How a CPG giant radically simplified its portfolio of SKUs using AI and tapped into value potential of $250M in just one year.
Clearing clutter
Clearing clutter isn’t just for closets - it’s for systems too. Simplifying SKUs needs focus, intention, and the right moment.
Not everything should stay.
The key is knowing what to keep, what to let go, and having the confidence to act. With clarity provided by data, teams can cut through the noise and focus on what truly drives the business forward.
AI doesn’t just automate - it creates space for smarter, and faster decisions.
One of the world’s largest CPG enterprises approached Fractal with a clear goal: simplify their SKU portfolio using AI.
AI-driven SKU delisting can save CPG organizations substantial time and costs, with up to 4X faster delisting and potential savings of $250M per year. This efficiency allows us focus on high-performing products, boosting sales and resource utilization.
Most CPG organizations are not aware of such potential opportunities with AI, though they could be hidden in plain sight.
How can you achieve it?
This is an illustrative story of how Fractal drove a radical SKU simplification program with AI and the lessons we learned from it.
What is SKU optimization?
An SKU (Stock Keeping Unit) is a unique identifier for a product variation based on size, weight, packaging, market, or even seasonal editions. Take a jar of chocolate spread, for example. It comes in multiple sizes (200g, 500g, 750g), formats (jars, glass cups, to-go packs), and regional or seasonal versions (Christmas or Halloween editions). Each of these variations is an SKU.
SKUs help CPG companies track and manage product variations. But over time, portfolios can become cluttered with low-performing SKUs. SKU optimization is a process of streamlining these portfolios - phasing out what doesn't sell and focusing on what does. For CPG organizations, this means reducing cost and complexity while focusing on growth - driving products.
It’s a practical lever to improve both profitability and revenue - often hidden in plain sight.
Large consumer product organizations create tens of thousands of SKUs every year. Over time, these easily add up to as many as 500,000 SKUs - each requiring manufacturing, distribution, and performance tracking.
Identifying which SKUs are underperforming isn’t easy. It’s like looking for a needle in a haystack.
And the real challenge?
Not just knowing which SKUs aren’t working – but understanding the cost of keeping them.
These SKUs quietly drain resources that could be redirected toward innovation, marketing, and products that truly meet consumer needs.
That’s where AI comes in.
How can AI help identify and reduce the long tail of SKUs...
...at the right time, in the right markets, and in the right way?
The AI problem
A CPG organization approached us, asking us to develop an AI model that could identify all poorly performing SKUs so they could delist them and save costs.
We implemented a model that considered multiple data points such as total sales, gross margin, monthly average turnover, SKU growth rate in the market, and consumer rankings.
The business team was thrilled to learn that AI pointed out poorly performing SKUs they hadn't noticed in plain sight. But...
When we followed up a few weeks after launch, we learned that no one in the organization was using the AI tool's recommendations to make decisions - a classic problem common with most AI tools: they stop at the proof-of-concept stage! We wanted to fix that.
A little quiz before we dive in.
Do you know how many new SKUs are introduced in a year by the topmost CPG organizations?
50 - 500
10,000 - 50,000
How many of these new SKUs mostly fail in the market and do not perform as expected?
25-50%
75-95%
How much loss do CPG organizations usually incur due to SKUs not delisted at the right time? (in USD)
100K - 100M
100M - 1B+
How long does it take a CPG organization to typically delist an SKU end-to-end from manufacturing and supply to distribution, inventory and sales?
3 - 6 months
12 - 24 months
One CPG organization was manufacturing 10,000+ SKUs in a year. The cost of maintaining SKUs at that volume is in the range of hundreds of millions. On an average, the organization was taking 24 months to find and delist a non-performing SKU.
In reality, most CPG organizations do not even focus enough on delisting.
Why?
Delisting SKUs is often complicated, manual, and heavily dependent on limited human judgment. Identifying underperformers can feel daunting.
To be or not to be
The problem of deciding what to delist.
Chris Anderson’s Long Tail Theory* suggests that the future of business lies in selling more niche products, not just bestsellers. CPG organizations have embraced this, constantly expanding their long tail through innovation. But while launching new SKUs is easy, knowing when to stop is much harder. Many struggle with decisions around when to delist an SKU - when to stop production, marketing, and distribution for SKUs that aren’t delivering.
For consumers, “more” doesn’t always mean “better.” The 80/20 rule reminds us that 80% of sales often come from just 20% of products.
Too much choice can overwhelm, turning a paradise of options into a paradox. That’s why, for manufacturers and retailers, the real question isn’t “what to launch” - it’s “what to kill.”
Who is to take charge?
The problem of deciding when and how to delist?
Consider this example: a fabric cleaner with mint and green tea launched successfully in Japan. Green tea, seen as a natural and gentle alternative, resonated well with local consumers. Encouraged by this, the company expanded the product to Switzerland.
But sales dropped after the first year. Despite clear signs, it took 24 months to delist the product - four times longer than it took to launch. The challenge wasn’t the data - it was the decision-making.
Delisting an SKU isn’t just a financial decision - it requires alignment across multiple functions. But with conflicting KPIs and siloed workflows, the process often gets delayed.
SKU delisting involves marketing, finance, sales, and supply chain
Often with conflicting goals. Siloed thinking makes timely decisions harder, even when the data is clear.
The human problem
The problem of deciding when and how to delist?
Teams faced several roadblocks when it came to delisting decisions:
AI surfaced an opportunity to delist over 100,000 SKUs.
The AI model datapoints synthesized across 200+ parameters to make the right recommendations.
We designed the decision process flow to enable accountability and action.
AI Value proposition
Enable people to find and delist poorly performing SKUs at the right time without incurring losses, and within the same financial cycle.
AI surfaced an opportunity to delist over 100,000 SKUs
This unlocked significant bottom-line efficiency across global markets. Until this visual emerged, no one had a full view of the potential. Teams were buried in fragmented data, scattered across spreadsheets and systems. What they needed was clear: a global, unified view of SKU performance - and the value it could unlock.
AI also identified the Crown Jewel SKUs that are performing well.
AI didn’t just flag underperformers - it also identified the best SKUs driving strong results. This gave marketing and business teams a clear focus to boost top-line growth.
By combining data across three lenses:
Customer (retailer performance)
(sales and likeability), and
Organization (costs and supply chain)
AI segmented the entire SKU portfolio into four actionable categories and turned a complex landscape into a clear roadmap.
The AI model datapoints synthesized across 200+ parameters to make the right recommendations.
The Garrett scoring system brought together KPIs from finance, marketing, sales, and business to assess which SKUs could be delisted for potential cost savings. More than just showing recommendations, the tool explained the "why" behind those recommendations, adding transparency and trust to every suggestion.
This gave decision-makers the clarity and confidence to act quickly and objectively.
We designed the decision process flow to enable accountability and action.
To support the client’s teams, we built an end-to-end SKU delisting process that offered clear visibility at every step. It enabled accountability across functions - with defined roles, timelines for stopping production, clearing inventory, and removing SKUs from shelves. Built-in reminders and prompts helped teams stay on track and complete the process on time.
We need to understand that just deploying an AI tool is not sufficient. We need to change and adapt our ways of working to use AI effectively.
Our learning
The hero isn’t AI. The heroes are the people making decisions with the support of AI.
AI is an enabler - what matters is giving teams the clarity, control, and confidence to act on its recommendations.
SKU simplification has two sides - the automation, and the human judgment. The real impact happens when both work together. It’s not just about the model. It’s about bringing data, algorithms, processes, and people in sync - to make smarter, faster decisions, together.
Value Potential
So, what changed for the company?
Teams faced several roadblocks when it came to delisting decisions:
$250M in potential savings: By delisting the right SKUs at the right time, the company freed up $250 million - cutting waste across the value chain and improving the bottom line.
Faster decisions, better momentum: What once took 24 months now took under 6 months. Teams moved faster, with clarity, ownership, and the confidence to act.
Focus where it counts: With a clear view of 100,000 underperforming SKUs and top performers to champion, the company could focus energy where it made the biggest difference.
And along the way, came unexpected wins.
Teams faced several roadblocks when it came to delisting decisions:
A culture shift: Teams began working together with greater alignment - making faster, data-driven decisions with shared ownership and clarity.
Stronger partner relationships: Retailers and suppliers could now focus on high-performing SKUs, improving their own margins and simplifying operations.
Confidence to innovate: With a clear path to delist underperformers quickly and safely, the company felt more confident to experiment and scale new ideas.
Illustration of value unlock for 1 SKU: the impact of AI + human-led SKU simplification
A variation of green tree fabric cleaner sells 188,000 units in year 1 - but drops to just 33,000 in year 2. It is sold in Switzerland.
AI flags it as one of 19,000 SKUs recommended for delisting. This time, the team acts faster - delisting it and saving $200,000.
Can you guess how long it took to delist the SKU this time around?
A variation of green tree fabric cleaner sells 188,000 units in year 1 - but drops to just 33,000 in year 2. It is sold in Switzerland.
AI flags it as one of 19,000 SKUs recommended for delisting. This time, the team acts faster - delisting it and saving $200,000.
Can you guess how long it took to delist the SKU this time around?
Your choices
3 Months
6 Months
12 Months
18 Months
24 Months
What next?
We’re exploring how AI can become even more intuitive and supportive through conversational interfaces - where teams can ask questions, get insights, and make decisions in real time through simple, natural interactions.
We’re also testing agentic AI - where the system doesn’t just recommend but can take low-risk actions like stopping production or clearing inventory, with human oversight. This frees up teams to focus on bigger, strategic calls to make - like product line rationalization.
There are other long tails - materials, suppliers, vendors, and retail partners - that add unnecessary complexity and cost. We’re exploring how AI can help rationalize these too, bringing the same clarity and impact we’ve seen with SKU portfolios.
And let us remind ourselves, the future isn’t just smarter - it’s more human.
Stop asking what AI can do. Start asking why.
AI is everywhere - but the real opportunity lies in how we choose to use it. It's time to focus on purpose, not just possibility. Let’s shift the conversation.
Drive value-led transformation. Move beyond incremental change to create meaningful impact.
Relook at how we build and use AI.
It is not about keeping human-in-the-loop. It is about giving human control where needed, enabling smarter decisions, faster action and stronger collaboration.
Because AI works best when it works with us.
Ready to unlock what truly matters with AI? Let’s lead the way, together.
How would you want to clean your house with Al?
