I suspect that I am not the only one scanning articles on the right electric car to buy. I am teetering on the edge of buying electric but keep feeling that my old reliable petrol car will do just fine for the next wee while. The current pandemic has definitely changed my driving habits. Only two weeks ago, my mechanic was able to tell me that my annual mileage had slipped back to 5,000 miles!
So the Wall Street Journal caught my eye with a fascinating piece, “How Volkswagen’s $50 Billion Plan to Beat Tesla Short-Circuited”. It’s an incredible example of how an industry giant approached software in the wrong way.
Software changes continuously
Software doesn’t industrialise; it evolves – it’s a biological system, not a mechanical one (even though it runs on machines).
Jeff Gothelf wrote a compelling book called “Sense and Respond” several years ago. He recently offered his analysis on the VW issue “Volkswagen’s electric car ambitions: A Sense & Respond case study”.
“What they didn’t consider: Electric vehicles are more about software than hardware. And producing exquisitely engineered gas-powered cars doesn’t translate into coding savvy.” Wall Street Journal
So what happened? VW had been building a huge software and mechanical platform for all of their Electronic vehicles. However, the software wasn’t ready when their new electric vehicle ID.3 launched. Remember that the ID.3 was supposed to take on the might of Tesla. VW endured a year of bad press. That is a lot of boardroom meeting headache time!
What added the icing to this cake is that new ID.3 owners have to physically drive their new cars back to the garage for a “software update” after purchase. When is the last time you have had to head back to a manufacturer for a software update? It is the equivalent of heading to your local Apple store for the latest software release!
In 2021 customers expectations mean that they never wish to see software updates let alone drive to a garage to experience one! Digital companies like Amazon release new software every second. And to really rub it in, Tesla push Over-The-Air updates to vehicles on average every two weeks.
The art of software delivery is simple – rapid and continuous.
What does the map say?
Regular readers will know that I’m a massive fan of Wardley mapping as it not only plots the current landscape and charts movement (evolution) and climatic factors (risk). When I map the VW experience, several observations jump out:
- The mechanical challenge (building a car well) is a difficult, but solved problem. You need to be good at it, but we expect cars to work (bottom of the map)
- With an electric car, customers love the tech so are looking for flashy new features – these must be software-driven (middle of the map). This is a critical driver that may not be there with all non-electric cars.
- There’s a different approach to “being great at manufacturing several million cars” and “continuously pushing out new and innovative software features”.
- If software (digital native platform) is custom, there is room for failure – it needs to be productised for scale. It should not be commoditised.
- Scaling software like a factory doesn’t work – production over evolution is not desirable. You need to embrace change in software, not eradicate it.
The map highlights that scaling software production over adapting to change will be a blocking factor in this environment.
What could be different?
Finally, the real interesting point is “what could have been done differently?”:
- MVP – what is the simplest shippable system? An Over-The-Air update must be part of this – as an essential requirement.
- Evolutionary Architecture. Similar to the MVP, but more important. Tesla ship software that’s incomplete, but is continuously evolving. This is understood. VW shipped software that will sit in the vehicle for a year before an update – this is last generation.
- Leverage Cloud Native services. The lack of OTA update implies that there is a low-level system work that is not complete. It’s a guess – but are there any parts of the software system that they “didn’t need to build” – i.e. they could have rented from the cloud. I understand they use both AWS and
- Integration is hard. I bet this is old school integration as opposed to flexible eventing. Old school integration with 14 vendors is never going to happen smoothly. Could more development be in-house?
- The whole system – was the software an equal partner with the mechanical system, or a peripheral? I’m not sure, but it’s tough to design systems in silos.
With all that said, I’m sure the figure quoted in the WSJ is not all down to software. VW should be applauded for launching such a significant and ambitious project. This is not a multi-billion software error, but the article suggests that “software issues” were a problem.
Let’s be clear; software is never the problem. The approach to writing the software is always the problem.
Would Serverless have saved the day?
If the entire approach was Serverless-First, surely this would have saved the day? Maybe, but let me attempt to call out some fundamental issues that may have been highlighted earlier if they had used a similar approach.
- Clear business KPIs – connectivity, driver safety, diagnostics all would have been high and measured from day one. Surely Customer Satisfaction would have been a KPI also?
- Fast feedback loops – this is essential. Good software delivery needs to prove that critical KPIs are being met continually. I expect the “bad software” should have been flagged very early.
- Observability – with good telematics, the software system should be visualised early and initially, this tests “the concept” and unifies understanding. Ultimately it evolves into live insights to the system.
- Single team – even if sub-components are out-sourced, there’s one team that stands over the whole system. That responsibility will result in the ultimate feedback loop – a red project status early.
The ultimate impact
Fortunately for VW, they will move beyond this quickly. This issue will be forgotten about and only a few new cars owners will feel slightly peeved. But, they bought their cars pre-launch, so they are probably VW super-fans. A few billion dollars is wasted and an exec will get a wrap on the knuckles. Tesla press are delighted (“Volkswagen, a rocky $50B EV bet, and the bid to chase Tesla’s software prowess“). Simultaneously, the plans for an all-encompassing electric car platform in the cloud by VW (“How Volkswagen Automotive Cloud will help shape the connected car of tomorrow”) are still in place and it will be successful. The issues described here will be side-lined as hiccups.
The real issue is what of the future competition? Could be Tesla, could be a Chinese company or a newcomer. Like many industries, the auto industry is ripe for a software disruption. If VW can struggle with a software transformation, how will the other manufacturers cope? There is a fundamental mindset change required when the software is at the heart of your product.