Ever visited a retail customer service desk looking for a specific item and been told “The system says there are three in stock, but if they aren’t on the shelf, we don’t know where they are.”? Most shoppers have been in this frustrating position at least once (and most retailers, daily).
While virtually every retailer uses inventory management software, they still suffer from stockouts, and often find that physical inventory counts don’t match what the software is reporting. There are many reasons for this, including changes that occur between the time a measurement is taken and when the reporting takes place, and shoppers or staff placing items on the wrong shelves.
In recent years, new computer vision technology has emerged that enables retailers to take instant, accurate inventory snapshots. These near-real-time measurements are far more accurate than physical inventory processes, and generally more accurate than inventory management system counts, because they are based on what’s actually on shelves and in stockrooms — even if it’s in the wrong place.
Retailers are turning to these new systems to:
- Reduce or eliminate stockouts. Retailers call stockouts a billion-dollar problem. A huge increase in “buy online, pick up in-store” orders fueled by the pandemic made this issue even more pressing. Many retailers use their stores as distribution centers for online orders. When staff can’t find a product to fulfill an order and must spend time either finding that item or finding a substitute, any potential profit on that order disappears — that time (and lost sales from items that can’t be substituted) ate the profit. Gaining real-time inventory snapshots of store shelves, stockrooms and warehouses ensures that shelves can be kept fully stocked.
- Get visibility into buying velocity. Buying velocity is a measure of how quickly items are selling, and both retailers and brands want that data, ideally by time of day. It helps retailers with demand forecasting and order accuracy, and enables them to optimize stock levels, practicing “just in time” replenishment that prevents overordering while ensuring products are available when customers want them. Real-time inventory data, taken at multiple points throughout the day, gives retailers timely and accurate insights into stock levels.
- Ensure planogram compliance. Every CPG pays retailers to follow a specific planogram that they’ve invested significant money into. These planograms specify both the position and number of items on a shelf. When retailers are not in compliance with planograms, CPGs can claw back money. With hundreds of millions of dollars at stake, CPGs do random checks of retailers’ shelves. Taking real-time inventory snapshots throughout the day helps retailers ensure they are in compliance (and quickly address any potential issues).
- Ensure price and label integrity. Erroneous prices and labels are another area of concern and potential lost revenue for retailers. First, they can lead to government compliance issues. Second, they lead to customer satisfaction issues and hold up lines in the store while cashiers check on a price. Often, to keep customers happy and keep the line moving, cashiers will simply give customers the price they claim was on a shelf, causing retailers to lose revenue. Real-time inventory snapshots — which also capture price and label information on the shelf — can help identify potential issues so store staff can quickly address them.
So what’s preventing retailers from taking real-time inventory snapshots today? Incredibly, many are still doing inventory using handheld scanners, having staff traverse the store. Shelves might be scanned once a week at the most. This labor-intensive process relies on staff who are incented to be fast but not necessarily accurate — retailers tell us that on average this approach results in a very low 30% accuracy rate, making it almost useless.
Those that use roving aisle robots report a 60% accuracy rate — better, but still not great. Other companies have piloted webcam-style static cameras mounted on shelves, which can be even worse than roving robots in terms of accuracy and integrity — the cameras are easily blocked by people, boxes and carts, making the images useless.
Inventory measurement frequency is also an issue. Retail is a chaotic environment. Shoppers are constantly picking things up and putting them back, often in the wrong place, and it’s impossible for stores to prevent or control that. Even taking inventory once a day is not enough to do it right.
Growth in ecommerce and curbside pickup have brought these issues to the forefront and made them more urgent for retailers to fix. Retailers that want to gain real-time inventory snapshots should look for the following features:
- Look for shelf-mounted cameras with a moving field of view. Static shelf cameras are easily blocked, and floor robots can pose obstacles to shoppers and staff as they move through the aisles. Roving shelf-mounted cameras are unobtrusive while still providing full views of items on shelves. If you must use floor robots, use them for auditing only.
- Keep image processing in-store. There’s no need to upload all raw video to the cloud. Keeping it on premise reduces storage costs. You can do image processing at the edge and only upload summaries to the cloud. This approach turns terabytes into kilobytes, saving tens of millions in hidden transfer costs.
- Avoid systems with batteries. Batteries will last six months if you’re lucky, and no retailer wants to replace thousands of batteries twice a year. Hard-wired systems are easier to maintain and significantly better for the environment.
- Look for something that’s easily upgradable, rather than something you’d need to rip and replace to improve.
- Make sure you’ll own your own data. Incredibly, some vendor contracts give them ownership of and control over the retailer’s own data. Don’t allow it.
A large CPG recently told me, “If we had accurate data on what’s on retailer shelves, we’d need one-third the trucks and one-third the people, and we’d know exactly what to put on each truck instead of guessing. We’d save $1 billion a year.” Retailers are moving toward that point, as real-time inventory snapshots have moved beyond a nice-to-have.
Skip Howard is Founder and CEO of Spacee. He founded Spacee the company in 2016 with the idea of building natural user interfaces and reactive intelligence into the physical world. Howard drove the development of Spacee’s computer vision technology, and the company grew to be one of the leading technology companies in the spatial augmented reality and robotics sectors, with multiple customers in retail, automotive, and manufacturing. Howard has worked in the technology industry for 20+ years. Prior to founding Spacee, he was Founder and CTO for Cancer Gene Connect, a hereditary cancer risk assessments leader, and Co-founder of Pave Systems, a judicial software company. From 2007 to 2015, Skip served as an integral member of Ross Perot’s technology team at Hillwood Development Company.