Moses Mwangi, known online as Mr Bingo, walked into his electronics shop at Twiga Towers on a Sunday morning in April 2026 and found the shelves empty. Boxes lay scattered across the floor, evidence of exactly how new most of the stolen stock had been. Within hours, footage from his own CCTV system was circulating online, showing the entire crime in painful detail. What that footage could not do, no matter how many times it was replayed, was stop the theft from happening in the first place.
That gap, between recording a crime clearly and actually preventing it, is the single most important lesson from the Mr Bingo case, and it is a gap that a growing number of retailers in Tokyo have spent the last several years working to close using a very different approach to CCTV.
What Happened at Twiga Towers
According to police and CCTV footage reviewed by local media, the operation began well before the actual break-in. At around 9.59pm, a man believed to be the ringleader was recorded engaging the building's security guards in casual conversation. Investigators believe the guards may have been drugged during this encounter, since footage roughly an hour later shows one guard appearing unconscious, with the suspect seen dragging him away. The actual break-in occurred at about 1am, and CCTV footage showed the gang forcing their way into the shop and systematically emptying it within minutes, carrying sacks of computers and loading them into a waiting vehicle parked outside. By the time Mr Bingo arrived the following morning, more than 500 computers and over 100 iPhones, worth in excess of Sh16 million, were gone.
Every part of this operation was captured on camera. The conversation with the guards, the guards being incapacitated, the break-in itself, the loading of stolen goods into a vehicle, all of it exists as footage. And none of it triggered any response while it was actually happening, because the system watching it was doing exactly what it was designed to do: record. It had no way of recognising that a late-night conversation lingering suspiciously long near a guard post, followed by a guard slumping unconscious, followed by an unfamiliar vehicle idling outside a shop after hours, formed a pattern worth flagging in real time.
Why Passive CCTV Fails at the Exact Moment It Matters
This is the core limitation of what security professionals call passive CCTV, meaning a camera system that records footage for later review but does not analyse what it sees while it is happening. Passive systems are excellent at providing evidence after the fact, which is genuinely valuable for investigations and insurance claims. They are structurally incapable of stopping anything in progress, because there is no mechanism built in to interpret the footage and act on it before a human happens to look at the right screen at the right moment.
The Mr Bingo case illustrates this with unusual clarity because the warning signs were not subtle in hindsight. A stranger spending an unusually long time talking to guards late at night is exactly the kind of behavioural pattern that trained analysis, human or automated, is built to notice. But noticing after the fact, while reviewing footage the next morning, does nothing to prevent the theft that has already happened by then.
Tokyo's Answer: AI Edge-Analytics That Watch and Interpret
Japan's retail sector, particularly in Tokyo, has spent recent years building exactly the kind of system that could have changed the outcome of a case like this. AI edge-analytics refers to artificial intelligence processing that happens directly on or near the camera itself, at the "edge" of the network, rather than sending raw footage to a distant server for analysis later. This matters because it allows a camera system to interpret what it is seeing in real time, flagging specific behaviours as they occur rather than simply archiving footage for someone to review eventually.
VaakEye: Teaching Cameras to Recognise Suspicious Behaviour
Tokyo-based AI security firm Vaak, whose VaakEye system has been deployed in dozens of stores across Japan, trains its software on more than a hundred hours of CCTV footage showing both ordinary shoppers and known shoplifters, teaching the system to recognise more than a hundred behavioural indicators, including gait, hand movement, and lingering patterns, that distinguish suspicious activity from normal browsing. In trials at convenience stores, the company has reported reductions in shoplifting losses of over 75 percent, with staff receiving real-time smartphone alerts the moment the system flags a concerning pattern, giving them the chance to intervene, often simply by approaching the person and offering help, before any theft actually occurs.
Tokyo's Unmanned Store: AI Analytics at Scale
Tokyo has taken this even further in some locations. An AI-powered unmanned convenience store operating on the city's Yamanote loop line uses roughly fifty cameras working together to track customer movement and flag suspicious behaviour automatically, including attempts to conceal items or avoid camera coverage while placing goods in a bag. The store's developer has been candid that the system is not entirely foolproof, but has described it as making shoplifting almost impossible to carry out undetected.
Translating the Tokyo Model for Kenyan Retailers
None of this requires a Kenyan electronics shop to install fifty cameras or build a fully unmanned store. The underlying principle, that a camera system should interpret behaviour rather than simply record it, applies at a much smaller and more affordable scale too. Modern AI-assisted CCTV platforms increasingly offer three capabilities worth understanding on their own terms:
- Loitering detection flags when a person lingers in a specific area, such as near a guard post or a staff entrance, for longer than a normal interaction would take. Applied to a case like Mr Bingo's, a system with this capability would have flagged the extended conversation with guards as unusual well before the situation escalated further.
- Tampering detection identifies when a camera is being obstructed, moved, or disabled, which matters directly for cases involving guards being compromised or cameras being tampered with during a break-in.
- Unauthorised entry detection, sometimes called perimeter or zone intrusion detection, flags movement in areas or during hours where no legitimate activity should be occurring, which would have caught the 1am break-in itself the moment it began rather than leaving it to be discovered hours later.
Critically, all of these features depend on the alert actually reaching someone capable of responding, whether that is on-site security, a remote monitoring centre, or the business owner directly. A system that detects a problem but sends the alert nowhere useful offers little real improvement over passive recording.
What Kenyan Retailers Should Ask Before Upgrading
Retailers considering this kind of upgrade should ask specific, practical questions rather than assuming any CCTV package labelled "smart" or "AI-powered" delivers genuine behavioural analysis. Does the system flag specific behaviours like loitering, tampering, or unauthorised entry, or does it simply offer basic motion detection rebranded with newer marketing language? Where do alerts actually go, and how quickly can someone realistically respond once one is triggered? And has the system been configured specifically for the business's own layout and risk points, rather than installed with generic default settings that were never tailored to where guards are stationed or where high-value stock actually sits?
Comparing providers with genuine experience in behavioural analytics, rather than assuming all CCTV installers offer comparable capability, is worth the extra diligence this kind of investment deserves. Platforms such as Secuwatch Tech can help Kenyan business owners find and compare vetted security providers offering AI-assisted CCTV with genuine real-time alerting, rather than committing to a system that looks advanced on paper but still functions as passive recording in practice.
A Grounded Conclusion
The Mr Bingo case is a sobering, very public illustration of exactly what passive CCTV can and cannot do. It gave investigators clear, detailed footage of a well-planned crime, which matters for eventual accountability. It did nothing to stop that crime while it was still unfolding, because nothing in the system was designed to interpret what the cameras were seeing until it was already too late. Tokyo's retail sector has spent years building exactly the capability that gap describes, and while Kenyan businesses do not need to replicate Japan's most elaborate deployments, the core principle, cameras that recognise and flag suspicious behaviour in real time rather than simply recording it, is now genuinely within reach for retailers serious about closing the exact gap that cost Mr Bingo Sh16 million in a single night.
