An object in possession seldom retains the same charm that it had in pursuit. – Pliny the Younger (62AD – 114AD)

As with all other industries and disciplines, retail supply chain planning is about to be rocked to its core by new AI technologies. Or maybe it’s going to fail spectacularly. The answer is probably somewhere in the middle.

Well, maybe.

We’re still at an early stage where multiple experts (people with far more knowledge about AI than I do) can have diametrically opposed views of where the technology is headed and they’re all looking at the same information.

Personally, I (along with nearly everyone else) avoided buying Amazon stock in 1996 when they were just an online bookstore hemhorraging cash every quarter. My point is that I’m not a futurist or an expert in AI, so I won’t be weighing in on whether or how it will ultimately impact how supply chains are planned in the years to come.

But what I think we can do right now is try to determine what the potential size of the prize actually is in terms of business results to a typical retailer.

First, we need to set a baseline of what retailers can achieve without the use of AI technology. Continuous time-phased planning has been around in manufacturing and distribution for quite awhile, but it’s still a relatively new concept in retail. And here I’m not just talking about forecasting and replenishment, but truly modeling the current and future policies and constraints of the supply chain from the retail store to the manufacturing plant over a long term time horizon – we call it Flowcasting.

It’s been well documented for at least the last 30 years that retailers who don’t plan in an integrated fashion from the store shelf back to their supplier base achieve in-stock levels in the 92-93% range. Those same studies estimate that every 3% improvement in in-stock leads to a 1% increase in top line sales.

Through our work with numerous retailers over the last 3 decades, we have proven that in-stock can be sustainably increased from the 92% range into the 98% range while simultaneously improving inventory turns by 20-40% (depending on item velocity and minimum merchandising requirements in the stores). This can be done by following some proven principles and strategies from manufacturing and distribution (tailored to apply to a retail business), namely:

• Create a forecast of true consumption for every item at every selling location
• Calculate long term replenishment plans for every item at every stocking location and link plans up the chain (i.e. from stores to DCs to suppliers)
• Keep the plans up-to-date with any known current or future external or internal impacts to demand or supply
• Make sure that everyone (including suppliers) always has an up-to-date version of the plan so everyone is working to a single set of numbers

This is all standard, old school DRP/MRP stuff. No magic. No AI. And while technology (especially the ability to process large data volumes) is enabling these results, the shift in mindset on the part of the retailers is as important, if not more so:

• If you know about something today that will happen in 3 months (e.g. a promotion, new product launch or store opening), get it into the plan today
• Don’t keep the plans a secret – let everyone see the plans and assumptions that go into making them so that everyone can be aligned on how to execute them
• Work by exception – let the plans tell you about potential problems that you need to start working on now before they become actual problems

This brings us back around to the actual size of the prize of AI in supply chain planning. The primary gap that needs to be closed in retail is the stubborn and pernicious 8% out-of-stock rate.

By applying process discipline, planning technology that enables it (without a single drop of AI) and connecting up-to-date plans up and down the supply chain, this gap can be reduced to 2%, while simultaneously making substantial improvements to inventory turns.

Even with highly advanced artificial intelligence that hasn’t even been developed yet, getting to 100% in-stock is a practical impossibility – with millions of item/locations in the retail supply chain, there is too much that can go wrong operationally outside the control of the planning function. So, at least as far as in-stock is concerned, the net incremental benefit above continuous planning is less than 2%.

Once the technology matures, there will undoubtedly be ancillary benefits of greater automation to planner productivity (hopefully not at the expense of accountability), but you don’t need sophisticated AI models to squeeze the vast majority of the juice from the lemon.

A lot can be achieved by just giving people the right processes, tools and a single plan that they can control and execute together.