Case Study: Using Buffers to Reduce Cancellation Rates and Offset Inaccurate Inventory

I’ve been interested in the relationship between better inventory data and fulfillment rates for some time now.

My assumption has always been that more accurate and frequent inventory data should lead to better fulfillment numbers. Not only does better inventory data mean less chance of cancellations from out-of-stocks but it also suggests an overall higher operational proficiency that bodes well for higher fill rates.

That’s why I was excited when early returns from our data warehouse this year flagged two footwear suppliers with similar levels of order processing but vastly different fulfillment rates and data exchange levels.

Even though a sample of two suppliers meant that the data was only anecdotal, it was still an opportunity to test my assumptions.

In this post I’ll share what I learned.

(Spoiler alert: at least in terms of these two suppliers, I was off the mark)

The Suppliers

• Supplier A
  1. Retail Category: Footwear
  2. Has processed 3400 orders in 2018 to date
  3. Data Match with Dsco: 92.2%
  4. Avg time to correct data: 6.3 hours
  5. Inventory Buffer: 0
  6. Cancellation rate: 8.8%

• Supplier B:
  1. Retail Category: Footwear
  2. Has processed 4200 orders in 2018 to date
  3. Data Match with Dsco: 89.6%
  4. Avg time to correct data: 12 hours
  5. Inventory Buffer: 10
  6. Cancellation rate: 0.14%

Safety Stock Buffers and Cancellation Rates

As you can see, for these two suppliers, safety stock buffers (or the lack thereof) have a huge effect on reducing cancellation rates. More real time and accurate data exchange, meanwhile, doesn’t appear to have a noticeable effect at all.

This wasn’t something I expected because I assumed better data would lead to greater inventory visibility and possibly even remove the need for safety stock buffers. I also didn’t anticipate safety stock buffers having such a large effect on fill rates.

But when you think about it, the results make sense.

Because Supplier A doesn’t use any buffers with their inventory counts they send “out of stock” messages only when their real inventory reaches 0. This creates a lot of opportunity for overselling products to multiple retailers, or for a retail partner to keep selling an out of stock product due to the lag time required for inventory data to wind its way through their system.

Supplier B, meanwhile, with its buffer of “10” sends “out of stock” messages even when their actual inventory level is 10 of a particular item. This means that even if they oversell a product to multiple retailers or a retailer partner continues to sell an “out of stock” item, they have a bit of runway to continue filling orders. In effect their safety stock buffers smooth over any data exchange problems they might have. This translates into much lower cancellation rates.

Does This Represent a General Pattern?

While this small sample size doesn’t take into account a lot of important variables, I think it might still be representative of supplier behavior in general.

Since our data research is continually expanding, however, in the near future I hope to offer analyses of much larger data sets to see whether or not this anecdotal evidence truly reflects industrywide trends.

In the meantime, it’s intriguing that a single factor such as the use of safety stock buffers had such a large effect on cancellation rates between these two suppliers.

By imposing a 10 item buffer, Supplier B had a cancellation rate that was 62 times better than that of Supplier A. This was true even though Supplier A had stronger inventory data exchange with its retail partner.

Further Questions

The problem with safety stock buffers, however, is that they represent unsold inventory. A 10 item buffer across all of a supplier’s inventory means absorbing a lot of opportunity costs to prevent cancellations.

This raises some interesting questions.

First, is there a lower inventory buffer threshold for various items that would allow a supplier to sell more inventory without significantly harming its cancellation rates? Would a buffer of 5, 6, or 8 items be almost as effective as a buffer of 10? What’s the optimal sweet spot between buffer levels and cancellation rates for various skus?

Second, does better quality and frequency of data (here represented by data match and time to correct values) reduce cancellation rates? If so, how much would a supplier have to improve these variables in order to lower their inventory buffer threshold and sell more product without harming their fill rates?

Third, do safety stock buffers lose their effectiveness during periods of high demand such as the holidays when inventory stocks are strained to the limit? In such situations does better inventory data exchange improve fill rates?

As I mentioned above, in future posts I’m hoping to use a larger sample size and much more data from Dsco’s platform to find the answers to these types questions.