It’s no great revelation that used cars are way more complicated to shop for than new models. And it’s sometimes even more difficult to determine whether or not a customer got a “good price” on a used car. A recent look into buying habits revealed that online retailers can, theoretically, save buyers some money when shopping used, but the details reveal a different story.

Our friends at analyzed millions of 2012-2018 cars sold in dealerships and on the online used car startups Carvana, Shift, and Vroom. ISeeCars.com found that buying online could possibly save buyers money over getting used cars from traditional brick-and-motor dealerships, but in the aggregate that average savings was a mere 0.8 percent, or $185 overall.

There were specific models that have more savings potential buying online, however, according to iSeeCars:

The Chevrolet Silverado takes the top spot with an “average savings” at online dealerships of 6.6 percent. But “average savings” in this context is somewhat misleading, since iSeeCars is not basing the “savings” on the final transaction price, but rather the average of the listed prices for specific models within the 2012 to 2018 model years.

The sample size also isn’t great. On iSeeCars.com, for example, I looked for 2017 Silverados within 200 miles of Austin, Texas which reveals 472 results (which is a lot of trucks, because Texas and all that.)

Carvana’s website, on other hand, only came up with 11 results when I filtered by zip code, Shift only turned up one Silverado in Austin, and Vroom, which doesn’t allow you to search by location, only has 20 listings for 2017 Silverados anywhere.

Moreover, while the iSeeCars listings have a variety of trims, from a super basic work truck all the way up to the loaded High Country, the Carvana listings all seem to be either work truck or LT models (the LT is a mid-level trim that sits above the the W/T and LS trims). Vroom has mostly LT and LTZ (the LZ trim adds upgrades like leather seats and larger wheels)trucks, and doesn’t show any work trucks or fully spec’d-out models.

All of which makes it hard to draw an accurate conclusion about “savings,” since the higher trims offered in traditional dealerships and not online would also drive up their average cost in this analysis.

Meanwhile, if you’re actually searching for a good deal, these numbers are kind of meaningless, since your concern is not the “average savings” but getting a good deal for yourself on a single sale.

At traditional dealers, for example, I was able to find trucks that were dramatically cheaper than the online-exclusive listings. Carvana, for example, has this 2017 LT with 20,000 miles with a “no haggle” asking price of $30,000.

A local Nissan dealer, on the other hand, has a similar 2017 LT with 19,672 miles, and an asking price of just under $28,000.

But that $2,000 price difference also doesn’t reveal the whole story when it comes to trucks. If you have ever spent time on an automaker’s website configuring a domestic truck, you know there is a seemingly infinite amount of configurations with various options, trims, and engines. One of these LTs could be minimally equipped, while the other could have upwards of $8,000 worth of options.

In other words, sometimes slightly more expensive used cars are actually the better “deal” when you factor the equipment along with the miles and condition. And so while online retailers can dramatically increase the convenience factor, if a used car buyer is trying to get the most car for their money, the key is to focus on value, and know that every deal is different.

Determining value always takes a bit of work, and it’s never as simple as looking at miles and price.