Dynamic Dual-Period Multi-Criteria Inventory Decision Framework (DPMC-IDF): An Empirical Comparison with ABC and ABC-XYZ

Imagine you are the captain of a massive cargo ship, but instead of sailing the ocean, you are navigating a sea of thousands of products. Your job is to decide which items to pack in the most secure, easy-to-reach spots and which ones can be tucked away in the deep hold. This is the world of inventory management, a crucial part of keeping stores stocked and businesses profitable. For decades, captains have relied on two old, trusted maps: ABC analysis and XYZ analysis.

Think of ABC as a map that only cares about how much money an item makes in a whole year. It's like ranking your friends only by how much pizza they ate in a year, ignoring whether they ate it all in one night or spread it out over a month. XYZ is a map that looks at how "bumpy" the demand is. It uses a math tool called the "coefficient of variation" to guess if an item's sales are steady like a calm lake or wild like a stormy sea. The problem is, these old maps are static. They look at the entire year as one big, blurry photo. If a sudden storm hits in November, the map gets confused and thinks the whole year was stormy, leading captains to make bad decisions about what to stock.

This paper introduces a new, dynamic compass called the Dynamic Dual-Period Multi-Criteria Inventory Decision Framework (DPMC-IDF). Instead of looking at the whole year at once, this new tool splits the year into two halves (like two semesters) and watches how items move between them. It doesn't just ask "how much money did you make?" but "how did your share of the pie change?" and "did your sales jump or crash?" By doing this, it hopes to spot trends that the old maps miss, like a product that is quietly growing or one that had a one-time spike but is actually stable. The author tested this new compass against the old ones to see if it helps captains avoid running out of stock or wasting money on items nobody wants.

The Old Maps vs. The New Compass

The paper argues that the traditional ABC-XYZ method has a major blind spot: it assumes the world is "stationary," meaning it doesn't change much over time. The author shows that this is often false. Imagine a product that sells 10 units every month for eleven months, but then sells 100 units in the twelfth month because of a holiday sale. The old XYZ map looks at the whole year, sees the huge jump, and calculates a high "volatility" score. It labels the product as "erratic" (a Z-item), telling managers to treat it like a wild card. But in reality, the product was calm for 11 months! The old map got tricked by that single month, just like a weather forecast that says "it's always stormy" because of one hurricane.

Similarly, the old ABC map might rank a product as "High Value" just because it sold a lot in total, even if its sales are dropping every month. It's like ranking a runner as the "fastest" because they ran a marathon once, ignoring that they haven't run since.

The New Tool: Splitting Time and Watching the Shift

The DPMC-IDF framework fixes this by doing three clever things:

1. Splitting the Horizon: Instead of looking at the whole year, it cuts the year in half (Period 1 and Period 2). It checks how an item performed in the first half and then again in the second half. This is like checking your grades in the first semester and the second semester separately, rather than just averaging them for the whole year.
2. Percentile Ranking: Instead of using a simple average (which can be skewed by one crazy number), it ranks items against each other. It asks, "Did this item do better or worse than 75% of the other items?" This makes the ranking more robust against outliers.
3. The Volatility Index (VI): This is the secret sauce. It calculates exactly how much an item's share of the total sales changed from the first half to the second half. It gives a three-letter code (like LHV or HMM) that tells a story.
   • LHV might mean an item started as "Low" share, became "High" share, and is "Volatile" (growing fast but changing a lot).
   • HMM might mean an item was "High" share, stayed "Medium," and is "Moderate" (a steady but slightly declining star).

The Simulation: A Tale of Three Products

To test their idea, the author created a tiny, simulated world with just three products:

• Product A: Steady and reliable, but slowly losing its share of the market to a competitor.
• Product B: A seasonal product that starts slow but explodes in popularity in the second half of the year.
• Product C: A stable product that had one weird, huge spike in sales for one month, then went back to normal.

The Old Maps' Mistake:

• The old XYZ map labeled both Product B (the growing star) and Product C (the one-time spike) as "Erratic Z-items" because both had high volatility scores. It couldn't tell the difference between a healthy growth trend and a one-time glitch.
• The old ABC map put Product A and Product B in the same "High Value" bucket, ignoring that Product A was fading while Product B was rising.

The New Compass' Success:

• DPMC-IDF correctly identified Product B as LHV (Low-to-High, Volatile), recognizing it as a growing trend that needs investment.
• It labeled Product C as MLV (Medium-to-Low, Volatile), realizing the spike was temporary and the item was actually declining, so it shouldn't be treated as a permanent high-priority item.
• It saw Product A as HMM, noticing that even though it was steady, it was losing its competitive edge.

In this simulated test, the new framework made zero mistakes in classifying the products, while the old methods made several errors. The author calculated a "Misclassification Risk Reduction Index" (MRRI) and found the new tool reduced the risk of error by 100% compared to the old methods in this specific scenario.

The Real-World Test: StellarMart Store

However, the paper doesn't stop at simulations. The author took their new compass to a real dataset from a store called StellarMart, which had 5,000 transactions across four regions (East, West, South, North) and 25 products. They used sales data from 2023 to predict what would happen in 2024.

Here, the story gets a bit more nuanced. The new tool didn't win every single time.

• In the South region, DPMC-IDF was a hero, cutting the error rate almost in half compared to the old methods.
• But in the North and West regions, it actually performed slightly worse than the old ABC-XYZ map.
• When they looked at all four regions combined (100 items total), the new tool reduced the simple error rate by about 5.5% compared to the combined ABC-XYZ map.

The author found that the new tool struggles when a product's trend reverses after the two periods it watches. For example, if a product declined in the second half of 2023 but bounced back in 2024, the new tool (which relies heavily on the second half) might wrongly predict it would keep declining. The old tools, which averaged the whole year, sometimes got lucky by smoothing out these reversals.

What This Means

The paper concludes that DPMC-IDF is a powerful new tool, but it's not a magic wand that fixes everything. It shines when products are undergoing structural changes—like a steady rise or a steady fall—because it can see the direction of the trend. It helps managers spot growing stars and fading lights that the old maps miss.

However, the author admits that if a product's behavior is unpredictable or reverses quickly, the new tool might stumble. They suggest that in the future, the tool could be improved by looking at more than just two periods or by combining it with other decision-making methods.

In short, the DPMC-IDF framework offers a more dynamic, "time-aware" way to manage inventory. It suggests that by splitting the year and watching how items move, businesses can make smarter choices about what to stock. But like any new technology, it works best in specific conditions and needs to be used with a clear understanding of its limits. The author's work suggests that while the old maps are still useful, a new, more flexible compass is needed for the complex, changing world of modern supply chains.