Most vendor ROI decks for dealership video are noise. Open rate lifts. Click rate lifts. “Engagement” graphs with no unit on the Y-axis. None of that lands a car in your showroom.
A dealer CFO evaluating personalized video against manual-record tools (TradePending Video, VentaVid, Covideo classic, CarFilm) or AI avatars (Covideo AI Video Agent and similar) needs to know exactly three numbers to make this call. Not seven, not twelve. Three. We’ll name them, show you a worked example on a realistic 3-rooftop group, list the things that kill the ROI calc, and hand you a one-line decision rule.
No vendor pricing in this post — the framework works regardless of which tool you pick. The goal is giving you the math, not selling you on our line item.
The three numbers
Everything else is a derivative of these.
1. Show rate on confirmed appointments
Industry baseline without personalized video sits near 33% — a 67% no-show rate is the number most dealer groups would quote you if they’re being honest. Manual-record video shops typically report 40–50% show rate on appointments where the video went out. VoxRefine’s measured outcome on the real-face-plus-cloned-voice pipeline is 40%+ show-rate lift above the baseline, which moves a 33% starting point above 46%. Every ROI calculation starts here. If your show-rate delta is real, the rest of the math works. If it’s not, nothing else matters.
2. Cost per appointment kept (not cost per lead)
Industry figures commonly cited in automotive marketing press put average cost per lead around $300. That number alone is useless for decision-making. What matters is cost per appointment kept — CPL divided by show rate. At a 33% show rate, a $300 lead costs $900 per kept appointment. At a 46% show rate, the same $300 lead costs $650 per kept appointment. Same lead cost, different universe. The show-rate lever is usually the single highest-leverage number on the P&L, and personalized video is one of the few ways to move it without touching media spend.
3. Rep-hours consumed to get there
The number nobody puts on their vendor’s ROI deck. Manual recording eats rep time at a linear rate — roughly 15 minutes per video including retakes, upload, send. That time comes out of the same rep’s day you’re also paying to sell cars. An hour of rep time disappeared isn’t free. Price it at the fully-loaded cost of that rep — salary, commission structure, benefits — and it shows up fast.
Worked example: 3-rooftop group, 5,000 outbound touches/month
Take a 3-rooftop group running a 5-person BDC. Combined outbound volume across the three stores is 5,000 touches per month (new leads, follow-ups, appointment confirmations, service reminders, equity mining outreach).
Baseline (template email only): 5,000 emails at a 2.1% BDC email open rate clears ~105 opens. A generous 20% reply-to-open rate is 21 replies. Half convert to confirmed appointments — call it 11 appointments set per month from this volume. At a 33% show rate, ~3.5 appointments actually keep. That’s the number on your P&L. Three and a half kept appointments out of 5,000 sends. Nobody wants to stare at it.
With personalized video: Same 5,000 touches. Video-attached emails routinely drive 3–5x open lift vs template text — call it 4x, so ~420 opens. Reply rate climbs with engagement, so assume ~63 replies and ~32 appointments set. At a 46%+ show rate (the personalized-video floor), you’re landing ~15 kept appointments per month. That’s ~12 incremental kept appointments vs the baseline, every month, same lead sources.
Rep-hour cost: If you manual-record at even 10% coverage of 5,000 touches, that’s 500 videos a month at ~15 min each = 125 rep-hours absorbed. That’s three-plus full work weeks of rep time burned before you sold a single car off the effort. VoxRefine takes that to roughly one rep-hour per month per team member on camera — a single source recording fires thousands of personalized outputs automatically.
Twelve incremental kept appointments, multiplied by whatever your group’s average gross per sale runs, minus the platform subscription, minus near-zero recurring rep-hour cost. If your average front-end gross on a unit is anywhere close to industry norms, that math closes inside the first month.
Want us to run the math on your actual numbers — your CPL, your outbound volume, your average gross? We’ll walk it on a call.
Run the math with us →The capacity ceiling nobody talks about
Manual-record personalized video has a hard ceiling, and every dealer running it at volume has hit the wall. A motivated rep can record maybe 20–30 personal videos a day on top of their actual day job. Past that, quality drops — they start mumbling the customer name, skipping the vehicle, reading half the script. The personalization stops being personal.
At 5,000 outbound touches a month, even 10% video coverage means 500 videos. At 15 minutes each, that’s 125 rep-hours a month, or roughly 6–8 full work days of pure recording time — not sold in the deal, not handed off to F&I, not out on a test drive. Every hour of rep recording is an hour they aren’t closing. The math stops working somewhere between 1,000 and 2,000 videos a month for most dealer groups.
The compute-bound tier (one recording, AI-generated personalized audio in the rep’s own voice) removes the ceiling entirely. 10,000+ videos per hour across a GPU cluster. That’s not marketing copy — it’s the architecture reason the ROI math survives at volume.
What breaks the ROI calc
The math above assumes operational hygiene. Here’s where real dealerships actually blow it up:
Bad source recording quality. If your on-screen salesperson recorded the base video in a back office with fluorescent buzz and a phone mic, every downstream video inherits that quality. Your customers just see a grainy, tinny clip with their name spliced in. Show rate does not lift. Spend 20 minutes doing it right with a clean mic and decent lighting.
CRM/DMS integration failure. If the platform can’t reliably pull customer name, vehicle of interest, and appointment time from CDK, Reynolds, Dealertrack, VinSolutions, or DriveCentric, the “personalized” video is neither personal nor useful. Worst case: wrong name, wrong vehicle, wrong time — which is worse than sending nothing. Test the integration on 50 leads before you scale.
Scripts that sound generic. “Hi [name], looking forward to seeing you on [day] for the [vehicle].” That’s template email with a face on it. Scripts need to reference something specific — the trim, the trade, the financing conversation from earlier — to actually land.
BDC not knowing video is going out. If your BDC rep calls a lead an hour after the video fires and says something that contradicts it, you’ve just confused the customer. Coordinate the touch sequence. Video should complement the BDC cadence, not collide with it.
Simple decision rule for a GM
Forget the feature matrices. This is the whole decision tree:
> 1,500 personalized outreach videos/month: Compute-bound automation (VoxRefine tier) wins. Manual record is burning rep-hours you can’t afford, and AI avatars cost you show rate because customers clock the synthetic face.
< 500 videos/month: Manual record is fine. Covideo classic or TradePending Video is the right tool. Don’t overbuy infrastructure you won’t use.
500–1,500 videos/month: Transition zone. Most groups handle it personal-first (reps record for high-intent leads) with automation backfill (voice-cloned video for appointment confirmations, no-show follow-ups, service reminders, equity mining). Run both. Revisit when you break 1,500.
See the math on your own numbers
Send us your outbound volume, your blended CPL, and your average gross per sale. We’ll walk the ROI on a call with your own numbers — not a generic vendor deck.