The average franchise dealer appraises somewhere between 15 and 40 trades a day. At $800 average loss-per-trade from appraisal error alone — a conservative number — that's $12,000 to $32,000 walking out the door every single day before you've looked at a single front-end deal.
Nobody talks about this number because the losses are distributed across hundreds of transactions and the mistakes are invisible in real time. You see them in your wholesale pack and your used-car gross erosion and your auction lane write-downs six weeks later. By then the desk manager who appraised the trade is already on the next deal.
Where the money goes
Appraisal loss comes from three places. Understanding which one is eating you tells you how to fix it.
- Reconditioning blindness. The appraiser estimates recon at $400. Actual recon runs $1,100. That $700 gap comes directly out of your used-car gross. The fix: enforce a recon estimate model that includes regional labor rates, historical actuals by vehicle class, and current parts pricing. Nobody does this manually. The AI can do it in seconds.
- Market position lag. The appraiser pulls Manheim and KBB ICO from yesterday's data. The market moved three days ago. In a softening market, yesterday's guide is today's overpay. In a tightening market — say, after a hurricane clears inventory in a region — yesterday's guide is today's missed opportunity. You need real-time comp pricing from your actual 100-mile market, not a national average.
- Retention incentive distortion. The closer the customer is to walking, the more the desk over-allows on the trade to save the deal. This is the most expensive one because it's invisible — it looks like a won deal. What it is: a funded deal with a trade that's $1,200 upside down before it hits the lot.
What 30 years on the desk taught me
The best appraiser I ever worked with had one habit that separated him from everyone else: he never appraised from the curb. He wrote down three numbers before he walked to the car — what he thought the recon would be, what he thought the market would pay, and what he thought the customer paid for it. Then he reconciled those three numbers against what he actually found on the walk. The car told him if he was right or wrong.
That discipline is what an AI appraisal assist replicates. It doesn't replace the appraiser's eyes. It gives them a pre-calculated anchor based on real recon actuals, real comp pricing, and historical auction results for that specific make/model/trim — so the appraiser is correcting a number instead of generating one from scratch.
Correcting is faster and more accurate than generating. Every time.
The retention distortion is the hardest to fix
The over-allow on a hot deal is a management problem, not a tools problem. If your desk managers are trained to save deals by over-allowing trades, no AI fixes that. You fix it by separating the appraisal decision from the closing decision — structurally, in your process.
The appraisal goes to one person. The trade allowance decision goes to another. The delta between them is a visible number in your DMS every day. When that delta is tracked, managed, and discussed in your morning briefing, it stops being invisible loss and starts being a managed line item.
That's the whole game: making invisible loss visible. The tools help. The process has to change.