Every past-due account on the Peak platform gets worked two ways: AI text messages and phone calls. That gives us a clean side-by-side — same dealers, same customers, same 30 days. Here is how many people responded to each.
It holds at every dealer
This isn't one good dealer pulling the average up. At every dealer texting on the platform, the reply rate landed between 32% and 43%. At every dealer, phone calls reached the customer only 2% to 4% of the time.
| Last 30 days | By text | By phone |
|---|---|---|
| All dealers | 40% replied | 3% reached |
| Lowest dealer | 32% replied | 2% reached |
| Highest dealer | 43% replied | 4% reached |
The customers didn't change. The channel did.
Why a reply matters more than a dial
A call that rings out does nothing. A text reply is a customer in a conversation — and on the Peak platform, the AI is on the other end of it. In the same thread it can tell them what they owe, send a payment link, take a partial payment, or set a promise date inside your rules. Every note and promise goes back to your DMS, so your team isn't re-keying anything.
That's also why we don't treat texting as a blast. A one-way reminder that nobody answers just creates callbacks for your staff. The reply rate only pays off if something on the other end can finish the conversation.
How we measured it
- Source: the Peak platform's own records, across our dealer clients, for the 30 days ending October 30, 2026.
- Text reply rate = texts customers sent back ÷ texts we sent. Rounded.
- Phone reach rate = outbound calls that got the right person on the line ÷ outbound calls dialed.
- DMS posting = notes the platform posted to the dealer's DMS without a person touching them. A handful of failures rounds to 100%.
Earlier case study: when Peak took over a dealer's collections desk
Before our texting platform launched, a multi-rooftop dealer moved its collections operation to Peak. This is a straight read of the dealer's own records, including the parts that don't flatter us.
The headline
Across the full portfolio, total collections rose from $84,994 per day to $91,448 per day — up 7.6%. That happened while the dealer's in-house collector count fell from 25 active IDs to 20, and Peak held steady at 8.
The more interesting number is the phones. Answered inbound calls went from 153 per day to 310 per day, a 103.2% increase — against an inbound volume increase of only 59.1%. In other words, the calls were already coming; a much larger share of them started getting answered.
| Measured per day | Before (Jul 1–16) | After (Jul 17–25) | Change |
|---|---|---|---|
| Total collections | $84,994 | $91,448 | +7.6% |
| Total payments | 229.1 | 245.0 | +6.9% |
| Average payment size | $371 | $373 | +0.6% |
| Inbound calls received | 285 | 454 | +59.1% |
| Inbound calls answered | 153 | 310 | +103.2% |
| Answered share of inbound | 53.6% | 68.4% | +14.8 pts |
| Missed-call rate | 12.5% | 3.8% | −8.6 pts |
| Abandonment rate | 12.9% | 8.5% | −4.4 pts |
The improvement came from more payments, not bigger ones. Payment count rose 6.9%; average payment size moved 0.6%.
That distinction matters. A jump in average payment size usually means a few large payoffs landed in the window and the "result" is noise. A jump in payment count means more customers paid — which is what a collections operation is actually supposed to produce.
Did Peak replace the in-house output it displaced?
This is the question every dealer should ask, and most case studies skip. The in-house team shrank during this period. If Peak simply absorbed work that in-house collectors used to do, total collections would be flat and the whole exercise would be theater.
On the agent-attributed payment platform, here's how the two groups moved:
| Team | Before / day | After / day | Difference | Change |
|---|---|---|---|---|
| Peak | $9,634 | $14,033 | +$4,399 | +45.7% |
| In-house | $19,674 | $18,551 | −$1,122 | −5.7% |
| Combined | $29,308 | $32,584 | +$3,277 | +11.2% |
Peak's daily increase was roughly four times larger than the in-house decline. Net of both movements, agent-attributed collections rose 11.2% per day and payment count rose 8.7% per day — with five fewer active collector IDs across the combined operation.
Efficiency: same eight people, materially more output
Peak's headcount on the account did not change. Eight active collector IDs before, eight after. What changed is what those eight produced.
| Per recorded agent-day | Before | After | Change |
|---|---|---|---|
| Answered inbound calls | 21.4 | 45.1 | +110.8% |
| Total inbound calls | 29.3 | 52.5 | +79.0% |
| Talk minutes | 87.8 | 130.1 | +48.2% |
| Ring-no-answer events | 6.6 | 4.8 | −28.2% |
| Dollars per active collector | $1,204 | $1,754 | +45.7% |
| Payments per active collector | 3.07 | 4.92 | +60.1% |
Looking at the whole operation rather than just Peak: active collector IDs across both teams dropped from 33 to 28, while dollars per active collector per day rose from $888 to $1,164, up 31.0%.
Where the money came in
Self-service payment channels moved too, which matters because those payments aren't attributed to any individual agent — so they can't be explained away as one team's numbers being reclassified as another's.
| Self-service channel, per day | Before | After | Change |
|---|---|---|---|
| Customer digital | $40,503 | $41,360 | +2.1% |
| Cash location | $11,966 | $12,963 | +8.3% |
| AutoPay enrollment | $2,992 | $4,213 | +40.8% |
| Agent-assisted | $226 | $329 | +45.5% |
The AutoPay movement is the one worth watching. Getting a delinquent customer onto recurring payment is the single most durable thing a collector can do, and it doesn't show up in this month's collections — it shows up in next quarter's roll rate.
The trade-off we're not going to hide
The operation shifted decisively from outbound dialing to inbound coverage. Over the same window, total collections per day rose 7.6% and self-service collections rose 5.7%. Those things happened together.
We are not claiming outbound calling has no value — the payment records don't carry complete call-to-payment attribution, so nobody can honestly make that claim from this data. What the records do show is that a 91% reduction in outbound volume did not reduce the daily collection run rate during the measured period. For a dealer currently spending most of its collections labor on outbound dialing, that's worth sitting with.
Who actually produced
Ranked by paid transactions on the agent-attributed platform after the switch, Peak collectors held 7 of the top 10 positions. Peak's top performer ranked second overall with $23,287 across 71 payments.
One honest note on that ranking: Peak's average payment size ran lower than the top in-house collectors ($328 vs. $746 for the number-one seat). Peak was working a higher volume of smaller payments. Whether that's better depends entirely on which accounts each group was assigned — which the ranking does not adjust for.
Methodology and limits
Sources: the dealer's payment-platform paid-transaction exports and completed-payment exports (July 1–22 and July 19–25 pulls); phone-system daily statistics, user statistics, call logs and group statistics for July 1–25, 2026. Peak did not supply the underlying data; it came out of the dealer's own systems.
Why per-day averages: the before window is 16 days and the after window is 9. Raw totals would be meaningless, so everything is normalized per calendar day.
- The after period is nine days. That is a short window and results should be re-measured as more weeks accumulate. We publish it because it is what exists, not because it is definitive.
- Collector counts are inferred from names with at least one paid transaction, not from an official payroll roster.
- Self-service payments are mostly not attributed to an individual agent, so they cannot be reliably split between Peak and in-house collectors.
- The two windows contain different weekday mixes. Per-day averaging corrects for the different day counts but does not remove every calendar effect.
- Correlation, not proof of cause. The switch and the improvement occurred together. No single-variable experiment was run, and we're not going to pretend otherwise.
What this should and shouldn't tell you
It should tell you that a specialized collections team, given the phones and the delinquency list, moved a real portfolio's numbers in a measurable direction within nine days — and that the biggest single lever was answering calls that were already coming in.
It should not tell you that you'll see 7.6%. Your portfolio, your delinquency profile, your current answer rate and your existing team are different. If your inbound is already well covered, the phone-side gain here simply won't be available to you — and we'd rather establish that on a 15-minute call than four months into an engagement.
Want to see it on your numbers?
Fifteen minutes on Teams. We'll look at your delinquency together, show you real AI text conversations, and tell you honestly whether there's room to help.
Book a 15-minute walkthrough