- AI and Machine Learning
- Analytics & BI/Benchmarking
- Blue Ridge Product Toolbox
- BLUEPRINT 2021
- Career Diaries
- Change Management & Adoption
- Cost Control & Reduction
- Data Science
- Discrete Manufacturing
- Events & Conferences
- Resiliency & Disruption
- Hero Stories
- Integrated Business Planning
- Inventory Optimization
- Multi-Echelon Inventory Optimization
- Net Pricing
- Planning + Pricing Solutions
- Price Optimization & Management
- Pricing Advanced Analytics
- Promotion Pricing
- Sales & Operations Planning
- Specialty Retail
- Supply Chain Planning
- Support & Services
- Value-Based Pricing
- Wholesale Distribution
Is Your DOM Forecasting Strategy on the Wrong Path? Part 2 of 2
AI in Supply Chain Planning and Demand Forecasting
Earlier, we talked about the shortcomings of Distributed Order Management (DOM) systems as a standalone demand forecasting strategy, and why you should start your forecast at the point of fulfillment vs. where the sale takes place. We shared Gartner Analyst Tom Enright’s position on this and suggested the concept of going more granular with your supply chain planning calculations and forecasts.
Today, Tom’s research got us thinking… How can we take this further? Can Artificial Intelligence (AI) be integrated into a supply chain solution to future-proof demand forecasting and extend the reach of DOM?
AI is the next generation of supply chain planning, and it’s a sure bet that your competitors are looking into it. According to an IBM study shared at NRF 2019, supply chain planning and demand forecasting are the industry’s top-two candidates for AI investment:
AI rocks. So why haven’t we figured this out earlier?
Retail investment in AI is not new. It has been applied to nearly every nook and cranny of the retail customer journey and is pretty much kicking ass at it. So why not use AI to fine-tune your supply chain planning and demand forecasting?
You’ve probably seen plenty examples of AI in retail, but here are a few just for fun:
- In-store robots: Japan’s “Pepper” was placed in two of SoftBank Telecom’s stores in Cali. During the one-week beta activation, Pepper yielded a 70-percent increase in foot traffic in Palo Alto and 50 percent of the Santa Monica store’s product sales were attributed to the in-store ambassador. The hip apparel store, Ave saw a 98-percent increase in customer interactions, a 20-percent increase in foot traffic, and a 300-percent increase in revenue from Pepper!
- Sales assistants that identify & converse with Internet leads: Star2Star Communicationsimplemented its Conversica-powered sales rep “Rachel” in 2016 and saw a 30-percent email response rate within hours. The customizable sales assistant can also cross-sell or re-engage existing leads. Another brand, New England-based Boch Automotive, employed Conversica, which it attributed to an average 60-sale increase per month at one Toyota dealership.
- Cognitive computing: North Face adopted IBM® Watson technology to help consumers decide on the best jacket for them based on variables like location and gender. The pilot results from 55,000 users landed a 60-percent click-through rate and 75-percent total sales conversions.
Bring this AI coolness to demand forecasting, and it’s game on! AI technology gives distributors and retailers something DOMs lack – the power and efficiency to hone-in deeper on analytics to put inventory exactly when and where products will be fulfilled.
AI in fulfillment-based forecasting
Fulfillment-based forecasting requires highly granular demand visualization and transaction-level analysis. Good news is, some cloud-native supply chain solutions already incorporating machine learning and AI into their platforms. Unfortunately, many organizations are not even taking advantage of it. That means there’s massive competitive advantage to gain here.
Imagine the awesome sauce here…
With machine learning, you could automate your analysis of constantly evolving logistics, such as long lead times from suppliers and fluctuating transportation costs. Historically, these were just casualties of the business. Now you can use detailed visualizations and transaction-level intel to incorporate unprecedented factors into your allocation process, such as:
- Complementary items frequently purchased together
- Demand by specific ZIP codes
- Time of day; are there blackout hours to consider?
The net-net is having the ability to fulfill inventory quickly with MINIMAL shuffling around between DCs, stores, lockers, etc. and MINIMAL unintended consequences – including unnecessary safety stock, inflated transportation costs, stockouts, and disappointed customers.
Examples & what-ifs
- Autonomous car delivery. What if pharmacies like CVS or Walgreen’s could use AI to auto-magically deploy autonomous cars to deliver just-in-time flu meds during flu season? The system would monitor flu epidemic reports by county and have pre-stocked, self-driving cars ready to deliver relief on-demand to homes in that area.
- Smart pallet placement. What if AI could predict the path of orders for perishable items such as food for grocery stores? The algorithms could forecast when orders will arrive and leave a warehouse, triggering employees (or robots) to place pallets in the most optimal position from the start – rather than moving pallets around like a giant game of Tetris. Slower-moving items get placed further in the back of the warehouse, and fast movers toward the front. We guess this could increase efficiency up to 20 percent. For a company that moves 20 to 30 billion pounds of food a year, imagine the financial impact!
- Drones. What if ecommerce stores could differentiate demand by the demographics of each neighborhood, then deliver the appropriate items via drones in under an hour? Like Amazon, only better… because it’s driven by fulfillment. The system would differentiate the starkly diverse preferences between, say, residents in a high-rise apartment, vs. the preferences of a family of four in the ‘burbs, vs. an elderly consumer who can’t leave her home. AI would track those differences and trigger the system to position inventory with rock-star efficiency and precision.
Being able to accurately forecast these fulfillment preferences and stock inventory at the right places, at the right time, is how brands will compete with the big guys.
Whole Foods is another great example of fulfillment-based forecasting. In their Midtown Atlanta location, they’ve piloted a new program to do home delivery via Amazon lockers. Using store-level analytics and machine learning, they’ve backed up the planning process so they can stock very different items at the pilot store than in other stores. This accelerates delivery time and cuts costs in the fulfillment process. Soon this location will be the biggest Whole Foods in the southeast because of smart planning.
The time is ripe for AI
If you’re thinking about DOM, and you’re not incorporating AI-enabled fulfillment-based forecasting into your strategy, you’re not seeing the whole picture.
DOM as a standalone has too many constraints and too many unintended consequences. We’ve talked with people whose strategy during the holiday season was to fulfil orders for online customers from store inventory. Disaster. In-store inventory was depleted to fill online orders and walk-in shoppers became angry, walk-out shoppers. Quantifying that loss is excruciating.
Anyone can place safety stock everywhere; but getting more granular and strategic about where you position items will accelerate fulfillment, increase customer happiness, and save beaucoup moolah!
How do you envision retail AI and machine learning technology in your process? What if you could achieve the perfect balance between meeting your customers’ needs and being more profitable?