Five surprising lessons from studying 5000 EVs for 12 months

Plus one bonus lesson on starting a remote-first company in the middle of the pandemic

This article is adapted and expanded from an article originally published on InsideEVs.

Recurrent was incorporated in June 2020, and in September, we welcomed the first wave of EV drivers to our platform. One year later, as five connected cars became five thousand, we have learned a startling amount about EV batteries, about the used car industry and about building a brand new company as a remote-first startup.?

If you’re not following along at home, Recurrent provides battery reports for electric cars, particularly used electric cars. We offer free monthly reports for EV owners and we work with used EV shoppers and car dealers to give them the ability to get instant battery analysis on cars they are considering, with just a couple of dashboard photos.?

The data from all the connected drivers in our program powers our buying reports and has created an incredible knowledge base on used cars across brands and models that we’ve begun to share.

After a year, I thought I’d lay out the key lessons that have really stuck out in Recurrent’s journey so far.

1. EV odometers don’t matter

For as long as used cars have been bought and sold, people have valued them based on the odometer reading. A low mileage car that’s five years old goes for a premium because it’s presumed to be in good shape.

That rule of thumb no longer applies when it comes to used EVs. Our early data science work confirmed that odometer is no better than the 4th or 5th best predictor of battery condition, behind such factors as calendar age, exposure to extreme heat over time, charging behaviors and driving patterns.

We’ve seen vivid examples of this on both sides: A 2013 Nissan Leaf with less than 20,000 miles -- incredibly low mileage for an 8 year old car -- with a battery that was so degraded it barely got 35 miles of range in cold weather conditions. But we’ve also seen 2013-2014 Leafs that get better than their original EPA range today in ideal temperatures, with much higher odometer readings.

The recent full recall of all Chevy Bolts underscores this. Chevy will be putting new battery packs in every existing Bolt. In the 1000 Bolts that we’re connected to and providing monthly battery reports for (representing approximately 1% of all Chevy Bolts on the road in the US), the average current odometer reading is 27,000 miles, with some up to more than 100,000 miles. Within a few months, every one of them will have a brand new battery, turning back their effective age to 0 and with substantially improved ranges. They should be valued accordingly.

2. EV dashboards lie about range

Range is king for a used EV. The EPA range for an EV is wrong on day 1 because it doesn’t account for temperature fluctuations, driving efficiency differences or small variances in the manufacturing quality of battery cells.?

Why can’t you simply charge it up fully and look at what the range says to understand what you’re going to get if you buy it?

The current temperature and recent weather conditions can impact the dashboard estimate of range. Every manufacturer’s battery management system works differently in how they attempt to accurately predict range at any given moment, based both on recent history but also on a prediction of what the driver might do next. And there are clear cross manufacturer differences on how volatile, optimistic or pessimistic the remaining range predictions are. For a great example, compare Chevy and Tesla.

Recurrent’s approach to our range reports on used EVs relies on a lot of data. We have over 10 million unique data points from 29 different makes and models of EVs on the road in many different temperatures, so when we score a vehicle for sale, we’re comparing it using machine learning models and can statistically model for an accurate range estimate.

3. Data matters a lot

As we work towards our vision of a highly-functioning used market for EVs, which is so important to encourage the automakers and dealers to push rapidly into new EV sales, it’s become clear how difficult it is to access battery data in a scalable way necessary to build confidence for the second and third buyer of each EV. We’ve been fortunate to build key partnerships that have helped us with driver acquisition and data access and see the results in our data models. The more vehicles, the more cross-referencing data sources and the more time that passes, the better we are at modeling battery and range degradation for every make, model and year of EV at every temperature.

We’re excited about our new instant battery reports for EV shoppers that’ll provide value to today’s shoppers while providing additional hard-to-get data for future Recurrent data models. We’re just kicking off a new National Science Foundation funded research project to cross reference our remote-data statistical models with plug-in diagnostics (thanks to our most dedicated volunteer drivers). And in the coming months, we’ll be announcing new data partnerships to further improve our data modeling precision to further increase buyer confidence.

4. Used car market is huge, complicated and not ready (yet) for EVs

Used cars sales in the US are over twice new car sales in a normal year and represent nearly $1 trillion dollars of sales annually. There’s an incredibly complex ecosystem of dealers, wholesale auctions, fleet remarketers, valuation and inspection providers, financing options, insurance providers, car service providers, technology vendors, auto marketplaces, digital retailers, and on and on.?

None of them have figured out how they’ll adapt their business models as EVs increasingly flood into the used car ecosystem, going from less than 1% today to 8% of all used car sales within 5 years, while at the same time new EV sales race from less than 3% to 18% market share on the new vehicle side.

This year, as has been well documented, has been dominated by historic supply shortages of used cars caused by ripple effects from the COVID-19 pandemic. Used car prices, including used EV prices, have been soaring, and the industry has been consumed in the short term with finding supply to meet market needs. That has meant that relatively few existing ecosystem players have been able to prioritize looking around the corner and getting ready for this transition, but we’re excited to be talking with the few folks in the used car industry that are thinking ahead.

5. Early adopters care a lot about their EVs

Most of the 5000+ current EV owners that have signed up to share anonymized data with Recurrent in exchange for monthly battery reports are early EV adopters. Think of your friend who has a Tesla and you don’t want to go on roadtrips with them because they won’t shut up about their Tesla the whole time! (Spoiler: our company is full of those people, too.)

It’s not just the Teslarati. There are passionate communities of Leafies (Nissan Leaf drivers), Bolties (Chevy Bolt drivers), The Unbridled (Ford Mustang Mach-E drivers), The Riven (Rivian’s future first gen adopters), and on and on. Early EV adopters have lots of motivations, including environmental benefits, gas savings, they’re fun to drive, and more.

But regardless of why someone is an early adopter, we’ve noticed that they are tolerant of a lot of changing and rough-around-the-edges experiences, both with their cars and with us. As we stumble through a myriad of early user experience challenges, data connectivity issues and data science edge cases, it’s impossible not to appreciate the patience of our people who love EVs, calmly working around bugs and diligently suggesting ways we can improve our software.

This is about to change. The 70% of people who are now considering an EV for their next car purchase are mainstream consumers. They won’t have the patience for bad charging experiences and they won’t have the same interest in diving into data science along with us. They need simple answers to the questions they are increasingly asking: “How’s the battery in that used EV I’m considering? What kind of range will I get now compared to when the car was new? What will that look like 3 years from now in the winter?” That’s who the entire EV industry is ultimately working to serve.

6. Remote-first startups work. In-person time is what’s hard.

Bonus 6th lesson! Why didn’t I title this article “6 lessons…”? Honestly, “Five lessons from 5000 vehicles had a better ring.

Recurrent was incorporated in June 2020 as a pandemic startup. Some people moved, some people got puppies, and some people learned how to bake bread. I decided to start a new company. To be fair, I also moved and got a puppy. And I already knew how to bake bread.?

When my co-founder Kyle and I sat 10 feet apart from each other at an outdoor table last July to write down what we thought out company values should be, number 6 was this:

We’re a remote-first company because we believe it brings advantages in cost, diversity/inclusivity, work-life flexibility and health. However, we have to be deliberate in addressing the challenges (team cohesion and communication) of maintaining a remote workforce.

Timing dictated this decision and frankly we didn’t even think hard about it. Had we started the company in January 2021, we might have established a mental model of having an office and always wanting to get back to it post-pandemic. But we didn’t and we don’t.

What has that meant for us over the last 12 months?

• No rent! Starting with an obvious one: it’s been nice to not pay rent. Within the Seed stage of the company, that’ll add up to roughly a half a junior person for a year, which isn’t that much, but at such a tender age for our company, it makes a difference.
• People everywhere! We have employees in five different states within our first 15 people. Every time we hire someone else in a new spot, I sheepishly apologize to our HR and financial folks for the new headache they’ll have dealing with a different state’s regulations (“Sorry y’all -- but this new person is SO good!”). Our talent pool isn’t limited to just the not-particularly diverse tech workforce of Seattle, and the people we hire can work from anywhere that has a good internet connection whenever they want.
• Remote works! The remote productivity boost is real, across all functions. Having run one in-person week, we all realized how efficient we’ve been working heads down remotely. The commute, the inefficient meetings, ping-pong, etc, etc, etc. When you measure yourself (and managers measure you) on output and results rather than the number of hours spent at the office, it’s a clear winner.

• In-person is hard! Following the examples set by Gitlab and other remote-first companies that have always been that way, we decided to bring the whole company together 3 times a year in-person. At our first work-together week in June 2021, it was clear that we were all out of practice at being in an office environment. Too many meetings, too many social events after work. At the end of the week, we had a post-mortem on the week and all felt like we’d had our least productive week at Recurrent in terms of individual work. However, in the weeks and months that followed, we’ve all realized the value of those meetings and social events in team alignment and connectivity.?
• What’s next? We’re a year into an experiment in company and culture building. There are so many long-term unknowns about how to do the remote-first startup well across all dimensions. We’re committed to the path, but pragmatic and flexible about how to do it. Let’s see where it takes us.