A self-service store, without staff, needs to handle access, product detection, payment, and customer trust all at the same time. Big Fish has achieved this with prefabricated modules of 18 square meters, using weight sensors, ceiling cameras, and a bank card to open the door. In chapter 63 of “Toque de Ingenio,” David Casanovas explains how an initial failed attempt led to five stores operating 24 hours a day, and the technical, legal, and communication decisions that underpin the model.
Guest: David Casanovas, who was introduced in the episode as the CEO of Big Fish. Interview published: September 30, 2025. Episode: 63. Duration: 1 h 10 min.
In this episode:
- Why the store was initially created as a module produced within the group’s own factory.
- How they combine weight sensors and cameras to determine what each customer is taking with them.
- What did they learn about theft, regulations, and trust after more than 6,000 purchases?
When the conversation was recorded, Big Fish had five stores located within its own group’s shopping centers, with the last two having opened in Ibiza in July, and they were finalizing the management software intended to offer the model to third parties.
An automated unstaffed shop: starting with a prefabricated module
David Casanovas explains that Big Fish is a brand that stems from a project initially developed within a pioneering energy group in Spain, starting with self-service gas stations. The same group later incorporated car wash centers, such as the Master franchise from Elefante Azul, and a factory where they manufacture the prefabricated modules housing the machinery for those centers.
The major oil companies already had convenience stores at their stations, and the founder of the group wanted to offer the same service without staff. According to David, the project began about three years before the interview, with a technology company in Madrid. The module was produced on the same site as the car wash facilities. However, this initial attempt failed because the technology wasn’t robust enough, among other reasons, and the project ultimately stalled.
Six months prior to the interview, the project was revisited, and David joined the team shortly after. The team analyzed technologies from various countries, redesigned the module, and selected a solution from Huawei Cloud. Within ten to eleven months, they had the first prototype.
Creating the container at home transforms the shop into a product that is finished at the factory – this is the logic being applied. product design and development When defining from the outset where a solution will be implemented.
Weight sensors and cameras: how do they detect purchases
The shelves are equipped with sensors that monitor the weight of the products, and the ceiling, which features cameras that analyze purchases. The hardware comes from Huawei Cloud and its partners in China; payment processing is handled through Adyen’s POS system, while another provider sends SMS messages, and a software company integrates all the components.
David insists that the system doesn’t record individuals. He tracks movements within the store as if they were an avatar, following them as they move through the aisles and noting what disappears from each shelf. The artificial intelligence then interprets what each customer takes.
“Do not record people inside the shop; what it records are the movements”.
David Casanovas 21:34.
The cameras impose conditions on the module: they require a minimum shooting distance of 2.5 or 2.6 meters. Its external dimensions – approximately 6 meters long, 3 meters wide, and 3 meters high – allow it to be transported on a standard trailer without special transport. It can be unloaded with a light truck, connected to electricity and data, and then immediately begin sales.
Asked about Amazon Go, David notes that its system used only cameras in premises of 300 or 400 square metres with thousands of product lines, whereas Big Fish chose smaller spaces and combined cameras with weight sensors. This is a common approach in Machine vision and automation: Restricting the scope of the scene offers more reliability than simply relying on a sensor.
A purchasing process designed with the customer in mind.
The team dismissed projects that required users to download an app and register, as their initial client is someone who simply follows instructions. The process involves entering only the phone number on the entry screen, without requiring a name or registration. Then, the card is passed over the payment terminal, which holds it briefly as if during a refueling stop. Once the door opens, the purchase begins; when the door is closed again, it ends. Within a few seconds, an SMS message arrives with the receipt, and the payment is released.
There is no receipt printer or basket with RFID readers, because they wanted to eliminate paper and unnecessary steps. Customers can pick up a product, read the label and put it back without being charged, provided it is returned to its place.
Multiple people can use the same card, or different customers can use their own identifier. The technology allows for more, but the store limits the number of buyers to five to ensure a comfortable shopping experience within a 18 square meter space. These are decisions that fall under both engineering and User experience design: The flow must remain clear at three in the morning.
Test it in your own facilities before offering it to others.
The prototype was unveiled at a Huawei event in Paris at the beginning of November, being tested in the factory just 24 or 48 hours beforehand. It worked, and it was then installed in a petrol station belonging to the group in Molins de Rei. The decision was to set up the first stores in its own premises to test the technology and observe how customers were buying.
According to David, the surprise was that the focus was on the technology, while the issues arose elsewhere: ensuring the customer understood the process and trusted that they would be charged correctly. They put up signs saying “smart store,” explained the three steps at the entrance, and utilized the fact that the petrol station employees were able to answer questions.
At the time of the interview, the five stores had a total of over 6,000 transactions, with a rate of 0.1 or 0.2%, partly due to customers attempting to mislead the system.
Theft and regulations: the limits of the model
A shop open 24 hours without staff encourages a certain kind of cunning, and David doesn’t deny it. He explains that the problems have been minimal: there are more honest people than those who try to cheat, the shops coexist with well-lit and monitored centers, and the door only opens with a bank card – and there’s no prize within that justifies a robbery. The two shops in Ibiza, which were open in July, had gone almost two months without any incidents.
Regarding regulation, David describes a fragmented system involving the European Union, the state, regional authorities, and local councils. Legally, the shop is considered a vending machine. The group’s petrol stations already have a license covering commercial activity; a shop requires the standard “premises open” license, while a location in another area requires municipal permission.
“Often, the project comes ahead of the regulations.”
David Casanovas 52:46.
Scaling the model: sizes, turnkey pricing, and management software.
The standard module measures 18 m², but the store at the headquarters occupies 28 m² within a physical space, and one in Ibiza measures 22 or 23 m². In addition to modules, they also offer corner solutions for corporate buildings, airports, or renovated shops, with approximately 250 or 300 references per store.
They offer third parties two options: Big Fish operates the shop under its own brand on the client’s premises, or supplies a turnkey shop for the client to trade under their own brand. As a cautious estimate, David puts a complete module with fridge and freezer at around €120,000, and adapting existing premises at around €80,000, because the metal structures are made in-house and very little building work is required. An annual maintenance and software contract of around €6,000 is added to this.
That software allows price changes to be made from a mobile device, updates electronic labels, displays the planogram for each shelf, alerts to stock shortages, and connects with the client’s ERP system.
David also recounts his biggest mistake: trusting a local tech developer to create the software and bring in external technology. He didn’t manage to do it, and he didn’t realize it in time, which cost the company money, months, and nearly the entire project. The lesson he’s learned is to insist on real testing, rather than relying on promises.
What can another development team learn?
The “Big Fish” case offers four key lessons for machines or spaces that need to operate without supervision:
- Define the scope before perfecting the details. A small space, with limited references and supplementary sensors, is more reliable than a large area monitored solely by cameras.
- Reduce the process to its essentials. A phone number and a card are enough; any further step moves the client further away.
- Test at home and measure. Installing the initial units in our own facilities allowed us to learn from 6,000 operations before selling the model.
- Requesting actual evidence from the technological partner. An intangible development needs verifiable milestones, not just faith or reliance.
At the time of the interview, Big Fish operated five shops in Catalonia and the Balearic Islands, was expanding its range with health and personal care products, sports nutrition and frozen food, and was preparing its offering for third parties. Its best-selling product was the one-and-a-half-litre bottle of water.
At i-mas, we develop automation, machine vision, and electronic systems for machines and spaces that need to operate independently and deliver excellent results from the outset. If your project requires the integration of sensors, cameras, access control, and software into a installable product, Tell us about your project..
To continue reading: How Awayter developed a fully automated, fresh food vending machine – without any staff..
Source of article: Interview with Edgar Guerrero and David Casanovas on Toque de Ingenio. The excerpts cited link to the minute of the conversation. The recommendations for other projects are a summarised editorial from i-mas.
