This article is about the broken business model of robotic piece picking companies. If you are looking for the broken business model of Q-commerce companies, please read this article or this one.
For a long time, piece picking robotics was a niche segment in warehouse automation, but the general consensus was that it was rapidly gaining in importance. When I entered the industry in 2015, it was already a thing. SSI Schäfer had launched its first piece picking robot called “Robo Pick” in 2010 [1], which was blazingly fast even by today’s standards. An impressive piece of technology (with some limitations) that unfortunately nobody wanted. It won a few awards, but very few orders. In the years that followed, every other warehouse automation company added piece picking robots to their product range. Several independent robotics companies also entered the market. The best known is probably RightHand Robotics, which was to some extent the poster boy of piece picking robotics companies.
Although robotics companies get a lot of attention and VC money, although they are ubiquitous at warehouse automation trade shows and although every second warehouse operator asks for a robotic solution, the total number of piece picking robots installed in live systems is staggeringly low. You know those surveys where people are asked if they would be willing to pay more for meat to support animal welfare? Most people say yes. Then they go to the store and buy the cheapest meat they can find. These people are the equivalent of all the warehouse operators who say they want robots. None of them buys them. (Well, often you can’t “buy” them, which may be part of the problem. More on that later).
In recent weeks, news has emerged that RightHand is having some financial challenges and Covariant has reached “some” agreement with Amazon [2]. The latter is a bit confusing, but some observers have described it as a sort of stopgap measure for Covariant to ensure its survival. I don’t know if that’s true, and I don’t have the details, but it’s not that important to the point I want to make. The point I want to make in this article is that this and quite a bit more is to be expected in the robotics picking domain. Because despite all the glamor and robots being “the future”, some robotics companies are in a pretty bad position to do business. This has nothing to do with the fact that they are running out of VC money and that the environment for start-ups has generally become more difficult since the end of “free money”. They are in a bad position because of their strategic position and their business model.
A Broken Business Model
One thing that should be clear from the outset is that no warehouse operator wants a robot. They want cheap, reliable and fast picking, but they don’t care if that is achieved by labor, a robot or a magic fairy.
As obvious as this point may sound, it has important implications for the business model of robot vendors – because they sell (well..) robots. And these robots are entirely useless unless integrated into a more comprehensive, and often very expensive system consisting of AS/RS and plenty of conveyors – which the robot vendors don’t sell. So, in addition to the robot vendor and the customer, there is another party: the integrator. Integrators usually act as general contractors and are the “one face to the customer”. Most customers like this because it reduces transaction costs and effort for them.
This means that while piece picking robotics companies got all the attention (and funding) they could wish for, they can’t complete a project on their own. They always need an integrator. And the integrator must be willing to work with them.
Some integrators have their own robotics solutions (e.g. SSI Schäfer, Swisslog). Some have entered into partnerships with robotics providers – which usually excludes cooperation with other robotics providers. So not only do robotics companies need integrators, but the number of integrators available to them is shrinking and will continue to shrink as more and more robotics providers enter the market and more and more integrators develop in-house solutions.
And how should an integrator deal with the situation where a potential customer prefers a particular robotics company, – but the integrator has a contract with another company? Perhaps the customer and the robotics vendor have already conducted picking tests with the customer’s SKUs and these tests were successful. Now the customer and the robotics company want to work together. And the integrator tells them they can’t? What an unpleasant situation for each of the three parties.
The Integrator Perspective
Let’s take a closer look at the integrators’’ perspective. What are the integrators interested in?
- Rapid contract conclusion and project realization
- A reasonable profit margin
- Low risk of delays or non-performance
The project lead time in the warehouse automation industry is considerable. For larger projects, it often takes 12 to 24 months before a contract is signed. And robots only make sense for large projects. Some projects fall apart after 18 months of planning and thousands of man-hours. The last thing an integrator wants is for its success in closing a deal to depend on yet another outside party that (a) overpromises in customer meetings or otherwise makes an incompetent impression, or (b) runs into financial difficulties halfway through the project, or (c) finds out it doesn’t have the resources to complete the project on the schedule agreed to by the customer and integrator, or (d) has few or no references, undermining the integrator’s reputation as reliable. For the robot vendor, by the way, this means not only that it must have a considerable amount of cash at its disposal in order to be able to survive long periods without new project deals, but also that the projects must be large enough to generate sufficient cash to sustain long periods. How is this supposed to work for a – let’s call a spade a spade – a component supplier whose essential value proposition is that its component is so cheap that it pays for itself in a relatively short time?
Also, the planning of the system gets a lot more complicated when robots are involved, because robots cannot (!) simply replace humans, but require some additional processes, some additional equipment, more detailed data analysis, a whole bunch of changes to the warehouse control system, etc., none of which is free or cheap, and none of which the customer is willing to pay for.
In terms of profit margin, integrating a third-party robotic solution is in some ways cash neutral. The robotics providers can’t offer their solution too cheaply because they are start-ups and need the money. The integrator, on the other hand, cannot add a large profit margin to the robotic solution or it becomes economically unfeasible. More importantly, most (all?) robotics companies insist on some variant of a subscription model to generate recurring revenue. After all, why shouldn’t something that works for Spotify also work for warehouse automation? Investors like it too. And it certainly helps to bridge the aforementioned long periods without new project deals.
But the subscription model basically means that the integrator does most of the work and bears all of the risk (see below), while the robot provider cashes in afterwards. That’s not a good deal for the integrator.
On top of that, some (most) integrators have their own component production and are interested in utilizing their production assets, which is not helped by the integration of additional third-party equipment.
As far as the project risk is concerned: The system integrator bears the project risk. If there are delays or non-performance in the project, he is the one who has to pay penalties and suffer reputational damage. No problem, you say: Why doesn’t the contract with the robot supplier stipulate that any penalties for non-performance or delays attributable to the robot must be borne by the robot supplier? Okay, fine. But in an integration project where the functioning of the robot depends on the functioning of many other parts of the system, how exactly are you going to determine who is at fault for a delay or non-performance? Without going into detail here, there are many ways to not achieve adequate robot performance, and a certain amount of finger pointing between integrator and robot supplier is to be expected. And good luck making the (much smaller) robot supplier pay, because that would simply bankrupt them in a heartbeat.
All in all, there is no reason for integrators to be thrilled about the prospect of working with a third-party robot supplier.
The Customer Perspective
Let’s look at the customer’s perspective for a moment. What are customers interested in? They want
- A decent return on their investment,
- Reliability and predictable performance,
- Clear accountability and quick troubleshooting if something goes wrong.
The idea of replacing (scarce, sometimes expensive) labor for a repetitive task with a robot is appealing. It’s not quite as simple as the robot vendors make it sound (see my earlier video on the complexities and shortcomings of most robotic piece picking systems), but let’s put that aside for the moment. Because the business case is not so easy to sell to customers because of the subscription model that robot vendors have adopted. Potential customers don’t seem to like it very much. Because they realize that with the money they pay for the service, they could have bought the whole robot every two or three years. And even if they would still save on labor costs and perhaps even get an acceptable return on investment, they don’t seem to like the fact that they are obviously being ripped off. At least none of the potential customers I’ve spoken to would want that. Also, the robot companies often want to own their data (to train their software, and who knows for what else), and many potential customers don’t like that either. Customers also have to negotiate another contract, directly with the robot supplier. However, they are not interested in dealing with an additional party, which is why they primarily use an integrator as a general contractor in the first place. Or, if the robot supplier works closely enough with the integrator, the customer’s contract with the integrator is at least complicated by one more element.
As for reliability and predictable performance: that’s complicated. I don’t see reliability as a big problem because it’s a technical problem that excellent engineers are working on. Also, many components of picking robots are standard components (robotic arms, cameras, suction cups, conveyors) that are well tested and reliable.
However, the predictability of performance is the complicated part. Because how well a robot performs and how fast it can pick depends very much on the incoming orders – their structure and the items required (see my earlier video for more details), as well as on the way it is operated and fed by the system into which it is integrated. What robot suppliers advertise is capacity (“1,200 picks/h!! 1,800 picks/h!!!1”), and capacity should not be confused with performance. Actual long-term performance is most likely a fraction of these figures, especially if only a fraction of orders can be picked. The moment realistic performance figures are put on the table, the business case can go up in smoke. Furthermore, some benefits (such as “24/7 operation”) only work if the order structure is suitable.
And as far as clear responsibilities are concerned: no customer is interested in having a separate contact person at a component supplier in addition to their integrator. Customers want a single hotline to call, a single maintenance and repair team and a single company they can complain to if something doesn’t work as promised. They don’t want duplicate transaction costs for troubleshooting, maintenance and repair.
Unsolicited Business Advice
This is not normally my style, but if I may, I would like to offer some unsolicited strategic advice for robotics companies that find themselves in the situation described in this article.
First, the ratio of project complexity and duration to pay-off on is really bad. If you have to work on a project for 12 months to earn a few hundred thousand Euros, it’s not going to fly in the long run. So that’s a problem that needs to be solved. Several options are conceivable:
Simplifying your systems so much that an integrator – any integrator – can just use them as a component in their broader, more comprehensive material flow system so you don’t have to stay involved is probably not going to work. It would be great if you could do the AutoStore thing where you sit back and count money while a bunch of integrators compete against each other to sell the same (your) product, but I doubt that will work as the overall complexity of integrating the robot in a meaningful way is a bit too complicated. Maybe I’m wrong with this one, we’ll see.
But the alternative is to be taken over by a system integrator. At this point, your investors may be licking their wounds and like the idea.
And either way, offer an alternative to your subscription models. Keep it for those who like it, but give customers the option to simply buy your stuff and additionally have a normal maintenance contract. By and large, warehouse operators are very down-to-earth and not overly receptive to trendy ideas from Silicon Valley.
A Different Way: Pickr.ai
Some companies that have faced very similar challenges are tackling some of them very effectively. In a recent article, I mentioned Pickr.ai as an example of thinking outside the box [3]. While goods-to-robot systems – which were the subject of this article – typically require large systems to make technical and financial sense, thereby limiting the population of potential customers and projects, Pickr targets small and medium-sized systems. For these smaller projects, they can act as a general contractor or (!) work with their preferred integration partner. They therefore seek direct contact with customers and can take on the entire project, which frees them from the burden of having to work with some anonymous large integrators. They can still do that, but they don’t have to. I think that’s smart.
Conclusion
The current business model of robotic piece picking companies is fundamentally flawed, caught between ambitious technological promises and the harsh realities of market dynamics. While the allure of automation and the potential cost savings make robotic solutions attractive in theory, the complex web of dependencies among robot vendors, integrators, and customers creates significant barriers to widespread adoption. Integrators face increased risks and reduced profit margins when working with third-party robotic solutions, while customers are often put off by subscription models, unclear performance metrics, and divided accountability.
The biggest challenge for robotic piece picking companies is not technological. And the financial challenges are a symptom, not a problem. The problem is their business model.
[1] See https://warehousenews.co.uk/2010/07/the-schaefer-robo-pick-is-alive/ and https://youtu.be/jrHwzEogGJQ?si=6UGlHRrDReDPOaWS (last access: 2024-09-01)
[2] https://covariant.ai/resources/introducing-the-next-phase-of-our-ai-robotics-journey (last access: 2024-09-01)
[3] Full disclosure: I am an advisor to the company. But that’s not the reason why I’d like to highlight them here. It’s the other way around: Because I have been very impressed with the team and their work, I really wanted to work with them.
