Electric Car Chat

Can We Trust Data If The Future Isn’t The Past

• Graham Hill • Season 2 • Episode 18

Share episode

Send us a text

First of all, I'd like to thank you for subscribing to my podcast and sharing with others. I'd also like to wish you a very successful, happy and healthy New Year.

Forget tidy charts that “prove” the obvious. We open with a stork-baby correlation that looks convincing on a blackboard and use it to expose how clean numbers can hide messy truths about electric vehicles, insurance risk, and the way headlines get written. From there, we dig into the mechanics of data interpretation: why new registrations aren’t the same as sales, how identical EV datasets can support both bullish and bearish narratives, and what goes wrong when yesterday’s models are used to predict tomorrow’s road.

Our journey moves from market stats to real-world risk. Early EV insurance looked cheap because cautious first adopters and fewer young drivers skewed the data. Then AXA stress-tested assumptions, flagged higher accident likelihood in specific scenarios, and named a behaviour many drivers recognise: “overtapping” during brisk starts. Layer on battery pack vulnerability, scarce repair capacity, stringent isolation protocols, and upside-down salvage economics, and you get a claims picture that legacy ICE data could never forecast. The result is a sober look at why premiums rose and what has to change for costs to fall.

We close with a masterclass in survey framing via a dentist endorsement campaign that allowed multiple recommendations, creating a headline-ready “90%” without real differentiation. The thread tying it all together is context: definitions, lags, behaviour, incentives, and experimentation. When systems are stable, data sings; when technology and habits shift, experiments lead and datasets follow. If you care about EV adoption, insurance fairness, and honest communication, this conversation gives you the tools to spot manipulation and demand better questions.

If this resonates, subscribe, share with a friend, and leave a review to help others find the show.

To buy a copy of Electric Cars - The Truth Revealed visit grahamhilltraining.com. Buy the current copy and receive the totally updated version in early 2026. If you are interested in sponsoring this podcast or would be interested in working together please visit grahamhilltraining.com/contact


Hi this is Graham Hill and welcome to my podcast Electric Car Chat.

I would have been 16 at the time just starting my A levels. It was my first lesson in A level Applied Maths and being in the 6th form meant that we walked straight into the classroom without having to wait to be beckoned in by the tutor. As we walked in the classroom we noticed two copies of the same graph Selotaped to the blackboard. Younger listeners probably don’t know what a blackboard is but in my day it was a roller contraption with chalks and a rubber, but on this occasion there were two copies of the same graph Sellotaped to the board, very mysterious. 

We found our seats and waited for the tutor to arrive. Then in bounded a rather youthful chap, probably in his mid to late 20s. He went to the front of the class and introduced himself as I think John our applied maths teacher. The very next thing he said was Who knows where babies come from? Our class was a mixed class of girls and boys so there was a bit of giggling going on as we looked at each other and tried to work out what this question had to do with applied mathematics.

Before we could answer, he continued, ‘Because I know where babies come from, as my mum told me when I was a little boy and my mum would never ever lie to me. I heard lots of discussions about the new arrival and that my mum was something called expecting. And the word baby was being mentioned quite a lot. So I asked her where babies came from and without thinking she explained that the new baby would be brought to the house by a stork who would leave it at the bottom of the garden underneath the gooseberry bush.

Now, to a 3 or 4 year old it didn't sound unreasonable. And when my younger brother was born or delivered as they called it, I didn’t see the stalk but I did see my mum holding him in her arms. And that was good enough for me, I now knew where babies came from, no more to be said. And it wasn’t until I was 12 or 13 when rumours started circulating at school on how babies were created and born. I immediately thought what a bunch of idiots, they really don’t know that storks deliver babies. But I was now getting curious.

Now when it came to mathematics he went on to explain that he was very advanced for his age and he had already been studying statistics and data collection. He continued, ‘So I decided to test this theory of my mum’s that babies relied on stalks to be delivered to their new mums. So, I took myself off to the library and started carrying out some research. I first looked at the population of humans in the world over the past 10 years and I plotted the results on a graph. I then researched the world population of storks over the last ten years and again, plotted a graph. Now when I compared the graphs you can see the results taped to the blackboard behind me.’

These are of course the graphs that we saw when we entered the room, Selotaped to the blackboard. He continued, ‘This one over my left shoulder is the population of humans, this one over my right shoulder is the population of storks and as you can see the lines move in sync with each other. When there was a sharp increase in stalks there was also a corresponding sharp increase in humans and when the figures levelled off a bit for stalks they levelled off a bit for humans. The theory that more stalks meant more babies, had clearly been proven and as I’ll be teaching you, as we get deeper into data analysis and statistics, I’ll show how theories can be proven within a very very small margin of error, as demonstrated in this example, allowing you to prove the theory.’ In this case that the world population of humans depends entirely on the availability of storks to deliver babies. Thus proving the theory that storks are responsible for delivering babies to their mums. 

Now here’s the interesting thing. John immediately pointed out that as we all knew in the room, at our age, even with very little sex education, that stalks didn’t deliver babies and leave them under a gooseberry bush at the end of the garden. To be honest I don’t think we had a gooseberry bush at the end of our garden so a bit of investigation could have easily disproved the theory. He then went on to say, ‘Whilst the graphs behind me are absolutely 100 percent accurate, as I’ve actually taken the genuine figures from statistical reports that I found in the library in order to create the graphs. The theory that storks deliver babies is completely ridiculous and had I carried out an experiment to test the theory rather than just rely on the data I could have disproved his and my mum’s theory.

However, what this means, is that you can take any theory you like and interpret statistics or data in order to support your case. There are of course occasions when data and its interpretation can be hugely important but my point is never to take the conclusions reached by data analysis, its findings and interpretations by themselves. Because data and the conclusions are open to interpretation and manipulation. As I’ll explain throughout this podcast. 

In this example the conclusion that babies were delivered by storks was reached without context. Each stork would have to deliver many thousands of human babies every day which would be physically impossible but this crazy story certainly proved the point that the tutor wanted to make about the reliance on data and statistics and their interpretation. There are many that will only use data to support their theory, even when the theory makes no sense whatsoever. 

And frankly this goes on all the time. People using data to come up with false solutions and recommendations. One of the biggest problems with data, according to the psychologist Professor Clayton Christensen and our famous psychologist and social economist, Rory Sutherland, is that data only comes from one place, and that’s the past. So, as has been suggested by Christensen and other psychologists over the years, unless it is what it was, we can’t use data to predict the future. What do I mean by it is what it was? Well let’s consider Gravity. Gravity is today what it was yesterday and 100 years ago. That means That we can rely on data that relies on gravity because it is what it was. But in situations where things are constantly changing, we can’t accurately take what has happened in the past and apply it to, or predict the future. I’ll explain this in one of my examples that I’m going to give you.

So let’s look at an example. Some big mistakes have been made when applying data that was collected from internal combustion engine or ICE cars then applied to electric cars. I’ll get to this shortly but first let me show you how the same set of data can be used to misreport the situation in completely opposite ways. So, let’s look at the sale of electric cars. First of all, we don’t know how many cars are sold during the course of a month or year, as sales are not generally recorded. What is recorded very accurately is new vehicle registrations which some members of the press refer to as sales when in fact, a car that’s registered today could have been ordered or effectively sold 3 or even up to 12 months ago. 

But let’s look at 3 months of data and how they can be interpreted either positively or negatively using the same registration data which I’ll refer to as sales. So, in month 1, sale of new cars comes come in at 1,000 with 200 of those being electric, the headline may say, ‘pretty good sales of electric cars this month at 20%’ or negatively, ‘electric car sales struggling at just 20% of total sales, nowhere near the Government target.’ Those comments refer to exactly the same data. Month two sales of electric cars remains at 200 but total new car sales have dropped to just 800. Headline 1 may read ‘great news for electric car enthusiasts as sales increase from 20% last month to 25% this month moving closer to the Government target.’ Or, it could read, ‘Bad news for EV enthusiasts and the EV industry as the sale of new EV’s flatline at 200 cars this month, the same as last month.’ Finally, in month 3 we see sales of new cars take off to 1,500 with sales of electric cars up from 200 to 300. 300 as a proportion of total car sales of 1,500 is 20%. So headline 1 may read ‘Brilliant news for the EV industry as we see an amazing rise in sales, up a massive 50% on last month’ or ‘Disaster for the EV industry as we see sales of new electric cars drop back from 25% last month to just 20% this month.’ 

And, that my friends is what my old applied maths tutor would refer to as data manipulation. Getting it to say whatever you want it to say. 

So now let’s look at electric car insurance. When electric cars launched in the UK only specialist insurers were prepared to insure them. They looked at the premiums for similarly sized petrol and diesel cars and decided to pitch premiums at a similar level. With no EV specific historical data to use it seemed to make sense. Data was being shared between the insurers and as data on EV’s increased it was found that fewer electric cars were involved in accidents than petrol or diesel. So, if anything, the insurers reduced the premiums to below those of equivalent ICE cars.

As the numbers of registered electric cars increased some of the larger insurers started to get involved and using the available data they continued the precedent set by the smaller insurers and pitched premiums just below those of equivalent ICE cars. 

Then the large French insurance company AXA decided to enter the fray and create an electric car insurance policy. They were offered the available data but refused it. They decided to carry out their own risk assessments on the electric cars that were available at the time. Their conclusion was that electric cars were 50% more likely to have an accident, completely flying in the face of what the historical data was saying. Even worse was the conclusion that AXA came to regarding electric cars with high performance, suggesting that they were 100% more likely to have an accident than their equivalent standard petrol or diesel car. 

However, as I mentioned at the time, what was missing was context. Whilst early data suggested that electric cars were safer, resulting in fewer accidents, surveys suggested that the reason for the low accident rate was the extra cautiousness of new EV drivers as they tried to maximise their very limited ranges and get used to a completely new way of driving. Resulting in far fewer accidents. 

Bear in mind that whilst there is no gear changing in an electric car they are closer to driving an automatic than a manual car and over 70% of cars on UK roads at that time were manual so clearly moving to not just an auto but an auto with instant torque which was for many drivers frightening, so the cars were being driven with extreme caution. Hence an unexpectedly low number of accidents. It wasn’t the cars that were safer but the driving style. Add to this the high cost of electric cars making them pretty much unaffordable for younger drivers, up to 25, the age group that is most prone to have accidents and it was clear as to the reasons why electric cars were the cause of very few accidents. 

Other information started to emerge. The slightest damage to the battery pack was causing some insurers to write the car off completely rather than either repair or replace the battery pack. And this is still happening as there are too few battery repairers able to repair batteries and too little information on the effect of replacing some damaged batteries within the pack with new batteries mixed with the older batteries with less capacity. Add to that the potential time it would take to carry out the repairs, if it was possible, and the cost of providing a loan or hire car to the driver whilst the car was off the road along with the higher cost of repair engineers etc. Not to mention the inconvenience. As a result, insurance premiums have been reviewed on pretty much a monthly basis following the findings of AXA.

I have also read reports that some breakdown and recovery companies are refusing to collect damaged or broken-down electric cars, given the risk of the battery igniting, unless being escorted by a fire crew. Police and fire brigades in the UK and USA have expressed concerns if recovery trucks, carrying electric vehicles, travel through towns and cities and are suggesting re-routing. Again, adding to the cost to the insurance companies. 

Then to add further pain it has since emerged that when electric cars are written off, unlike petrol and diesel cars where the wreck can generally be sold to a breaker’s yard, now called a recycler, for a few hundred, if not thousands of pounds, thus recovering part of the money paid out, this is not the case with EV’s. In fact, quite the opposite as the cars need to be stored safely in the event the batteries catch fire with the possibility of the fire spreading to other vehicles if in close proximity. Specialist dismantlers are needed to strip down the car and the law requires safe disposal of the batteries. So rather than receive from a few hundred pounds to thousands of pounds for a petrol or diesel write-off the insurance company is often having to pay from hundreds up to thousands of pounds to dispose of the EV write-off.

So, back to the main theme, with few accidents during these early days there was very little repair data so assumptions were being made that it would cost the same or less with fewer expensive parts to be damaged in an accident such as engine, gearbox etc. However, this couldn’t be further from the truth, from collecting the damaged vehicle from the roadside requiring the car to be isolated to avoid an electric shock to having the wreck professionally dismantled the costs were looking incredibly high.

Then  AXA identified a serious issue when accelerating. Something they later referred to as Overtapping. This happens when moving off unexpectedly fast, in a reaction the driver lifts the right foot followed by a quick depression causing an accident. A kind of tapping on the accelerator without realising it. So, whilst historical data suggested that electric cars were safer and less likely to have an accident, the experiments disproved the theory, so when AXA introduced higher premiums and the reasons for the increase, it resulted in other major insurers increasing their premiums as well. 

This caused my recently departed adversary, Quentin Willson and his followers, whom I lovingly referred to as the Willson Whingers, to complain about the unfair increase in premiums. However, due to the higher likelihood of accidents in electric cars, the higher cost of repairs requiring specialist engineers and the necessity to isolate the cars before carrying out any work, the extremely high cost of battery replacements and even the cost of disposing of written off cars running into many thousands of pounds the industry felt that the increased premiums were justified.

Over time, AXA was proven to be right when latest information from the Association of British Insurers revealed that for every pound collected in motor insurance premiums £1.10 was paid out in claims. The point I’m making here is that using data collected on ICE cars bore no relationship to the performance and characteristics attached to electric cars. They are clearly completely different modes of transport with completely different sets of risks. And with development of new electric cars making old EV models obsolete within as little as a year we can’t accurately use the old data collected on previous models of EV’s to predict the future risks which is all the insurers are interested in, not the past but the future.

As you can see, it all depends on the questions asked and how the data is interpreted. And as I mentioned earlier, unless it is what it was, we can’t use data from the past to predict the future.

Finally, let me tell you what happened a few years ago in Australia when one of the top toothpaste manufacturers started losing market share. The company had done all they could in development and on the factory floor, different flavours, whitening, mouthwash included in the toothpaste, sensitive gum toothpaste and so on. As a result the board called in their marketing experts and told them to find a way to increase sales. As they sat down to brainstorm the options and how some creative marketing could increase the sales of the brand, one of the group suggested that if they could get recommendations from dentists for their brand, that could be a great basis of an advertising campaign.

The suggestion was to send out a survey to the dentists they had on their database and list the top 10 toothpastes asking which of the toothpastes listed they would recommend to their patients. If over 50% recommended their toothpaste they had the basis of a great campaign. They had just over 1,000 dentists on their database so it would be a strong survey if the results were good. Now, in Australia, Advertising Standards are very tough as the makers of Neurofen found out, that’s another story for another day. The cost of the survey was low, some printing costs and postage, so it was a low cost low risk idea. If the survey came back with poor results the company hadn’t lost much money and the results may provide a different angle for the future asking why didn’t the dentists recommend their toothpaste?

So, the survey went out to the 1,000 dentists with reply paid envelopes addressed to an independent firm of auditors to meet the advertising standards minimum regulations. Four weeks later the results were in and the executive in charge of the survey raced back to the office from the office of the auditors with the findings. The marketing team assembled quickly to be given the news that of those who responded, just over 50% of those surveyed, an amazing 90% plus said that they would recommend their toothpaste brand. 

The marketing campaign pretty much wrote itself and proved to be a huge success increasing sales of the toothpaste brand and the perceived worth of the marketing team. So, what’s wrong with that? Seems like job well done so let’s scratch a little below the surface. Here’s some of the wording used in the adverts. 

We surveyed over 1,000 dentists and of those who enthusiastically responded, over 90% said that they would recommend XYZ brand of toothpaste. In fact, just over 50% responded which was enough to satisfy Advertising Standards as a strong sample. At the same time readers read the 1,000 number which was included in the marketing and this was the number that stuck. Psychology at work.

But the industry was totally confused. The various toothpaste manufacturers worked with dentists all the time to promote healthy teeth and gums, subsidising events both local and national, providing freebies from toothpaste samples to dentistry equipment so why would over 90% appear to turn their back on these most generous toothpaste manufacturers. The question was answered at a dentistry exhibition a few months later when a toothpaste manufacturer asked one of their dentist customers about the survey. 

He happened to have a copy of the survey in his briefcase explaining that he didn’t respond to the survey and passed the questionnaire over to the director of the toothpaste company. The director saw his brand on the list causing more confusion but it was when he checked the question at the top of the questionnaire that all became clear. It asked, ‘Which of the following toothpaste brands would you recommend to your patients?’ and each brand that made up the top 10 selling brands in Australia had a small tick box next to it. On the face of it there was nothing wrong with it and after all it had been approved by Advertising Standards. 

But here was the kicker, underneath the header was another line saying, ‘Please tick as many boxes as you feel appropriate.’ Total data manipulation. So, the question was, ‘Which of the following toothpaste brands would you recommend to your patients?’ Followed by, ‘Please tick as many boxes as you feel appropriate.’ The director found that after asking several dentists, most of those who had taken the time to respond, had actually ticked every box. So there you have it. How data can be manipulated. 

And as I’ll explain in another podcast, you can’t rely on surveys because as David Ogilvy of Ogilvy Mather, the famous advertising agency on which it’s believed, the series Mad Men was based on, was reported to have said about consumer surveys and the responses given, ‘people don’t think how they feel, they don’t say what they think and they don’t do what they say.’ Highlighting the inaccuracies of data collected via consumer surveys.

Look, I’m not saying that all data is misleading and of little use because there are times when data can be used to save time, save money, make things safer and even save lives. But with so much reliance on asking exactly the right questions then accurately interpreting the resultant data should we be so reliant on data over experimentation? Just asking.

I’ve been Graham Hill still making a ruckus. I have tons more to share with you in future podcasts so please subscribe and share with as many people as you can. And if you think that the only way to increase sales of electric cars is to continue to discount and subsidise with Government grants then you need to listen to my podcast on Air Fryers. My next podcast will take a deep dive into the psychology of moving drivers of ICE cars into electric. By for now.