iKassa, or How We Tamed the ‘Digital Cashier’
The evolution of technology might go unnoticed to ordinary customers — let’s admit, nobody cares about yet another release of your application and why it is better than the previous one. However, when you walk into a regular Pyatorochka store to get some bread, and instead of an ordinary checkout lane with a human checker you see a digital cashier that resembles the biggest iPad in the world, you get thrilled and captivated with the power and boundless opportunities of these digital technologies. And then you either find yourself among the early adopters of these digital cashiers and finish up your shopping in just a few minutes, or look for a good old staffed checkout lane and (possibly) complain in the comments of this article asking why does someone once again make you do the job yourself… But there will be no spoilers today, so, you’ll have to read the whole story of how the X5 Innovation Laboratory traveled its own road of trial and error in developing its homegrown self-checkout counter.
Why did X5 need self-checkout lanes
X5 Group started experimenting with self-checkout systems (SCOs) back in the early 2010s. Even before the experiments it was clear that self-checkout lanes could bring tremendous benefits to retail through reducing checkout lines, increasing the throughput of the cash zone, and saving customers’ time. But first self-checkout machines were cumbersome and slow, and even in high volumes their price was out of sight.
However, the idea proved right — digital self-checkout kiosks definitely come in useful. These first efforts gave us the key — the understanding of those thresholds at which the equipment pays off and becomes profitable to the chain. If we plan to install these machines in thousands of our stores throughout Russia, not just in beautiful showcases in the center of Moscow, our digital employee needs to cost several times less than the market average, be fast and easy to maintain, attractive and reliable. Basically, everything as we love — cheap, efficient and quick. In short, this is a practically unachievable task.
We started by looking for ready-made solutions. We looked through different markets, not missing out even the most exciting ideas like shopping carts with cameras. By the way, there is no reason to be surprised: today, any technology beauty parade features equipment of international giant vendors on the one hand, and Chinese innovations — on the other. Moreover, in Asia, natural selection in technology is happening at much faster pace than in Europe — due to the mere size of the market and the highly competitive environment.
As a result, often even the pilot samples and prototypes of Asian manufacturers are better-thought-out, since vendors from the Celestial Empire always try several options at once and only then, having chosen the best one, begin to replicate it.
We decided to try Chinese self-checkout systems on ourselves and at the end of 2018 went on tour to assess the level of digital technologies in Chinese stores. It’s hardly surprising that thousands of stores across the country were equipped with compact locally manufactured all-in-one SCO systems running Android, without scales (since they have everything packed in advance) and with non-cash payments. We sent some requests for information to manufacturers, and it turned out that the most advanced usual system costs about $1.5 thousand.
So, we’ve got a certain vision — a reference for our future system, at least in terms of its appearance. Its internal structure, however, did not match our expectations in any way. One slip-up of the hand after opening the kiosk, and everything is broken, since inside it’s usually like a laptop, with all the boards exposed — although this approach is technologically simple, it is not suitable for field repairs.
How Pyatorochka plunged into DIYing self-checkout lanes
At first, we tried to take the shortest possible path at our Innovation Lab. We requested for samples of kiosks from a suitable price range and prepared to choose the best ones on an aggregate basis, having somehow moderated our expectations for their maintainability.
But that’s not how it turned out! Almost all cheap Chinese SCOs are based on ARM processors and, accordingly, work on Android. However, the X5 Group’s cash department infrastructure has been built on our own architecture for several years. Changing the foundation for a single innovation is simply impossible, at least without huge investments (i.e. for a pilot launch it’s impossible in principle).
We continued negotiations with manufacturers, and it turned out that microprocessor architecture is not that big issue at all — moving to Intel chips would add a moderate sum of $200–400 to the price of a kiosk. However, me faced another difficulty: these devices were designed for the domestic market, and Chinese retail legislation was much more liberal than in Russia. The local players even did not know what a fiscal registrar is. Not to mention the POS terminal.
The new device was supposed to be convenient both for the buyer, and for store employees, as well as for the chain’s service contractors. It would also be nice to make it future-proof and lay the groundwork for further development: new ways of payment, neural networks, or whatever else we will may come up with…
Mikhail Kucherenkov, co-creator of Digital Cashier
When creating a new device for a store, you cannot predict all the scenarios for its usage, which will be in demand in two or three years. Today large vendors are making equipment to meet only the current needs of retailers. And nothing else — otherwise the cost of the device will rise, and they will lose their competitive advantage in tendering process. We, on the other hand, considered both the actual tasks and where technology is moving in retail. That is why, for example, we added stereo speakers, a microphone, and a 3D camera to our terminal. And most importantly, we designed it in such a way that, if needed, we can gradually upgrade it, replacing its components with more powerful ones. Moreover, it will be possible to upgrade the self-service checkout to a whole new level within 10 minutes right in the store. This means that its life cycle will become even longer.
Our engineers at the laboratory made their first intermediate conclusion: to get a device that fully meets your requirements and specifications, you need to design it from scratch, either outsourcing the whole process to a large vendor, or developing it in-house. What’s interesting, these two options are equivalent in terms of costs. But if you want to do it well (and on a large scale) — you have to do it yourself! Therefore, we decided to focus on building our own expertise and developing hardware and software inside the company, and the production will be outsourced to OEM manufacturers.
How we assembled the digital checkout system with our own hands
At first, few people outside the innovation laboratory believed in this project: in the X5 Group (for obvious reasons) no one had developed their own hardware before. However, we agreed that we should try.
The initial understanding of how SCOs should be designed was based on the best models available in the market and common sense. In particular, when analyzing Chinese devices, it became clear that the most reasonable way is to make the exterior of the checkout system as simple as possible and at the same time close to modern user gadgets.
And then we started on our own adventure of developing the SCO kiosk. Its interface was written from scratch by the engineers of the innovation laboratory. The hardware and enclosure were also designed on our own with the support of one of the self-service systems manufacturers. The physical implementation of the ideas was commissioned to the same contractor with its own R&D department and production facilities. And we started prototyping the checkout system.
An important prototyping challenge was to bring together our expertise on classic retail equipment gathered by various X5 divisions and implement it in the device. On average, it took us two weeks to carry out each iteration, with each of them being evaluated by the working group.
Each element of the hardware was selected separately. There were no difficulties with peripherals models — due to the basic requirement that “everything should be like at regular Pyaterochka checkout lanes.” Fiscal registrars and POS-terminals were blended in the overall design of the device. All other components were chosen to our taste, based on our experience and practical intelligence. For example, we wanted the barcode scanner to be robust and large enough, so that customers do not have to struggle “aiming” at the barcode — and believe us, they suffer from it, which “kills” the shopping experience.
The hardware stack was composed of the most well-executed and efficient components available on the market at the design period. But when it comes to a serial production, it will be using a slightly different element base — such are the realities in this industry. That’s an old story: you’ve just picked up the Intel processor that looks most suitable for your application and want to place an order for several thousand pieces within the next three years, when it suddenly turns out that it is being discontinued. As a result, you have to settle with a comparable chip — with the reserve that it will be possible to buy it in large quantities in the next five years.
We decided to make the SCO unit fan-less: the fewer the moving parts inside — the less dust it collects, and the higher is its reliability. I We deliberately choose the model with low TDP: it makes us to write efficient software, with an eye on our limited hardware resources. Besides, it imposes a more meticulous approach to the choice of other technologies. Let’s say you can’t use a heavy software component in the system designed for checkout lanes.
We tested everything, even how deep the pinpad is installed inside the enclosure as well as its optimal angle. The ergonomics of this solution was tested based on life situations: is it convenient for persons of various height to bring a payment card to the reader, is it easy to tap it with a large smartphone?
We were looking for the right solutions based on tasks at hand. For example, we wanted to equip our SCO terminal with a camera. It can be mounted in different ways. Many vendors place it flush with the main enclosure. We used this mounting scheme for our first prototype — but it turned out that it was not that good for the chosen design of the SCO unit. We tried to arrange the camera as a stand-alone ‘head’ on top of the SCO. It got better, but the tests showed that the “head” should be raised higher and its angle changed so that it would be convenient for customers to work with the display. We tested different mounting heights on a prototype during live tests on the trading floor.
In fact, the whole prototyping process was a series of quick iterations on models made from … mud and straw. Or, more precisely, from electrical tape, cardboard, boxes, wooden wedges — in general, everything that was at hand and helped to evaluate the solution here and now. After all, if you do something for a long time, especially in a sphere, which is new for you, there is a high probability that you won’t achieve it at all. Therefore, we were constantly checking every iteration, mainly on ourselves and on our colleagues.
By the end of June, we finalized the concept of our SCO solution, prepared the design assignment for our OEM-manufacturer with all the needed technical drawings, and developed the software for our hardware. Then we ordered a pilot series of equipment and corresponding furniture.
Making the checkout lane more attractive
However, assembling the hardware and designing enclosure is just half the battle. It is the software that makes the self-checkout systems usable, and usable with pleasure and satisfaction. Last but not least, it is its user interface. In retail, self-service checkouts created by large vendors have basically evolved from equipment used by cashiers. And the user interfaces in such devices have been originally designed for store employees. No wonder, such a UI (and we can confirm it) drives ordinary people into a stupor: all the buttons are square, the color scheme is unusual, al the working sequence is obscure — this is definitely not a modern mobile application with clear accents on the UI.
In the Innovation Lab, we evaluated the SCO software that was generally available on the market, and realized that it would be better to build on familiar technology that has already been implemented, like that in the fast-food industry. People are accustomed to such terminals and use them with pleasure. That is why, from the very beginning, our SCO team was looking to the usability of just such devices, as well as on modern web interfaces, using them as a reference for their development.
The same arguments dictated the choice of the technological stack: Electron JS framework, HTML, CSS. In addition, web technologies make it possible to quickly implement changes. The design of the interfaces (it also went through several transformations) was outsourced, and the layer of communications between the GUI and the core of the POS system was developed in-house — it would take third-party developers a lot of time to get acquainted with the platform and its functionalities.
In essence, the cash register software serves as a front-end add-on over the existing platform. For example, Pyaterochka has long been using its own software solution — a platform named GK. It is impossible to rewrite this platform from scratch — too much is tied to it. Likewise, you can’t make daily changes and support them with each release (see the very beginning of this article). So, we decided to wrap our CSO software core in a shell that basically emulates the work of a human cashier. Or, better to say, it converts all user’s manipulations with the screen into operations that are understandable to the backend.
This approach slightly complicates the system, hence, reducing its reliability, but it allows you to quickly create a finished product. All without the risk of “collapsing” one of the operations that are important for retail business, such as exchanging fiscal information with tax authorities and working with bank cards.
One step forward, two steps back
Just like in the case of hardware R&D, developing user interfaces has turned into a whirlpool of iterations. And in the very same way, after several successive transformations, sometimes we had to move a few steps back.
In the early stages of UI development, the desire to make the interface look “like on a smartphone” first led the team down the wrong path. The ‘back’ button was placed at the top left corner, the ‘menu’ button was at the top right. But it quickly became clear that such a layout is inconvenient on a large display: you still need to reach these buttons. It becomes especially clear when you look at an old lady next to the cash register. Okay, she can see these buttons, but it would be more convenient for her when everything is somewhere in the middle of the screen. That is why all controls have been moved down.
In addition, it was vital to ensure that the interface responds quickly to any customer action. In a world where a person is thinking about changing their smartphone, if social web opens for five seconds longer on their current model, few people will have patience to checkout using a cash register that is ‘too meditative’.
What we came up with in the end and how our SCOs work today
The team that created our SCO systems did a great job and made a device that meets all the requirements that we’ve put forward at the start of the project — in terms of production cost, usability, reliability, maintainability — a device, that is future-proof and is ready for the evolution of technologies, which supports quick upgrades, etc.
The pilot deployment started in 2019, when 75 “digital cashiers” were employed in 14 Pyaterochka stores in different districts of Moscow. The results show that customers love it!
Of course, it is still impossible to remove the staffed checkout lanes (and we actually don’t want to): firstly, these self-service systems are intended primarily for those who have relatively few purchases, and secondly, the devices are not yet able to automatically calculate the cost products that are sold loose. We have installed the scales on the second stage of the pilot.
SCOs work a little faster than a regular cash register. The digital cash desk was built on a new hardware base with a margin of robustness for several years ahead. In addition, the software for the device is written in such a way as to minimize delays.
As expected, the youth and middle-aged customers were among the most eager users of SCOs. Some people like the idea of making a purchase using a gadget, others are attracted by its look and feel (white terminals in the current iteration are really nice). However, older people also use SCOs more readily than expected. In their own words, the main advantage of this way of shopping is that no one rushes them, it is easy to check the product prices when scanning, if the product is too expensive, they just can put it aside without buying it. Most regular customers become familiar with SCO’s user interface after one or two interactions.
Today there are shop assistants working in the cash zones equipped with SCOs. Their duty is to come to the aid to customers, to support and suggest actions to those who are not yet familiar with the new kiosks. From the UX point of view, a person who purchases just 2–3 goods — a bottle of water, a loaf of bread, a chocolate bar — rarely faces any difficulties. It is more psychological support that is important — customers need someone nearby and ready to help. Usually, real help is needed when a store visitor is faced with tasks that, in the case of regular checkouts, are hidden from him and that are normally solved by cashiers. For example, the law requires that in case of cigarettes or alcohol, a store employee must verify that the buyer is an adult person. Therefore, after scanning a bottle of alcohol, the SCO terminal is blocked with a pop-up notification on the screen: “An assistant will come to you now.” Perhaps in the future, with improved legislation, this will be made possible with either biometric authentication, or at least with remote authentication, similar to car sharing: the client opens the passport to the SCO, the built-in camera takes the photograph of his face and the document, and a remote employee verifies his/her data.
The SCO terminals are able to add points under X5 loyalty program. However, there is no loyalty card reader in the current version of the kiosk: the support of this program and the work with bonus points is implemented through the Pyaterochka mobile app.
What do we have now?
Today, more than 17,000 self-checkouts operate in 3,800 Pyaterochka stores and 570 Perekrestok supermarkets in all regions where X5 operates. The digital cashier has become a real hallmark of our stores. This is one of the most economically successful innovative projects — more than 40% of our store visitors switch to SCOs. This means that the employees of the outlet free up time for other things. In 2023, we plan to launch another 8,000 SCOs.
Our SCOs also work with payment using biometrics: at Perekrestok, biometrics work at 1,130 SCOs. We plan to further actively develop our iKassa.
