In-Store Customer Decision Making

In-Store Customer Decision Making
March 2, 2022 Omnibus

First published Mar 2, 2022. Last update: May 31, 2023.

Every successful retailer plays a game with the customer decision-making process. Shoppers’ choice. Our in-store behavior: how we search for → choose → buy products in a store environment.

It’s a key game in retailing where the goal of the retailer is to meet shopper expectations and solve his / her pain points. Maybe cross-sell and up-sell a little.

The rewards of playing the game well are what retailing success is all about:

  • increased customer satisfaction
  • rising value of the shopping basket
  • increased store profitability
  • higher stock turnover
  • revenue growth

The basic elements of the game are simple. We have:

a) a shopper need (“something for breakfast”)

b) an appropriate offer that meets the expectations (“milk”, “cereals”, “yogurt”, “toast”, “eggs”), maybe up-sells a little (“milk & cereals combo”)

c) a place in the physical store where retailers provide the solution (“shelf space”).

Then comes complexity.

Add 1000s of customers with many different shopping needs and missions, expand the offer into an assortment of 60 product categories, and 20.000 different products, and limit a physical space to say 1000 m2 … For a normal grocery supermarket.

And more. Shopper expectations change while the flow of products from different categories rapidly expands. And there is limited space in physical retailing. The rising pressure of time, space, and costs.

Also: today milk isn’t “milk” anymore – it might be dairy-free no-lactose almond milk! Sorry, we forgot, a blend of almond milk, that is.


In reality, meeting customer expectations is a highly complex retailing game.

To make the process manageable, to decrease assortment confusion, and raise the level of tactical proficiency,

retailers need the right strategic tool.

A solid data model like Customer Decision Tree (CDT) fits the complexity of the task. It is kind of a silent worker that links all the elements described.

On the one hand, it is a hierarchical model of a shoppers’ decision-making process while they shop.

It follows the shopper’s needs by breaking them down into a series of steps they go through when making a purchase.

Simple example: A shopping motive could be split into decision attributes top-down.

Motive: “Something for breakfast” → Category: Milk → Type: Low Fat → Brand: FinesseMilk → Package: transportable by foot → Promotion: monthly catalog → Final Product FinesseMilk 0,5 lit.

On the other hand, CDT is a strategic tool that assigns importance to different elements of the offer.

If you prepare a CDT for “Something for breakfast” you might analyze the past sales, market shares, and trends, but also recognize special importance for your strategy – eg. you might give more importance to organic products.


For a retailer, CDT is thus not only a tool for understanding shopper buying motives but also a strategical tool that serves as a guiding line for:

  • visual representations of shelf space – planograms, and assigning product positioning on shelves (micro space planning)
  • design of efficient store layouts (macro space planning)
  • assortment decisions and pricing strategies.

Shall we add: Customer decision tree also plays a central role in Omnibus’s store space management framework #levers for successful retail stores.


  1. Customer decision tree: what it is, definition, retail use, benefits, mechanics of the tool.
    Also customer decision tree design principles.
  2. Why do machine learning algorithms open up striking new capacities of customer decision trees
  3. Fresh perspectives on a “math trap”. Often overlooked, but crucial in so many data science projects. Past data – even machine learning insights – should be complemented with enough creative hypotheses, “human input”. The customer decision tree serves as a very good example of this principle.
  4. How to leverage successful CDT implementation for 2-digit store profitability and revenue growth



We’ll start studying the mechanics of customer decision tree with a very basic example.

Our shopper has the simplest of all goals → to buy a little milk. No bread needed, just milk.

It’s also one of the most regular shopping tasks. Our recent shopping basket analysis shows that 10% of all grocery tickets in a Central European city market include milk (another research proves that shoppers buy milk from the nearest convenience store 9.4 times per month).

So let’s follow our shopper – walking down the street with a recycled bag.

Entering the nearest grocery store, then approaching the shelves.

A little decision needs to be made. Long-last or fresh milk, right? Ah, and another decision. Will it be whole, reduced-fat, low-fat, or fat-free type? How about lactose? Organic or not? Is it cow’s milk, rice, plant, or nut-based? And size? And origin? And brand? And price? And then the expiry date?

Simple buying task causes an unexpectedly huge cognitive load. Especially in digital times when shoppers’ demand gets increasingly fragmented. While the older generation would still embrace the regular cow milk, the younger might lean in a completely different direction – towards the craft variations of coconut milk. Subchoices are galore. We might say shopper in digital times is thrown into a labyrinth of choice.


On the other side of the shelf, someone is pulling the hair out. The retailers.

Usually heavily armed with computers, sales data, specialized software, and market research, they try to decipher shoppers’ decision-making process.

They tend to organize products and shelves in such a way that shoppers can find their beloved milk product (finding means selling). And maybe – just maybe – even to divert them a little away from the intended purchase. Direction: higher margin or bigger size or – not forbidden! – local, organic, sustainable.

To make sense of a huge assortment mess, the main tool in a clever retailer’s hands is a customer decision tree.

The customer decision trees are used to map out the key decision factors that influence customers’ in-store behavior.

It breaks down product attributes like

  • category
  • sub-category
  • brand
  • price
  • package
  • size
  • origin
  • producer
  • types, subtypes, ingredients, etc

and combines them in the tree hierarchy.

Below you can see how attributes are organized hierarchically. Each attribute serves as a node – s single point which then branches into directions.

For example, the price level is a root node – a point in a chart – which branches or splits into four directions, leaf nodes – premium, mainstream, low-budget, and entry point.

The tree thus tries to follow the shoppers’ decision-making process and provides criteria for positioning products on shelves.

Sample Customer Decision Tree with 4 hierarchical levels (c) Omnibus – Ljubljana


Back to our milk shopper. One possible captured customer decision path from the top down goes like this:

“Something for breakfast”  → Milk Low → Dairy Free → Low Fat → Brand → Flavour: Hazelnut → Package: transportable by foot → Promotion → Final Product X 950g.

Of course, the hierarchically organized path stages are used just for a better illustration of the buying process. The shopper’s brain makes the decisions instantly. A choice between 2 products takes 0,2 – 1,5 seconds, though it includes different regions of the brain: visual cortex, prefrontal cortex, amygdala, nucleus accumbens, and hippocampus.

But the choice time significantly increases when given additional product options!

Returning to the content of our shopper’s decision path we might find a surprise. Milk searched is technically not milk but a dairy-free product.

Even at this stage, we see the dilemma:

  • some retailers recognize this substitutability and reorganize categories – they put non-dairy products close to dairy products.
  • other retailers still say: “No! Milk is (cow) milk and almond drink is something completely different.” They might place milk in the refrigerators and on the shelves nearby, while almond drinks could be found far away on the shelves with “healthy products”.


Here enter the quantifiable sales data that give weight to the importance of each node. The weight might be a volume or value share or some other performance indicator, possibly a combination.

With the addition of other data sources – like income range, use of loyalty card, promotion sensitivity, etc. – the data model could become increasingly smart but at the same time more complex. We’ve done live testing and included the importance of nodes as one of the key inputs of our methodology #levers for successful stores.

It’s a decisively important input for the solid customer-decision-making model but elaboration would exceed the purpose of the article. If you’re interested in how #levers could inform your store’s success, don’t hesitate to contact me.

But for the clarity of explanation, let’s just say that a data model “translates” sales and behavioral data into space values. Sales to space.

Thus, these customer decision trees serve as a key input for creating planograms – a visual representation of store and category layouts.

The retailers add a calculated share of shelf space – this finally brings sales and space share into relation and pushes the decision tree further toward implementation.

When planograms and store layouts are implemented, the customer decision tree finally brings together all the participants in the buying process.

  • For the retailer, they help him organize the space, and increase the store’s profitability. There are also other benefits, like a significant decrease in the amount of time store staff needs for shelf replenishment.
  • For the shopper, it eases up the choice. It helps her navigate from top to bottom until she finally finds the right supplier’s milk product. In particular case non-dairy milk products.
  • The suppliers also get their share of shelf space according to clear criteria.


As mentioned, CDT links shoppers’ needs to the assigned store space.

It’s immensely useful for clarifying a picture of assortment, the importance of the categories, and playing a tactically and strategically good retailing game.

However: CDT is hidden behind planograms, store layouts, and assortment decisions. That’s why the benefits of the lever might still be overlooked. Talking from experience, it’s not uncommon to find a national retailer who struggles with category management – and space management – very often because the tool of CDT is not properly used or not used at all!

How a good data model like CDT might help with the success of a retailer still needs some reinforcement. So let’s try to list possible benefits which properly done CDT brings to a retailer:

  1. Increased sales. By better understanding customer behavior, retailers can optimize the key variables: assortment, pricing, and marketing messages.
  2. Improved customer satisfaction. By tailoring the shopping experience to meet the unique needs and preferences of each customer, retailers can improve customer satisfaction and loyalty.
  3. More effective marketing messages. By targeting the right customers with the right messages, retailers can improve the effectiveness of their marketing campaigns and reduce waste.
  4. Reduced costs by greater operational efficiency. By optimizing store layout, product placement, and staffing levels based on customer behavior, retailers can improve operational efficiency and reduce costs.
  5. Better inventory management. As store layouts follow the customer needs, the ratios between sales and assigned space for brands and categories increase inventory turnover and at the same time lowers the average inventory needed.
  6. More accurate forecasting. By using data from the customer decision tree to inform forecasting models, retailers can make more accurate predictions about future demand and adjust their strategies accordingly.


I can best answer from my own experiences.

For years, I’ve been designing and implementing customer data models to support space management decisions. From macro-planning (described in the article Succesful Retail Store Spaces) to micro-planning (more in the article Planograms Meet Inventory Management).

The referential project is described more thoroughly in a separate article Efficient Store Layouts. For a project, we analyzed the current store space efficiency and then prepared a list of proposals that served as a framework for the overall store layout overhaul.

Repositioning of the categories in a supermarket (1500 m2) resulted in the following results:
– inventory stock decreased by 8%,
– store profitability increased by 5%,
– space profitability increased by 20%.

Based on this, we further developed the hypotheses and implemented improved store space strategies in another supermarket.

The result?

increased sales by 25% every year making the store in Ljubljana (highly retail competitive Slovenian capital with 300.000 inhabitants) nr. the 1 growing store out of more than 100 chain supermarkets.

While the growth gained is a result of a set of many different variables, designing and implementing CDT was a crucial step of the new strategy and pilot stores. We used it for macroplanning as well as for informing the cross-category positioning, and in-store promotions.


Customer decision trees are perfectly suited for machine learning algorithms.

Large volumes of data could now be automatically processed – almost live. Transaction data, demographic information, and research data are blended to identify the patterns and also trends in customer behavior. Once the algorithms are informed, the speed, power, and accuracy exceed the capabilities of even a highly trained person.

If old customer decision trees were black & white, the machine learning boost might be compared to color.

A perfect tool!?

As often happens, the idyllic view hides a trap.

The simple look of a tree often entices corporate practitioners – not to mention startup wizards creating all-purpose magical algorithms! – into believing that the decision-making problem could be solved by some winning mathematical equation.

I’ve heard their compelling arguments often oozing out over-confidence:

“How? Easy. We choose the decision criteria, gather the past data, connect different data sources, and set the genius rules. More sales more space, fewer sales less space. We turn machine learning algorithms with fancy charts on and … Iterations? Yes, might be some iterations … and voila! the space management problem of the store is automatically solved. Once and for all. Time to move over to other, more serious problems!”

But that’s a corporate illusion.

Transforming past data and growth opportunities into the store layout IS the ultimate challenge for a physical retailer. 

It combines all of the retail variables at once. Assortment, price, presentation, and place – all come together here. No magical formula will solve this once and for all. Sorry.


Good mathematics helps. Might be even a necessary condition for success.

Bad news: good mathematics is not enough! It should be complemented by other inputs. Even by some gut feelings. By understanding how customers buy, how they make decisions, and feeling their pulse. Also taking trends and future customers into the picture. By relentlessly assessing our market position and the value we bring to customers. And more: decisions have to be regularly checked, refreshed, and improved.

All those inputs could get a “weight” = importance, and thus be translated into numerical values. #levers do exactly this. The importance is adjustable, according to the needs of the retailer, but also leans on established best practices.

Machine learning can come only after we find a working solution.

There is also good news: once a valid CDT is set up and implemented the following iterations will take considerably less time. Plus: with a feedback loop and further improvements – expert input also – CDT will very likely accelerate positive results.


The answer to the question hides in human psychology.

All CDT approaches lean on the logical and hierarchical division of shoppers’ needs.

Then the products are hierarchically put in different “branches” (or nodes) following the attributes that represent the needs. You could see how this works in the case of frequently bought milk products in grocery stores.

But this is not sufficient.

We need to add something beyond logic. Something often overlooked.

Basic stuff, Watson.

Our human nature. Shoppers tend to find the easiest ways of fulfilling their shopping intentions, true. But they are not just calculating machines solving problems by attributes only.

We know it from experience – but economic and business theories overlooked it for a long time. As Dan Ariely nicely put it, shoppers are predictably irrational. New theories are successfully challenging the rational paradigm, and our in-store research has constantly proved the need to understand the emotional side of in-store behavior. More on the topic you can find in the article How To Engage Customers In Retail Stores.

Even in a place like a store, we as human beings are longing for a little surprise, a little diversion from everyday routine.

The good retailers thus successfully:

a) meet the shopper’s expectations by efficiently solving the pain points


b) in an unobtrusive way divert shopper expectations with a valuable offer of something the shoppers didn’t plan to buy (more emotional, surprise that diverts from logic)

When these retail surprises are backed by proper strategies, and successfully executed, the customers are satisfied. Even delighted. And, the store thrives.

Only a combination of logic following the pain points and surprise elements will satisfy the shoppers and convince them to award a retailer with many happy returns.


The concept of a customer decision tree gained momentum after the publishing of Peter & Olson’s book “Understanding Consumer Behavior” (1993).

Their key insight is that the customer decision tree represents the actual decision-making process when making a purchase.

But reality demands that we take this approach with a little salt.

Here are four challenges for designers of customer decision trees.

1) The decision-making is neither hierarchical nor purely logical & rational. The consequence: the above decision path doesn’t fit the process of shopping. Some neuro research shows that depending on the category up to 90% of our purchasing decisions are irrational.

Meaning: That even a regularly bought product such as milk is bought with a blend of your senses. You might get attracted by the colors or maybe the memory of the mountain valley once visited vividly overwhelms your mind when approaching the shelf.

2) As decision trees are prepared on a category level they fail to recognize our real shopping goals. Namely, our shopping basket is usually a pattern of products from different categories – only together do they represent a fulfilled shopping mission.

Meaning: Organic milk shoppers might like to add other organic products to their shopping basket, eg, organic cheese, organic butter, and even organic snack. It probably isn’t the high-turnover products, but if you as a retailer don’t provide them – the disappointment might permanently push the shopper towards a competitor.

3) Shopper decision-making process is not a system with one solution. On the contrary, it is heavily influenced by a context – place, time, and the goal of the particular mission.

See how shoppers’ attention varies regarding his / her goals:

  • Say you just want a quick breakfast solution on the way to the office: you probably wouldn’t be at least interested in heavy multi-packs of milk. The situation would completely change a few hours later when you return home and need to fill the gaps in your home wardrobe. The decision tree is different from mission to mission.
  • If you go to a specialty dairy shop, you’d be less sensitive to prices, but you’d also expect more diversity in terms of local variations. The decision tree is different from retailer to retailer.
  • Another important factor is familiarity with the store. In familiar places, people tend to skip the unnecessary options and buy habitually (auto-pilot buying), while in unknown places the decision-making process is more time-consuming.
  • One more context that heavily influences our behavior: is it a category you buy frequently (like milk 9.3 times per month in your nearest local store) or a category you buy every 5 years (like refrigerator)?

4) The case of buying milk showed us how fragmented the needs of the customer in the digital age are. By logically hierarchically splitting the branching, we can easily get to the tiny segments that the retailer can no longer viably fulfill.

A fresh article published by a consultant and shopper marketing expert Mike Anthony can further substantiate the problems of the CDT as a merchandising tool. In his well-thought-out practitioner’s view, he points out some of the key issues.

The key takeaways from his article could be:

  • There is no single ‘right way to layout products on the shelf
  • Much of shopper’s decisions are made subconsciously
  • Algorithm or Artificial Intelligence that exactly knows the right solution for a layout is highly improbable
  • “Any Shopper Decision Tree is going to be ‘wrong’ for some shoppers”
  • There is a difference between need and find
  • Shopper Decision Trees map the past, not the future
  • Shopper Decision Trees is not a guaranteed recipe for success


Does this mean we shall abandon CDT?

Our experience answers: No!

CDT remains the best framework we know to clarify a very complex assortment and space management picture. It helps manage the complex system and make decisions that connect retailers with shopper expectations. It brings so many tangible results, we can claim it is a central tool for (almighty in the retail universe) successful store layouts. 

Don’t abandon CDT. Rather upgrade & execute properly!

So how do we then properly grasp the shopper decision-making process? How do we fill CDTs with the right inputs to success in retail stores?

We’ve been studying the shopping process for years. Also learned from first-hand experiences. Many decision trees later, here is a solution that works for us – and our partners. A framework called “levers for successful store spaces from tip to toe”.

The methodology consists of three big stages:


First a disclaimer: the word “strategy” itself is often something you approach with contempt. In our parts of the world (Slovenia, “Adriatic region” …), even for seasoned corporate people strategy means a set of bland words on paper no one cares about.

Let’s avoid the pitfall – the described is a bad strategy: long on goals, short on policy and action.

But good strategy is the opposite – it leads organizations across business challenges. It fundamentally provides the difference that matters for customers.

Here are some examples:

“Organic products at mainstream prices.” (one of the discounter’s strategies like Aldi / Hofer)

“Buy fast, eat slow” (Eataly)

“Furniture is not an investment, but a lifestyle” (Ikea)

Store and category layouts should mirror and grasp your store strategy.  In this perspective, each category layout should be a piece of a store layout necklace.

Even lower levels of product management should be aligned with strategy. Therefore, setting up a particular category must start from

  • well-defined store / retail strategy
  • key competitive advantages
  • identified growth opportunities


This stage links retail sales data and customer behavior data (segmentation).

We analyze the data, choose the criteria, define the shoppers’ decision path, and create data insights that will lead to actionable findings.

A mathematically sound customer decision tree (CDT) is an essential result of this stage.


Synthesis follows the analysis.

Here we add the tactical part to the customer decision tree.

Here is a glimpse into the “how” of tactical improvement:

  • The relative importance of particular nodes should reflect the strategy from stage 1. If you’re trying to position yourself as the number one organic provider, emphasize organic products, etc.
  • Balance past data and trendlines  with informed hypotheses about future developments
  • Link inputs to objectives by informed space hypotheses. Those are ranked according to the expected performance and feasibility. After prioritization, we get to the actionable solution.

Mathematically and logically correct customer decision trees from the previous step should now also be made more shoppable. How? By combining navigation logic, sales data, AND emotional value for core users. One very successful practice from the past: we created blocks of confectionery snacks by blending brands & colors (the Swiss cow has its recognizable color! and Ferrero and …). Essential is to feel the pulse of the customer – which is more than just the logically correct process of step 2.

As a result, we get tactically sound CDT that reflects shopper expectations and clearly emphasizes key competitive advantage. With the addition of shoppability, it is finally ripe for execution.  


Why do I call for fresh perspectives on something so data-heavy as customer decision trees?

Because humans – and shoppers are human beings – don’t just solve problems as calculating machines. Buying problems included. Now and then we want to be surprised and find something new, something which exceeds our expectations and diverts us from everyday routine. In other articles, I prove how former discounters like Lidl and Aldi (not to mention Ikea) use this principle for their store’s success.

The content of this article is completely free even though it grew up from experience in countless commercial projects. Why? As in other Omnibus endeavors, I believe only a good balance between data analysis and creativity can bring us closer to the demands of the future.

Not Artificial Intelligence but creativity can lead us away from the dry, machine-like, boring physical experiences of many of today’s stores.

Only a holistic approach that brings together organic use of data and conscious creativity will serve our Nature, Environment, People, and yes, even, Business purposes.

Profit shouldn’t be seen as a purpose, but as a result, the reward for the well-played game.

Refreshment of a CDT with a human perspective might provide us with a tool that helps people build better stores for satisfying other people.


  1. Proper management of shopper options is instrumental in improving customer experience and sales success.
  2. A customer decision tree is a central analytical and tactical tool for the design of efficient & successful store layouts.
  3. A data model that links sales data with spatial data is a very important intermediate step of retail space management.
  4. For better “shoppability” – stimulation of purchases – and efficiency, additional tactical decisions should be added. By tactical adjustment of the importance of particular elements retailers can easier achieve the goals of their sales and brand strategy.
  5. Based on CDT and identification of key factors that influence customer purchase decisions, businesses could develop targeted product assortment better aligned with customer preferences. That has a direct positive impact on the value of the shopping basket.
  6. By using machine learning, retailers can develop sophisticated decision trees that take into account a wider range of variables, enable quick rearrangement of hierarchies, and provide more accurate predictions.
  7. Algorithms behind CDT should be adjusted regarding the retailer’s brand. Store space organizing criteria should also clearly link to the key competitive advantage of a retailer. This can serve as a strong leverage for market success.

All of the above is collected in our shelf & store layout methodology called #leverages for retail success.

Our workshop serves as a primary facility to get through the stages of strategy → analysis → synthesis → successful layout. Combined with analytical projects, it’s a proven methodology with results supported by different success cases.

CDT, properly done, is a powerful leverage that can provide your retail store thrive also during digital times.


The implementation of CDT depends on different variables.

Some retailers lack coherent product classification linked to point of sales data. Or the product classification serves technical purposes only. Others are weak in data point integration. Smaller chains and independent stores usually lack sophisticated data models. Customer segmentation is based more on the intuition of store managers. No dedicated analysts etc. This is something that should be addressed.

On the other hand, there are rising retail chains that developed fast, and have some organizational and technological resources available, but are now facing a new stage of their development. If data from different sources are available, but not structured properly, we can develop an alternative structure – using tags – specially prepared for store profitability cases.

It’s also extremely important how the technology is used for retail space management. Using space management software might serve as a very good springboard for tactical improvements. The organization of retail space management varies a lot, as does the executive power. And there are other considerations like size and type of the retail organization. There are many possible situations, for rural stores, specialty stores, and independent chains.

Let’s not forget – of course – business objectives.

Based on the above, there are different starting points for the implementation of CDT.

During the years of retail consulting and store layout project experience, I’ve developed procedures that support tailored solutions for different situations.

The common thing is: regardless of the type, the retail chains could all leverage a customer data model (CDT) for store organization and additional revenue growth.

Anyway. It’s never too late to start with customer segmentation. I’d recommend even the most basic CDT implementations. After all, retail is at its core about satisfying customers’ needs.


The steps I recommend are the following:

  1. Fill in the basic info in the contact form here.
  2. You’ll get a 10-question initial survey that will help us determine your basic needs.
  3. We’ll arrange a 30-minute online discussion where we’ll discuss your current situation and expectations in a structured way.
  4. After the discussion, I’ll prepare

    1-2 page assessment paper with the recommendations for further steps towards CDT implementation.

    1 quick win proposal for revenue growth included.

It’s 100% free. Believe me, I’m curious enough about your situation so this will both gave us a picture of the possible solution.

SPECIAL BONUS: If you’re dealing with sustainable brands, healthy products, or working on solutions that benefit nature, I’ll add:

a document with 5 retail store space strategies that will increase your profitability


2 hours of free consulting!

I won’t brag too widely, but I guess it’s worth getting this assessment report from someone with field experience who created pilot projects for companies like Coca-Cola and 9.4 bn yearly revenue national retailer nr. 1.

You can study the assessment paper and take your time to decide if you want to proceed with the next steps direction of revenue growth with me or not.

I’d be really glad to help you understand your shopper needs, connect closer with your customers, and leverage their satisfaction for your store’s profitability!



Any feedback – liking, sharing, quoting, commenting on – is much appreciated!


Would you like to grow your business and implement a customer behavior model powered by machine learning in your project?

High-quality Business Case for a pilot implementation of a customer behavior model will:

1) articulate goals and objectives based on key stakeholders’ inputs

2) evaluate feasibility, risks, and benefits, and lead you to an informed decision about the project

3) seriously improve ROI by providing a roadmap, setting key metrics, and optionally putting our #levers methodology at work.

I can help you in each stage of the Business Case. From understanding the business problem, and gathering relevant inputs and feedback, to conducting key analyses.

We always start with a free discussion of your needs and expectations and the initial sketch of the project roadmap.

If you’re interested, please contact me by filling out this simple form.


1 Comment

  1. Rizka Firdhayanti 2 years ago

    Thanks for information

Leave a reply

Your email address will not be published.