Leif Johnson Ford · Manor, Austin, Buda · Impel review

Impel: What It Did, What It Costs, What's Next

For the group conversation · Prepared by Will Robbins (GSM, Manor) · data as of 10/15/26

$108K
a year for 3 stores: about $3,000 per store per month, $9,000 a month per Will, verify invoice
3.8 min
Impel's first touch at Austin and Buda, around the clock, on 96–99% of Internet leads
0
sales we can show Impel caused, at any store. Manor turned it off 8/21 after its test run
Manor timeline

Chloe ran about 90 days at Manor: 5/22 to 8/21

Apr–May
Impel set up at Manor. Website chat leads ("Chat-Impel") from late April. Our own machine starts 4/7–4/8.
5/22
First Chloe message in VinSolutions. Impel's dashboard rows start 5/27.
5/26–7/27
Our sending jobs paused (0 sends 5/26–7/6, a trickle to 7/27). Chloe is the only AI on Manor leads.
6/23
First read of Impel's numbers: 337 leads worked, 4 AI appointments, 126 handoffs, 0 AI-assisted sales.
July
Impel's impact report: 542 leads worked, 4,134 messages, 3 AI appointments, sold tile blank.
7/28
Our machine back on. Both AIs on the same leads.
8/17
Full census of every Chloe conversation (next slides).
8/21
Will switches off Chloe's first response on new leads. Leads 3+ weeks old still allowed.
8/25
90-day Chloe vs our machine comparison, at Will's ask.
9/4
"Impel = website chat only." Our Impel monitors paused.
10/7–10/9
Impel's lead nurture fires on 19 new leads (slide 7).
10/13
Paused on all 5 live conversations; Will asks Impel to disable it account-wide. 10/14 check: 0 new Impel notes.

The test window

5/22 → 8/21 = 91 days, from the first Chloe message to Will's switch-off. No written trial, contract term or price is on file anywhere we can read. For 6 weeks (5/26–7/6) Chloe was the only AI on our leads, and nearly so through 7/27.

What we can and can't see

  • Impel's own counts: the 6/23 pull (90-day lookback), July's report, a 6/30–8/17 pull and Will's own 7/23–8/21 read. No month-by-month before July.
  • VinSolutions notes catch about 1 in 4 Impel conversations at Manor (106 of 453 matched leads), and every Chloe note is cut off at ~450 characters. Nobody could read what she said in the CRM.
  • So speed from notes can't see Impel. VinSolutions' own first-outbound clock does (from 7/17 on); slide 5 uses it. Appointments, shows and sales come from VS and Axcessa and are solid.
The test: Impel's report vs our CRM

Lots of activity. Appointments in single digits. Sales: none credited.

Impel's own countsto 6/23 (rows from 5/27)July7/23–8/21 (Will's read)
Leads worked337542 (400 new)562
Messages2,0604,134 (69 bounced)4,166
New leads engaged101 (30%)113 (28%)86
Handoffs to a person126126 (41 call requests)n/a
AI-set appointments435
Sales Impel credits itself0 AI-assistedtile blank0

"542 worked" is nearly every Internet lead in the CRM that month (real July Internet leads ~425 after removing a 2,081-record Ford mailing import). A 6/30–8/17 API pull agrees: 616 worked, 171 handoffs, 6 appointments. Will's 7/23–8/21 read was recorded in the 8/25 comparison, not re-pulled.

Checked against VinSolutions and Axcessa

~10%
of July handoffs sold (about 13 of 124), vs ~13% of all July Internet leads. Handoffs closed at or below average.
14 of 128
Chloe leads in the 6/23 pull that ever sold; 0 Impel-assisted sales on Impel's own report.
1
sale from the "Chat-Impel" source in July: 14 leads, 3 appointments.
41
phone-call requests and 33 after-hours conversations in July: the real value it showed.

Sold = name match to VS / Axcessa; a handoff that sold may have been sold by a rep who never saw the handoff.

Every Chloe conversation, Jul 1–Aug 16

Our machine got more replies than Chloe did

Full census, VS comm historyChloeOur machine
Conversations we could verify in VS401 (of 600 Impel claims)556
Customer replied (generous)7.0%10.4%
Replied to that AI (strict)4.2%10.3%
Real back-and-forth, 2+ reply days22
Sold in the window02

Strict = the customer's last message before the reply was that AI's, and the other AI hadn't sent anything in the 15 minutes before. 183 of Impel's 600 claimed conversations matched no VS lead at all. An earlier 41-lead sample said Chloe got 71% replies; it was pre-filtered toward engaged leads and is withdrawn.

78
times Chloe and our machine texted the same customer within 15 minutes (75 customers; 10 in July, 68 in Aug 1–16). Tightest: 4 seconds apart. 122 customers heard from both.
~450
characters: where every Chloe note in VS is cut off, and Impel's two tracking links use most of it. A rep opening the history saw "Hello [name], This" and nothing else.
  • Worst seen: 23 unanswered messages over 52 days to one customer, who had to ask twice to be opted out. Impel logged it as a successful handoff.
  • Recurring "Has someone contacted you yet?" loop (~8 times to one customer).
Manor before / during / after · Internet leads

When Impel left, speed dropped. Appointments, shows and sales didn't.

Period (lead created)Which AILeadsReached ≤1h (VS clock)After-hours first touch, medianPerson ≤1hAppt setShownSold
Jan–Mar 2026neither (people only)1,222n/an/athin notes12.0%7.0%11.3%
4/7–5/25ours alone586n/an/a55.9%16.2%11.4%12.3%
5/26–7/27Impel alone (ours paused)72094%*4 min*43.6%17.8%11.7%12.4%
7/28–8/21both28995.5%4 min37.6%16.3%9.3%11.4%
8/22–9/12ours alone (Chloe off)28970.2%241 min42.8%18.7%11.8%12.8%
9/13–10/13 too youngours (+ Impel's 10/7–10/9 burst)30583.3%12 min43.6%13.4%8.9%9.2%

What moved

  • Speed. With Impel, 95% of Internet leads got a touch inside an hour and after-hours leads in ~4 min. With it off: 70% and 4 hours. Our overnight emails brought it back to 83% and 12 min.

What didn't

  • Sold per Internet lead: 12.3% → 12.4% → 11.4% → 12.8%. Shown: 11.4 → 11.7 → 9.3 → 11.8. Inside normal noise.
  • Appointments per lead stepped up in April (12% → 16–19%), before Impel, and held either way. People's own speed didn't change with Impel.

Read it right

  • Not a controlled test: seasons, staff and lead mix changed too. It rules out a big effect, not a small one.
  • *VS clock exists from 7/17 only; the Impel-alone row uses 7/17–7/27.
Month by month · Manor Internet leads

Sold per 100 Internet leads, 2026

Jan · neither10.8
Feb · neither10.2
Mar · neither13.0
Apr · ours13.4
May · ours, Impel late10.9
Jun · Impel alone12.7
Jul · Impel, ours late12.4
Aug · both, then ours13.6
Sep · ours (young)7.4

Same months last year (people only): Jan 7.3 · Feb 6.7 · Mar 9.5 · Apr 9.6 · May 9.8 · Jun 9.6 · Jul 10.2 · Aug 10.8. September 2026 leads are still closing.

Appointments set per 100 Internet leads

Jan 11.2 · Feb 12.2 · Mar 12.8 · Apr 17.6 · May 15.8 · Jun 18.4 · Jul 16.7 · Aug 16.9 · Sep 17.6

Shown per 100 Internet leads

Jan 6.4 · Feb 7.9 · Mar 6.9 · Apr 12.6 · May 10.6 · Jun 11.7 · Jul 10.4 · Aug 12.1 · Sep 9.5

  • The step up came in April. Impel's months (green) sit inside the range of the months around them.
  • Every 2026 month beats the same month of 2025, with or without Impel.
8/22–10/13 · what we found when we looked again

After the switch-off, Impel kept texting new leads

  • Impel first-response notes still landed on 10 new leads 8/24–9/17 and 1 on 9/29.
  • 10/7–10/9: a burst on 19 new leads (27% of leads created those days), with a vendor test lead among them.
  • Impel's report page (its default range 9/13–10/13) shows the result:
20
new leads (9 after hours)
4
appointments (2 AI-only), 1 shown
1
sold, a customer our managers and machine were also working

The STOP case (customer Matt T., 10/8)

8:10am
Lead arrives. 8:11 a first email is logged in VS.
8:15
Impel texts him, then emails at 8:19. It quotes $36,000; the vehicle card said $34,768.
8:20
He texts STOP, to Impel's thread. VS never got it, so 3 more texts were logged to him 10/9–10/12.
+5 days
Impel still shows him "AI active" until we paused it 10/13.

In Impel's own settings (harvested 10/13)

  • Knowledge bank: "we can hold a vehicle for a short time with a deposit". Our rule: no holds, no deposits.
  • "Serving Manor and the Austin area". We never name a sister store to a customer.
  • No master on/off, no quiet hours, no daily cap. Every source and status had "responses" off, and it texted anyway. The hard stop is on Impel's side.
  • Cadence: 13 steps, text + email on days 1, 3, 5, then email to day 51. Step guidance blank.
  • 4 of the 5 live Impel conversations were customers our machine was also messaging. One got Impel's text at 11:30pm, its email at 11:31pm and ours at 11:32pm.
  • No Impel user at Manor has a recorded login; 5 reps have no Impel user at all. Handoffs had no owner.
Austin and Buda · Impel live

At the sister stores Impel touches almost every Internet lead in under 4 minutes

Internet leads Sep 1–Oct 12 (census)AustinBudaManor
Leads read743738395
Impel first touch on the lead99.3%96.1%7.1% (off)
Impel median minutes, all hours3.73.84
…after hours4.64.53.3
Impel leads with 2+ messages logged46.5%45.8%75%
Impel lead never touched by a person6530

Personas: Emma Reyes (Austin), Ella Moore (Buda), Chloe Davis (Manor). Customer replies to Impel are not logged in VS at either store (0 before any staff touch), so Impel's reply rate there is unavailable.

And the people behind it (first staff outbound)

AustinBudaManor*
Median minutes, all hours164.241.3128 person-only (5.6 if our machine's texts count)
…leads created 8a–8p61.615.13
…leads created after hours672.6617.816 h for 5–9pm arrivals, person-only
Staff touch within 1h34%53%42% person-only (77% blended)
No staff touch in 24h (or ever)24.4%5.8%1%
Customer replied after first staff outbound29%37%55%

*Manor blends our machine with staff (machine texts post under user names); Austin and Buda exclude Impel. Manor person-only: 14.5 min median for leads landing 8:30–5, 104 overall (nights and evenings). Not like-for-like on speed; the 24h and reply rows are the fair read.

The VinSolutions response tile

Austin's "99% in 15 minutes" is Impel stopping the clock

VS tile, Oct 1–14: answered in 15 min

Austin tile99.2%
Manor tile74.5%
Buda tile42.6%

Staff actually reached them within 1 hour

Austin staff34.3%
Manor77%
Buda staff53.4%
  • Austin: Impel posts "Sales AI email/text sent. Clock stopped." on 737 of 743 Internet leads. VS counts that as answered. Staff median is 164 minutes, and 65 leads (8.7%) never got a person.
  • Buda: Impel never posts the clock-stop note (0 of 738). Its tile shows 49% "Missed" in October although Impel reached 96% inside 4 minutes.
  • Manor: no Impel since 8/21; the tile reflects our machine and staff.

So what

With a bot in front, the VS response tile stops measuring people. Judge staff speed on comm history, store by store, whatever we decide about Impel.

Results across the group

Impel stores don't close more Internet leads than Manor

Per 100 leads, last 30 days (all types)ManorAustinBuda
Appointments set28.319.714.9
Shown17.19.19.0
VS sold17.819.713.4

Mature close, leads Jun 1–Jul 15 (91+ days old)

Manor Internet12.4%
Austin Internet12.4%
Buda Internet6.9%

Counts: 62/498 · 110/888 · 59/852. All non-bulk leads: 16.1% · 20.5% · 13.5% (Manor carries repeat-customer leads, Buda service referrals).

  • Austin, with Impel on 99% of Internet leads, closes them at exactly Manor's rate.
  • Buda, with Impel on 96% and faster staff than Austin, closes about half as many.
  • Manor ran Chloe until 8/21, so the June–July cohort isn't bot vs no-bot either.
  • Austin's show rate fell from 58% (Sept 1–14) to 37% (Oct 1–14); cause unknown: real no-shows or appointments not closed out.

The fair read

The data can't say Impel helps or hurts closing, at any store. What it does show is speed and coverage. What decides the sale is what people do after the first touch.

What it costs, what it buys

$9,000 a month for 3 stores. Here's what each store gets for $3,000.

$3,000 × 3 = $9,000 / mo
$108,000 a year. Per Will; verify against the invoice. Contract term, end date and whether Manor is still billed: unavailable
Per store, per monthAustinBudaManor (July, live)
Internet leads Impel first-touched~539~511542 claimed · 263 verified
Cost per lead touched~$5.60~$5.90$5.50 claimed · $11.40 verified
AI-set appointmentsdashboard not readdashboard not read3 (≈ $1,000 each)
Sales it can show it caused0 shown0 shown0

Austin / Buda leads = last 30 days of Internet leads × Impel coverage (X §1, §4). No Impel invoice, contract amount, term or renewal date is on file anywhere we can read; VS lead-source ROI has Impel's cost set to $0. The "$3,000" on Impel's own report page is estimated gross on its one sale, not a fee.

What Manor's own machine does for that job

  • Speed: first touch 4.6 min median; 3 min in store hours (blended with staff).
  • Coverage: 81% of Internet leads touched (9/13–10/13); left off on purpose where a rep is already working the customer.
  • Replies: 23.6% of first-touch texts get an answer with no person in between; auto-responder 36.7%.
  • Appointments: 7 of 371 set with no person before (5 showed, 5 sold). "Assisted", never "created".
  • Cost: no license fee; AI usage spend not recorded yet. Will's build time not costed.

Neither AI has a measured cost per sale. Formula once both bills are on file: monthly cost ÷ shown appointments it touched.

Impel vs Manor's machine

Same job, different strengths

Impel (Austin, Buda; Manor until 8/21)Manor's machine
Speed, store hours3.4–3.6 min median3 min first outbound (blended); first-touch 4.6 min
Speed, after hours4.5–4.6 min, texts any hour (one at 11:30pm)Email in ~12 min overnight (median, 9/13–10/13); texts wait for the 8am window; staff first outbound 94 min. Since 10/13, customers who write back 9pm–9am get an answer in ~5 min
Coverage96–99% of Internet leads81% (by design skips leads a rep is already working, recent buyers, credit apps)
Replies4.2% strict at Manor (Jul–Aug census); not logged in VS at Austin/Buda10.3% strict, same census; first-touch 23.6% attributable
Appointments3–6 AI-set a month at Manor; sister stores unread7 strict "assisted" in 90 days; setter not recorded in VS
VS clockStops it at Austin (tile looks great); not at BudaDoesn't stop it
Errors seenSTOP not honored for 5 days · quoted $1,232 over the feed · deposit-hold answer · same copy by text and email at once · 23 messages to one silent customer · notes cut off at ~450 charactersTold 8 customers a sold unit was "still here" (14 sends; sold-unit gate added 10/15, earlier sends being corrected) · cadence inside live rep threads 26.6% → 0 after the 10/14 gate · trade pitch after "no trade" 2 of 1,331
GuardsNo master off, no quiet hours, no daily cap. Per-source switches (it texted anyway). 13-step cadence to day 51Opt-out and STOP block · 120-day recent-buyer gate · no holds or deposits · no rates or negotiation (0 breaches in 391 sends since 10/10) · never names a sister store · steps back when a person is talking · quiet hours

Both AIs on the same lead is the worst case: 78 near-simultaneous double texts at Manor in Jul–Aug and 4 duplicate customers in Sept–Oct. Whatever the group decides: one AI per lead.

The honest read

Impel buys speed and coverage. Nobody can show it sells cars.

What Impel is worth

  • A first touch in under 4 minutes, 24 hours a day, on nearly every Internet lead. At Manor: 95% reached inside an hour with it, 70% the month after it left.
  • A safety net under slow staff: at Austin it was the only contact for 65 leads (8.7%).
  • A clean VS response tile, where it posts the clock-stop.

What it isn't

  • No sales lift at Manor: sold per Internet lead 12.4% with Impel alone vs 12.3–12.8% without.
  • No close edge across stores: Austin 12.4% = Manor 12.4%; Buda 6.9%.
  • Fewer replies than our machine (4.2% vs 10.3%), and errors our guards would have blocked.
  • The tile it improves hides how slow people are.

Where Manor's machine stands

  • Matches Impel's speed in store hours. After hours it was 4 hours when Chloe left; now 12 min by overnight email, with texts held for 8am. Overnight replies since 10/13.
  • Has guards Impel doesn't. Sold-unit error class closed 10/15 (gate at every send); earlier sends being corrected.
  • Also no proof it sells cars. Same standard for both.
For the group conversation

Three options

1 · Keep Impel at Austin and Buda

  • Cost: ~$6,000 / mo, $72K / yr.
  • Get: 24/7 first touch; no change for staff.
  • Risk: no proven lift; masks staff speed; guard gaps (STOP, deposits, price).
  • If yes: turn on Buda's clock-stop or turn off Austin's, so the tiles match; get Impel's dashboard export monthly; cap the cadence.

2 · Replace with Manor's machine

  • Save: up to ~$72K / yr, less AI usage (not measured yet).
  • Get: one system, one set of guards, more replies.
  • Needs: per-store setup (store name, inventory feed, reps, send identity, VS lanes) and a ruling: today the machine writes only on Manor records.
  • Risk: after-hours texting waits for 8am; one person maintains it.

3 · Run both, as a real A/B

  • One store or both, leads split 50/50 by lead number: half Impel, half ours. Never both on one lead.
  • 90 days plus 30 to let sales land.
  • Same scorecard, set before day 1 (next slide).
  • Decide on the result, not the dashboard.

Manor's exit, still open

Will's call stands: Manor stays off Impel (8/21, 10/13). Still owed by Impel, in writing: lead nurture off account-wide · billing end date and last invoice · final data export · Impel media off our vehicle pages. Website chat ("Chat-Impel") still sends leads (latest 10/14): keep it or cancel it?

The sister-store call is the group's.

What a fair test needs

If we test it, test it so the answer counts

  • Random split by lead number (odd / even), same sources, same reps. Not by store, not by month.
  • One AI per lead. The other is hard-blocked on that lead.
  • Same clock rule: both post the VS clock-stop or neither does.
  • Score on people outcomes: booked, shown and sold per 100 leads at 30 and 60 days; first person touch; opt-outs per 1,000; wrong-info sends (sold unit, price, hold, rate).
  • Log everything: every send from both, Impel's monthly export, both bills, appointment setter.
  • Stop rule: any STOP ignored or deposit/rate promise ends that arm.

How big it has to be

Setup, 90 daysLeads per armSmallest difference it can see
Austin only~815Shown 9 → 13 per 100 · close 12.4% → 17.4%
Austin + Buda~1,610Shown 9 → 12 per 100 · close 12.4% → ~16%

80% power, 5% false-alarm rate. Internet leads, last 30 days: Austin 543, Buda 532. A 1–2 point sales difference needs 6+ months.

Before day 1

  • Impel invoice and contract terms in hand.
  • Impel dashboard login or monthly export for both stores.
  • Our AI usage cost logged per message.
  • Group sign-off on the scorecard and the stop rule.