Buying a Car with Data Driven Decision Making

We have been dragging our feet on replacing our aging 2006 Pontiac G6 for years now. Our frugalness has extended the longevity of that car.

In a few years, we’ll replace it

It still runs

The repairs are cheaper than buying something else

Buying a new or used car is too much of a headache

It is better to stick with the problems you know

After several costly repairs earlier this year, the mental gymnastics to justify keeping it around were getting much harder.

It had been a few more years, a few years ago.

It was still running, but it had its share of quirks and when I forced myself to list them all out, I was a bit surprised at the size of the list. It is after all a 14-year-old car.

The repairs weren’t all that cheap, and instead of fixing the problems, they just revealed new ones.

Buying a car is a headache, and will continue to be a headache in the future. There is never a good time to buy a car, but there are bad times to buy a car (like when you need one right this instant).

As my daily driver, I knew all of its problems and could experience them on my commute to and from work every day. That familiar rattle, the untraceable knocking noise at highway speed, the newly added percussion of parts that were either too tight or too loose to sound the way they should. The headlight fluid was always full (condensation in a headlight assembly). The list went on and on.

It was time. We had the money saved to buy in cash and had had it for a while.

First, we narrowed down our selection to a particular make and model. We wanted something that would work well in town, something that was small enough to zip through parking lots, but with enough ground clearance to get out of our steeply sloped alleyway without scraping the bottom. Those two criteria eliminated a wide swath of choices. We also knew of several brands that we thought favorably of and some that we had a bad experience with or just had an unfavorable opinion of.

With all of that and a budget in mind, we were able to narrow it down to a Subaru Crosstrek. The Crosstrek underwent a redesign in 2018 and hasn’t changed substantially since then. That knowledge gave us a nice floor for vehicles and it just so happens that 2018 models are coming off their three year leases right about now so there are plenty of them out there for sale.

I went to the local dealership to test drive one. We had no intention of buying a new year model, but we did want to verify that we liked driving it before doing extensive shopping. One nice thing about car shopping during a pandemic is that you can usually test drive by yourself without the salesman. I was able to drive the vehicle over to a local park and meet Shae and the kids there instead of dragging the whole gaggle to the dealership. With four votes yea to that make and model, we started our data driven car shopping.

First, we needed data. We input our parameters, make, model, year, trim level, and sought after options into online searches. CarMax, Vroom, and Carvana all offer slick digital storefronts with no-haggle pricing. They also ship cars nationwide, so it is easy to gather many data points for a very specific vehicle. Then we broadened our search to include traditional dealerships. These listings were often trade-ins and were frequently at other brand stores like Kia and BMW.

Right-click and open the image below in a new tab to see the chart we based our good deal indication off. Each blue dot is a car for sale.

Any car over the blue trend line was a bad deal and taken off our radar. The perfect car would be at minimum X and Y values (mileage, price). One surprise that I had doing this was that the trend line valued miles at 11.2¢/mile. The IRS lists the mileage rate at 57.5¢/mile. Perhaps Subarus hold their value better?

One other data visualization that we added was a positive error bar to represent sales tax and other out the door costs.

What we ended up with was an easy guide to good or bad deals. Simply find the data points that are farthest below the trend line and buy that car.

Our first choice sold while we were inquiring about it. Our second choice was not as advertised and had extensive damage. There was a “too good to be true” deal at the higher miles range, but it had been extensively modified by a DIY for rally racing and didn’t seem like something worth pursuing.

At the end of a very long day of driving to the city and back with stops at two dealerships, we had our car. Yipee!

Haggling was very minimal. Most dealers were not interested in negotiating large discounts. The largest I negotiated was 4% off their asking price and I still didn’t buy that vehicle because it was only halfway to my number. 1% was far more common.