Is Your DOM Forecasting Strategy on the Wrong Path? Part 1 of 2

0
150

Gartner recently suggested that most retailers don’t do a good job of forecasting demand at the place where it really happens – fulfillment. Retailers are forecasting demand where the sale happens, vs. where customers receive products. And that has got to change.

In this two-part series, you will learn:

  • Why DOM doesn’t cut it as a standalone demand forecasting strategy
  • Alternative strategies for handling demand volatility
  • How AI and machine learning can be applied to supply chain planning solutions to augment DOM

Part 1 of 2:

Distributed Order Management Systems: You’re Looking in the Wrong Place!

Looking into a Distributed Order Management (DOM) System? Maybe one is on your radar. Or maybe you’ve already implemented DOM because it promises perfect inventory-management homeostasis. Pure bliss.

Oh yeah? Really? Then why do organizations using DOM systems still find demand forecasting so freaking hard?

In a 2017 survey,

  • 46% of respondents cited order fulfillment optimization as a top-three concern
  • 65% said it was product availability, and
  • 56% pointed to inventory carrying costs

And in a 2019 supply chain study, three-quarters of respondents struggled most with increasingly complex patterns of demand volatility. Even though retailers are investing in retail technologies like DOM, these figures tell us it’s not enough.

Traditional DOM systems alone tell an incomplete picture of modern retail’s volatile ecosystem of “buy-wherever, receive-wherever” shopping. A demand forecast which ignores the growing list of fulfillment channels is a surefire ticket to demand distortion, poor fill rates and pissed-off customers.

Gartner Analyst Tom Enright argues that if retailers expect to have the right inventory at the right place at the right time, they must get uber-granular in aligning their demand forecast to how people shop. This will require a data audit trail that goes far beyond the capability of any human or DOM system can handle alone.

Minutia matters

Up until very recently, a high-level view of sales history was really all buyers needed to forecast demand. For example, their analysis might look something like this:

What’s the ratio between items purchased in store vs. those purchased online?

From there, they would position inventory at those two touch-points accordingly. And that was that. Today it’s much more complicated. For every sale, there is growing list of “if / then” hierarchies to drill down into, much deeper than where the sale happened. For example:

  • If bought in-store, will the item go home with the customer, or will the customer have it shipped? Will they collect it from a locker, or even from a third party?
  • If an item was bought online, will the customer choose to ship it to their home, their work, a vacation home on Mars, into another dimension???
  • What other creative shopping options will segment fulfillment even further in the coming years?

The buyer’s thought process now requires a deep-dive into the dirty details of each transaction. Minutia matters, big time. Best-in-class wholesalers and retailers that apply new technologies to manage detailed allocation and replenishment calculations at the point of fulfillment will land the highest fill rates at the lowest inventory cost.

Boom! [mic drop].

Ok, cool. So how do we do that?

Fulfillment-based forecasting strategies

You need a robust demand forecasting solution that can handle the two core pillars of fulfillment-based forecasting with uber-efficiency:

  • Demand visualization by fulfillment node. Can the system visualize demand by where the customer wants to take possession of the item? It should capture demand where the customer wants to take possession of the item, so you can allocate or replenish inventory to the proper node in your retail network. Right time, right place, lower operational cost.
  • Transaction-level analysis. Tom says real-time analytics and AI/machine learning techniques will be crucial to predicting demand tendencies at the most detailed customer transaction level, such as: Timing – Who purchased what, at what time of day? Location Which neighborhood or area has clear demand spikes? Product Affinity What co-products are often purchased together?

Choose a purpose-built supply chain planning solution that continuously analyzes these granular details and makes recommendations based on what it learns every day. If you can factor in all the juicy details of your customers’ shopping patterns and fulfillment preferences into your demand forecast, you will significantly improve inventory positioning – with some pretty significant perks:  lower costs, higher fill rates and happy, loyal customers.

Here’s an example. Let’s say someone buys a new dress online, but elects to pick it up in a store. Using real-time transaction-level analytics, the solution posts demand for that purchase to the exact store where it will be collected. Algorithms tell the system what it learns along the way to calculate future inventory positioning or replenishment recommendations based on the requirements for that SKU.

Where DOMs fall short

DOM systems are good at fulfillment, but lack the forecasting piece.

If you’re thinking about DOM and you aren’t incorporating the concept of fulfillment-based forecasting with real-time inventory visibility, you’re not getting the whole picture.

Planners use DOM systems to choose their allocations and replenishment for each store. When a demand variance happens between the different stores, the buyers can’t see that – so they miss sales and disappoint customers. They might have enough supply of a certain item at the DC, but can’t react quickly to volatile customer preferences (such as ship-to-home, BOPIS, pick-up lockers, etc.) without the excessive cost of physically moving it around.

Fulfillment-based forecasting systems – in contrast to DOM – not only show what purchases consumers will make in the future, but also how they will shop and, subsequently, how you should allocate inventory for replenishment. Check out Tom’s example of how this works to improve inventory positioning:


Source: Gartner® “The Next Step in Improving DOM – Fulfillment Forecasting”. April 5, 2018, Tom Enright

If an online retailer – say The Gap – can use machine learning analytics and transaction-level forecasting to back up the line a bit and understand which fulfillment mode that demand is attributed to, they can ensure the inventory is there and waiting for customers. In a much more profitable way.

The payoff is huge

There’s a good reason why cloud-native supply chain solutions (and more recently, AI) are quickly bubbling up to the top of the technology wish-list. In a 2018 retail study with RIS News and Gartner on top retail technology priorities,

  • Over two-thirds (67 percent) of respondents have already implemented, started to implement, or will in the next 12 months implement a major tech upgrade in the area of real-time inventory visibility
  • AI rocketed from the bottom of the list in 2017 to number-five in 2018

And this year, we suspect more industry experts like Tom Enright will continue to explore the value of using fulfillment-based forecasting systems – and especially those that incorporate AI and newer cloud-based innovation – like Netsuite – into their platforms.

These real-time AI analytics-driven forecasting solutions will be the key to optimizing inventory positioning across your fulfillment channels – ultimately reducing the dreaded consequences that routinely run off customers, such as depleting store inventory to fill an online order.

The benefits are game-changing over DOM:

  • Visibility into both high- and low-level demand
  • Ability to react better to demand volatility
  • Fewer stockouts, overstocks and costly uncertainties
  • CX improvement: 3-5% increase in service levels without increasing (or even decreasing) inventory costs

In Part 2 of this series, we’ll go deeper into fulfillment-based forecasting systems, plus explore potential applications for AI and machine learning within these products.

Update 11/5/2019: There’s an interesting LinkedIn discussion going on this article. Please visit and express your thoughts.