amazing experiences, delivered consistently for their customers, across the channels they feel are important.

They need to work hard to convert 'Light Buyers' into 'Frequent Buyers' and also to introduce their brand, products and experience to new customers who have not yet purchased, but will benefit from engagement. Ask any of them and they will tell you this is no easy feat. And, it is only getting harder and harder.

But that is their job, which they do well. Our job is to enable them by helping to separate the signal from the noise in the marketplace and identify the often invisible connections between people, products, places and content.

To do this, we created our own simple story to introduce DecisionArts and our Knowledge Engine, FELIX.

Why call our Knowledge Graph FELIX? Well, Felix is name is latin for 'Happy' or 'Lucky'. DecisionArts' Mission is:

"What if you could learn from mistakes without ever actually making them?"

So, FELIX seemed appropriate. Take a look at this short introduction video on how FELIX works and why, if you are a CPG brand or retail executive, it may matter to you and your team.

Either click on the above image or this link to watch how FELIX delivers value.

DecisionArts. Using Decision Science to move AI from Cool to Commercial. If you are interested in learning more about FELIX or DecisionArts, drop us a line at hello@decisionarts.io to set up a time to have that discussion.

But then again, we were told there would be flying cars by now too.

The use of the term 'AI' is nearing saturation. For instance on LinkedIn, the vast majority of profiles tied to marketing, sales and consulting, as well as, software development now include 'AI' as a core tenant but without being clear as to what or how (or even why). And, often we would hazard to guess, without a clear understanding of the detailed underpinnings of what the science behind AI is and is not.

That said, there are LOTS of skilled professionals and technicians who publish innovative content that are bona fide subject matter experts in one or several (or even multiple) of the machine learning, mathematical, data science, decision science and software elements which are part of the composite which is 'AI'.

'AI' is simply a label, not a discipline, like 'car' is a label. A Toyota Prius and a Dodge Hellcat are both cars, yet very different. Different buyers, different carbon footprints, different purposes for use and so on. Saying you sell cars or you want to buy a car might lead to one choice or type of owner as to the other. So not useful.

It is like saying all software is good. Software is going to change the world. If you aren't embracing software, you are going to lose. See what we mean? 'Software' as a term, is important and an embedded part of our society but what kind of software are we talking about? To bring up some old terms for illustration, is it cloud based, is it thin client or perhaps thick client software? Ok, cloud based. A bit more specific yes, but far from helpful in determining, measuring and delivering commercial value.

Every technology has evolved from general use to specific application over generations of evolution; the development of 'species'.

AI has subsets that are starting to follow the evolutionary track of purpose-oriented existence and performance. Why? Because broad or blunt applications and labels are of limited use; like using a 16-pound hammer for every carpentry task.

At DecisionArts, we look through the lens of Decision Science. Specifically, the value of one choice being better than another for specific users with specific needs and specific outcomes relating to specific commercial goals.

For us, optimizing assortments, merchandising and campaign funnel conversions with respect to content, geographies, pricing options and customer segments within CPG. Uncovering previously unknown and otherwise invisible threads and triggers which yield new buyers and new sales conversions.

To deliver value commercially, being that specific isn't just important as we go forward, it will be a requirement.

We believe the value in decision making and content creation tools which currently fall into the broad category of 'AI' need to get more specific. 'AI' solutions need to evolve into 'species' in order to be adopted and relied upon as a new standard of operation.

These species must yield measurable results, consistently delivered, to warrant an evolutionary jump to AI being ubiquitous and accepted as a new normal-rather than novelty.

It's a long headline, but it accurately states our point.

Let's start this piece with one of many examples of fiction being perceived as fact from Freakonomics to set the table on the point of this article.

Did you know that over the years Listerine advertisers were able to convince tens of millions of Americans that bad breath was a disease and that Listerine was the only and then later, the best cure? "Kills the germs that cause bad breath", is the tagline.

Delve into the facts beyond the conventional wisdom pounded into the population's consciousness and you learn that not only is it not true...or at least in how it is perceived isn't true, the product an have a hidden negative effect. The ingredients, including alcohol, do wipe out a lot of 'germs' (a.k.a. bacteria), including the litany of bacteria that are there to help you process food, keep you and your mouth healthy (your bacteria 'biome' commonly understood as the 'gut biome', which today hundreds of products now help you grow and sustain).

Billions of dollars are spent each year bringing new CPG products to market and promoting their sale and adoption by consumers. Most of them fail. And, much of that precious capital spent to launch and grow products is poorly allocated. Why are stacks of cash and hundreds of thousands of valuable man-hours wasted?

Because we (ok, the industry of brand CPG and retail) are focusing on the wrong metrics and are still largely guessing as to where to place bets (investment), where and when...and why. Digital less so than physical space but even there, there is room for improvement in new and exciting ways that graph intelligence can inform against.

Quoted from David Ogilvy, the head of the famous Ogilvy & Mather advertising agency's book, 'Confessions of an Advertising Man (pp. 86-87)':

Half the money I spend on advertising is wasted, and the trouble is I don't know which half.

There is much new intelligence that we can extract from FELIX, DecisionArts hybrid GNN knowledge engine.  This intelligence can include products which appear to have little to nothing in common will resonate with the same consumer , doing so with uncanny accuracy.  Which content and incentives will drive conversion with specific customers, at specific times in specific locations.  Everything within FELIX’s knowledge graph has a unique ‘Fingerprint’.

Putting this approach to work can yield powerful new results for brand marketers.  In this piece we are going to break down some of the shocking conclusions emerging AI and in particular Knowledge Graph driven systems like FELIX are revealing.

FELIX can accurately divine what recommendations or substitutions will yield the greatest results, as well as where and when.  FELIX can accurately identify emerging  white spaces, where products that don’t yet exist will achieve superior sell through results.  It can fetch products based on identified p_Keys or even queried attributes or even bond strength linkages between items. 

FELIX can enter and interpret its graph from any direction or question/knowledge gap, as can the product marketer or even the end consumer with access to FELIX to get to the best or right answer best tailored to their particular needs and desired outcomes quickly. 

The novel approach found in FELIX’s structure and algorithms is what drives insights and results that prior to broad AI awareness in the market was thought to be impossible. 

What does all this mean? 

It means that what marketers have long believed to be best practice and the ground truth of product marketing, isn’t.

Sure, some of these are generalized statements. We get that. The point here is to put a stake in ground and build around that with empirical evidence. Hard data tied directly to definitive outcomes.

We do plan on filling in the gaps with data as we move forward, so stay tuned.

DecisionArts has also been able to analyze program performance data which lead to new expectations about consumer behavior, product marketing and incentives.  What are some of the key learnings FELIX is uncovering?   Here are a few nuggets to chew on:

• The bond strength between a brand and consumers is much weaker than currently believed. 
•  Loyalty as we know it is largely a myth.  What drives repetitive purchase is not necessarily what conventional wisdom and approaches suggest. 
• Loyalty behaviors can be short circuited just prior to, or at the moment of truth (purchase), through the use of certain incentives and merchandising, which can easily result in a competitive product substitution.
• Habit, entry points and friction play a much greater role to status quo consumer shopping behaviors. 
• Habit and friction can be more easily influenced than entry points.
• Price and friction have a different relationship than previously believed.
• Discounting doesn't necessarily work as commonly believed.
• Specialty or non-traditional merchandising, including non-alike items is materially more effective than traditional planograms in generating share growth for new products or products in maturity phases.
• ZIP codes adjacent to one another yield very different sales results using traditional assortment, merchandising and marketing techniques.  A more selective approach applied in identified ZIP codes can have 2x greater results over simply making a product broadly available within a regional marketplace.
• Characteristics and attribute linkages are consistently more important than 'brand strength' as we currently think about 'brand strength.'
• Characteristics and attributes drive selection and performance can be identical in geographically diverse ZIP codes in which surface data would indicate otherwise.
•  Look-alike models, PRIZM data, focus groups and the use of third party aggregated data sets are important tools to use as a starting point but the value and accuracy quickly falls away. Characteristic, attribute and criteria alignment is significantly more effective than Look-alikes, PRIZM segmentation and focus groups.
• Broad based market share growth can be accomplished using more surgical techniques. In essence, the current model of 'Spray and Pray' is no better than betting on the trifecta at the Kentucky Derby based on a horse's name.

Traditional product development, assortment, merchandising, promotion, marketing and advertising techniques do still work today.  It’s just the variables these tools and programs must account for means the effort required is much greater than ever before and results and their consistency is lower. 

But that is how it is today.  More channels, more noise, more tools, more options to choose from, smarter competitors, new entrants, market forces, economic dynamics, etc., etc. etc..  However, unlike Blackjack, card counting with AI driven solutions is perfectly legal, if you pick the right approach and put it to work correctly.  All AI solutions, like any solution or even like card players are not equal.

You can still guess and rely on historical methods, but the goal is to win, so why would you?

Everyone is touting AI. Everyone seems to be doing AI.

Word of warning, this isn't a short, fluff marketing piece. It is 1,600 words about things that can and will impact you, positively or negatively.

As a brand or retail executive, your professional experience is vast and deep. Your corporate technical team, as skilled as they are, is only as good as their last webinar or article consumed.

You are not in the business of AI, nor in the business of buying and using software systems and tools. You are in the business of designing, manufacturing, distributing, marketing and selling innovative products that customers love and rely upon.

You are in the business of delivering shareholder value. Marketshare acquisition and retention. Developing competitive advantages.

All of which is hard enough.

When adopting an AI enabled system, how do you know what you are getting? Really?

Like you and members of your executive teams, many on our team hav built and even bought complex and expensive 'enterprise systems'. We've built, marketed, sold and then had to support software with millions of lines of code.

We've also bought and used in previous endeavors six, seven and eight figure enterprise software packages from Microsoft to Salesforce to Adobe as well as powerful and promising niche solutions from small start ups.

As you well know, very little out of the box is plug and play.

Often, a consultancy with relevant expertise would have to be brought in to help navigate what a system actually did well versus what it was promoted to well in the webinars and sales materials.

There was a very specific incidence that illustrates this we will use as an example. As cloud based CRM systems were exploding just over a decade ago, a marquee software package from the market leader had been purchased, after nearly a year of analysis. There were issues that kept a team of engineering FTE's busy configuring modules to work with each other correctly after a very expensive purchase. But a series of mysterious data leaks and glitches in how the system escalated and fetched record data correctly began to bubble to the surface.

When this was escalated to the consultant's most senior subject matter experts, located in another market, the answer was quick and to the point. The module purchased was not yet stable, for a variety of reasons. Our patches were on top of the software company's improvements and re-deployments. All the result of brand new systems doing new things in new ways.

Another client who had devised a work around solution was introduced. Another firm with the same problem, just three months ahead.

The consultant's opinion was the module performed poorly at this lifecycle stage and should not have been purchased. However, the truth had been buried. There were signs. Dev forums here and there bemoaning the range of issues, however no clear business intelligence that could be easily deciphered against a set standard of performance. Because the module was part of a larger, multi-year subscription, the cost had to be carried without the benefit of a tool that did what it was advertised to do in the way it was supposed to work.

Still, we had access to another firm's work around, which was adopted (and paid for).

This process led to many unintended consequences, such is the case of new tech.

It's called 'bleeding edge' for a reason.

It was in a large executive meeting with a team of legal counsel weighing the options (in 2014) when a phrase which became DecisionArts tag line was uttered.

"It would have been great to have learned from this mistake without having to have made it."

The executive went on to say that having an independent evaluation against a set of industry performance standards (or regulations) was as important to software and services as it is for manufacturing.

He continued,

"The manufacturing company I invested in, to do business with the OEM's it supports, has to get ISO certified, them maintain against against that certification AND be audited by a third party and provided a performance score. That score determines their value and viability to their OEM clients."

The fact remains in each case, that unless you can validate what you've bought, you are left with figuring it out later. After the fact.

AI tools are too important and too powerful to leave it to chance. The regulation is already in play to require disclosure for any firm using AI tools. Quantifying performance against the standard is next.

So with all of this said, we would like to offer a very important few words on standards, AI governance, quality, and transparency compliance.

One area at this stage which presents a glaring gap is in the efficacy of the artificial decision process as well as security and accuracy of the inputs used in this process and the outputs the processes yields.  Without a clear understanding of the boundaries and scope of this efficacy, there is no way to definitively determine the safety of the input data (such as consumer PII) and the accuracy or malevolency of the output.

Artificial Intelligence Systems, to broadly commercially viable must be uniformly trusted within the sectors of broad commercial use. At the same time these systems must operate within defined boundaries of acceptable behavior (not lying, cheating, or stealing) and conforming with all applicable governmental standards, regulations and laws in place within the marketplaces the system operates. No some of the systems, not 'it depends', not 'trust us'. It is binary.

In essence, AI must adopt and enforce its own version of well adopted and time-tested standards such as GAAP (Generally Accepted Accounting Principles established to govern the common structures of financial accounting processes, and the financial standards for recording these transactions and values).

Standards such as GAAP can also operate in rapid change and evolution environments such as technology.   IEEE SA (The Institute of Electrical and Electronics Engineers Standards Association (IEEE SA) is an operating unit within IEEE that develops global standards in a broad range of industries).

Fortunately, this is under way. We also have emerging Artificial Intelligence standards such as NIST's AI framework, which is now accessible to industry.  It is important to keep this body of work and its applicability and use at the forefront of the work that all professionals operating in this space, as well as, who use AI driven tools (such as those of you reading this) understand and are compliant.

The link to the NIST standard is here.

The goal of the AI RMF (National Artificial Intelligence Initiative Act of 2020) is to offer a resource to the organizations designing, developing, deploying, or using AI systems to help manage the many risks of AI and promote trustworthy and responsible development and use of AI systems.

The Framework is intended to be voluntary, rights-preserving, non-sector specific, and use-case agnostic, providing flexibility to organizations of all sizes and in all sectors and throughout society to implement the approaches in the Framework. The AI RMF is intended to be practical, to adapt to the AI landscape as AI technologies continue to develop, and to be operationalized by organizations in varying degrees and capacities so society can benefit from AI while also being protected from its potential harms.

The more commercially valuable and far reaching the sector such as consumer packaged goods, the more important self-adoption, self-governance and self-adherence is to an open, secure and stable marketplace that can be trusted, regardless of speed of innovation or the sophistication of commercial AI advancements.

We’ve seen in the past the impact from industries whose self-governance and business behaviors didn’t keep up with public interest, customer, employee, shareholder or  individual interests, or even the industry’s long-term welfare (think Sarbanes-Oxley and the recent social media Congressional Hearings).

When things go sideways, this is what can happen. Let's avoid this with AI enabled systems. We have this ability now but it is important to constrain the Genie before letting it fully out of the bottle.

DecisionArts is committed to this standard and remaining compliance to NIST standards in order to ensure that our customers can universally trust and rely on the data feeds we provide as being truthful, accurate and in accordance with all applicable global laws such as privacy, consent validation and data sovereignty.

Specifically, here is some of what DecisionArts commits to deliver to our brand and retail clients as well as partners:

Global Privacy compliance to all regulatory body standards.

100% Consent Validation and associated audit certification and validation to global regulatory body standards.

Data Sovereignty audit certification and validation.

Full PII compliance and audit validation on all first and third party data.

Detailed performance score cards regularly made available for or prompted by all DecisionArts clients for all AI enabled DecisionArts graph and algorithm structures and outputs, audits conducted by embedded Truyo AI compliance software and data engines.

Audited AI program governance and performance compliance to NIST standards.

AI solutions and data usage are moving and evolving faster than ever.  Global brands and retailers, as well as consumers have as much to lose as to gain, so ensuring for protection from unintended consequences is critical, not just in the future but now.

Policies are only as good as the standards applied and the ability to demonstrate compliance with transparency.

We expect you to question how this is being done. We'd question it. So, here is how to do that if your are interested or if you want to learn more.

Simply drop us a line at info@decisionarts.io so we can set up a time to walk through your questions and share a few things with you.

Decision science, when viewed through neurological and cognitive lenses, allows researchers to delve deep into the intricate and complex processes that underlie human decision making.  Cognitive functions such as attention, memory and executive function play a crucial role in how we process information and ultimately make decisions.

 Neurological studies in decision science often use imaging techniques like fMRI to observe brain activity in real time to better understand how decisions are being made.  This helps us to understand the neural correlates of decision making, including bias influencers such as dopamine which influence risk-or perceived risk, or how the amygdala’s response to fear can influence and impact decisions.

Cognitive psychology contributes further to decision science by exploring heuristics, mental biases, and perception impact decisions.  These biases, such as ‘Confirmation Bias’, or even the ‘Availability Heuristic’ can lead to errors in judgement which lead to flawed decision pathways and outcomes.

 Recent work, led by Google’s DeepMind team in conjunction with Princeton University researchers, focuses on new methodological approaches for technology which mimic how the human brain works.  Furthermore, there have been several recent advances in decision mapping led by researchers at Carnegie Mellon University’s (CMU) Center for Advanced Process Decision Making, as well as, CMU’s Department for Social and Decision Sciences which carry value in building artificial intelligence models.

Marrying these emerging approaches allows data scientists, information architects, data and software engineers to advance Artificial Intelligence Models beyond our first and even second generation LLM’s. 

Interestingly, some of the most effective AI modeling doesn’t originate from new structures.  Multi-dimensional computing tools have existed for some time.  What makes ‘now’ both new, different and tremendously exciting? 

What is important today for artificial intelligence-based decision science and related systems, centers around three things:

1. Vast amounts of available cross-categorical, structured and unstructured data
2.  Material increases in relatively cost-effective computing power
3. Improvements in the fidelity and function of these multi-dimensional tools and the ability to synthesize tools such as these with specifically designed LLM ‘Species’ to create powerful hybrid intelligence tools working in concert without degradation in information quality or outcomes (NOTE: these stand apart from traditional ‘pure play’ LLM’s in which hallucinations and lies are prevalent)

These models resemble a network of interconnected points or branches which are adjacent or overlap one another.  It’s a model that illustrates how a single decision point can lead to multiple pathways or branches, each in turn offering the possibility of other

intersections, each of which can represent a different potential outcome or course of action. 

 In this model of decision making,  the process begins with the identification of a decision node, which is the point where a final choice which leads to a tangible outcome must be made. 

 This is the ‘Act’ step, even if the ‘Act’ is ‘No Action’.  Interestingly, we have learned

that few if any decisions are final.  They are influenced by the outcomes they inform, and the experiences encountered which reframes the decision node for when the process begins again.

A simple example of this is “Should I touch the flame?”  Burning a finger informs whether it was a good decision to stick your finger in a candle flame. 

The mistake reframes the decision node and the next outcome, which is to likely not stick your finger in the flame. 

 A flame is bad, but what does electricity from the outlet feel like?

A Decision Node consists of a center point, or a ‘nucleus’ from which the branches extend.  Each branch leads to further nodes which specific information encoded which influences the path of the decision. 

This can be broken down into several simple cognitive steps, which have until recently been the state of the art in this process:

• Identification of the decision to be made or identification of a series of decisions that lead to a ‘master decision’.
• Generation of the possible options or actions available or perceived.
• Evaluation of the consequences or perceived consequence of each option, considering factors such as risk, benefits, difficulty and probabilities of each, including influence of biases, acceptance or avoidance.
• Selection of the best perceived course of action.  Selection is never linear.  It involves a complex series of feedback loops where steps and outcomes are assessed and decisions or the steps in a decision are revisited.

The technical models which mimic the natural decision process are evolving quickly and new knowledge graph-based models are emerging.  DecisionArts leverages n-dimensional knowledge graph structures in novel ways to tease out how products, content and consumer characteristics and attributes align.

1. Nodes representing decision points and their proximity to other nodes.  In multi-dimensional knowledge systems, knowing what these nodes are comprised of and their proximity to other nodes and what the various nuclei are comprised of is crucial to the effectiveness of the system and the ensuing processes.
2.  Edges as pathways to outcomes.  Edges in a multi-dimensional model are analogous to branch pathways.  Some edges have stronger bonds, some have weaker bonds which inform possibility and likelihood the path will lead to some defined certainty.
3.  Optimize edge routes and bond strength through algorithm to identify navigation efficiencies.
4.  Exploring all possible then probably possibilities through branching or edge mapping.  Here the algorithms assist to explore all conceivable scenarios based on certain scoring models being applied.
5.  Cycles reflecting feedback loops.
6.  Probabilities and outcomes on weighted edges.  This is an extension of step 3 but here, outcome values are more heavily utilized. 
7.  Sequential decisions in a directed multi-dimensional structure.  Directed structured, with unidirectional edges can effectively model the sequence and information in each step of the decision process, as well as the progression from one step, decision and even outcomes to the next.

Within the graph structure, a measured item (a consumer, a product, piece of content, etc.) and its address within the graph is plotted along with that item’s embedded meta data (the characteristics and attributes) in n-dimensional space.

However, it is extremely important when viewing structures and linkages in graph spaces to what we call 'Distance Warping'. While managing graph structures with tens or hundreds of millions of dimensions can provide high levels of fidelity, when they are output and represented in 2 dimensions, actual distances shown can warp. What looks like close adjacency in a 2D map can warp or change significantly as the perspective point in the graph shifts dimensions.

Distance warping plays a significant role on information and intelligence the system can yield and needs to be accounted for in ways similar to how astro-physicists account for how much light bends around bodies with significant gravitational fields. Without accounting for this, their measurements and the intelligence it yields would be incorrect.

Any item in the graph is surrounded by its various criteria, characteristics and attributes, which give the item shape, like a human cell. And, like human cells, the shape of these items and how they move have patterns. The patterns and profiles of these things along with the myriad of other elements which FELIX tracks in the GNN which are of supreme importance to yielding actionable intelligence.

There is much new intelligence that we can extract from FELIX, DecisionArts hybrid GNN engine.  This intelligence can include products which appear to have little to nothing in common will resonate with the same consumer , doing so with uncanny accuracy.  Which content and incentives will drive conversion with specific customers, at specific times in specific locations.  Everything within FELIX’s knowledge graph has a unique ‘Fingerprint’.

 In our next post, we will show what putting this approach to work can yield for brand marketers, as well as some of the shocking conclusions emerging AI and in particular Knowledge Graph driven systems like FELIX are revealing.

The science is clear.  The decisions that humans make are never linear.  It is multidimensional and directional, is full of biases and shortcuts and it is rarely fully rational.  The process, when visualized, looks nothing like the well-recognized ‘decision tree’.  Using linear tools and statistical or ‘finite state’ models  (LLM) to augment decision making yields rapidly diminishing returns-as many have experienced using recommendation engines or current AI enabled services.

Relying on n-dimensional graph structures which emulate the human mind’s process of information organization and dissemination, DecisionArts enables brands to engage with a greater degree of hyper-personalization, utility and relevance at scale. 

 Decision science is an interdisciplinary field that relies on a broad range of quantitative and qualitative disciplines and tools to effectively evaluate and understand how decisions are made.

 This discipline combines tools from statistics, economics, psychology, neuroscience, physics, and data science.  Decision scientists help organizations to identify, prioritize, and act upon opportunities and challenges, often using data-driven techniques to predict outcomes and measure the risks which accompany those possible outcomes. 

 In essence, decision science aims to understand the often-opaque process of decision making and improve that process through the betterment of outcomes.

 Whether for an individual or within a corporate or governmental context, decision science attempts to provide a structured framework to assist in balancing judgment, experience, and evidence inputs while removing often hidden biases.

 It's about making informed and rational choices based on the analysis of complex data and the application of sophisticated models which can simulate potential outcomes.

 This is DecisionArts’ focus.  Our GNN (hybrid Graph Neural Network) knowledge graph, called FELIX, follows many tested and emerging decision science concepts and structures.

 DecisionArts is a Decision Science middleware application company focused on improving the quality and outcomes of end-user decisions across a variety of industry categories. 

 Our focus is helping commercial brands win in marketplaces where differentiation is difficult,  substitution is rampant, friction is high and customer acquisition and retention is expensive.

 DecisionArts identifies where optimal, yet hidden linkages between brands and consumers exist and the intelligence on what action to take.

 The field of decision science is particularly valuable in areas where stakes are high, and the consequences of decisions are significant.  We understand this and coined this phrase, which represents the value provided by DecisionArts and our knowledge engine, FELIX:

“What if you could learn from your mistakes without making them?”

In the upcoming series of articles here as well as , here on LinkedIn, the DecisionArts team will break down in bite-sized chunks how decision science intersects with artificial intelligence and outline DecisionArts enhanced approach to solving commercial problems, beginning in the CPG space using FELIX, DecisionArts’ Knowledge Engine, FELIX.

In the 70’s there was a television game show called ‘Let’s Make a Deal’.  It was exactly what you’d expect from the 70’s.  The host would pick a contestant.  Say one worked at a farm implement manufacturer, who was from somewhere in Iowa. 

 The host would begin to negotiate when the contestant was on the verge of deciding…did they want to pick door number one, two or three?  Behind one was  new car and another had something like a 100 pounds of potato au gratin mix.  The contestant could also choose a smaller prize box that was certain to have something nice in it but of lesser value.  Say a portable FM radio (hey, it was the 70’s).  The risk was limited even more.  But so was the reward.

Oh, the high drama and dilemma of what to select or not to select.  The audience would attempt to help by screaming “PICK DOOR NUMBER 3!”  What did THEY know that the guy from Iowa didn’t? 

Everyone is engaged.  No one is thinking that like a casino, the odds are stacked against them, and they are all guessing.  But it doesn’t feel like they are guessing.  That’s what makes the show. 

 The guy from Iowa exhales. He has made his decision.

"Door Number Two, Monte.  That’s it.  That’s the one."

 Dang.  The consolation music is queued.  It’s not the Camaro.  Everyone groans.  But as the host points out,  it wasn’t 100 pounds of potato mix either.  That was behind Door Number One. 

 The guy from Iowa stares at what he had just won…a full set of brand-new camping equipment.  Not bad, but not the hot rod either.  .

 When you have imperfect information, perceived trends,  limited tools and a bunch of people telling you what they think they know or what they believe, what do you do?

 Today’s state of the art in product management or merchandising is full of data, just like the game show. .  Reams of data.  But data is like opinions.  Neither by themselves are facts.  Just representations of facts.  Facts come from the insights.  But how many of the insights in today’s CPG world are certain?  Or even mostly certain.

 The terms ‘Spray and Pray’ is still used for a reason.  Profiles are built based on past results.  Look-Alikes are generated.  Focus Groups, whose members are paid to give an influenced opinion are tapped, Prizm data is ingested.  Prototypes, test markets. Incentives, coupons, ad campaigns, merchandising, planograms, slotting fees, co-op dollars, listening engine data feeds, retail data samples, third party rolled up data feeds.  The list goes on.

 Sure, it all informs, but it all comes in across a period of 3 or so months and is, well, less than precise.  As a member of the brand’s cross functional team, you are no better off really than the guy from Iowa.  But it is what you have.  Given that, winning the consolation camping set prize seems like a good day. 

 The problem is, when you do some research across the CPG category, spending is going up as reported in 10-Q filings and market shares are slowly eroding.  Down 1-2% anymore is a win.  Preserve rather than grow.  The CPG game has gotten a lot harder.  Hitting the jackpot required to gain and keep market share is now as much luck as skill. 

 Mind you, the skill is supremely important.  Brand teams MUST have highly talented people with great instincts and insights.  Perfect insights without great team members is as useless as having the opposite.

 In the past few days, I’ve had a few conversations with some former colleagues.  A former global CPG brand president of Unilever, a CXO of Proctor & Gamble, the former US Chief Revenue Officer  of a top 3 alcoholic beverage conglomerate and the former CMO of one of the top 3 soft drink brands in the U.S. 

 I also asked them each the same question: 

“If existing systems could be enabled by AI to identify where optimal customers were, down to the ZIP codes, and exactly what conversion triggers were required for conversion and that conversion drove cost effective growth, what would keep you from doing it?”

 Here were their responses:

“Fear of being first.  But then later, fear of missing out and losing market share to the organization whose culture was more innovation and change oriented.  We’ve done this before and have paid high prices for those mistakes.”

“Fear of making a catastrophic, unknown mistake by relying on AI too heavily.  However, it we could control and manage that, nothing”

“The fear found in my teams who would tell me, “we already do that” when they in fact don’t and can’t”

 See a common theme?  Fear.  Fear of uncertainty or mistakes. 

 I asked each of them, their level of certainty of Artificial Intelligence MATERIALLY impacting CPG teams, either positive or negatively in the next 24 months.  Materially means changing at least 1/3 of how and what is done…how much harder or easier it gets and the associated results.  They all said the same thing: 

100%

 Seems like the old adage ‘change being the only constant’ is still true.  However, change, like luck, favors the prepared.

 The game is not over.  We are still in the early innings.  Find the appropriate place to start a small fire.  Try something.  Control it.  Be strategic.  Integrate it properly into your legacy systems.  Train to use it.  Then throw some gasoline on it.