Credible - Abandoned Application Recovery & Form Conversion Proposal

Problem

Hundreds of thousands of starts. No way to bring them back.

Credible's personal loan pre-qualification form is fully self-serve. There's no outbound recovery, no alternative channel, and no second touch. Users who drop off are gone -- and you don't know why.

100,000s
Monthly form abandonments
Each is a borrower with demonstrated intent. Recovering even a small share is a meaningful revenue lift on the existing pipeline.

0

Recovery mechanisms today

No outbound exists for partial-form users. Every abandoned PQ is a lost lender match.

"We get hundreds of thousands of users that we don't even see because they abandoned our form every month. If we can capture a small percentage of that, the revenue opportunity for us is fairly meaningful."

-- What we've heard from Credible

Where users drop off

From form analysis -- the three steps where trust and clarity break down. Phase 1 of the pilot will measure each empirically.

1

SSN entry

The biggest trust barrier. Despite "won't affect your credit score" messaging, users are unsure whether this triggers a hard pull, whether data is sold to lenders, or whether this is even legitimate.

Highest risk

2

Account creation wall

After 11 steps, users hit a signup gate before seeing rates. Feels like a bait-and-switch on a "check your rates" promise.

High risk

3

Income disclosure

Ambiguity (bonuses? rental? alimony?) plus the sensitivity of sharing earnings online causes hesitation -- a recurring theme for self-employed and freelance users.

Medium risk

Solution

Outbound recovery for partial-form users -- text first, voice as backup

When a user starts the PQ form and doesn't finish, Lorikeet reaches out via operational SMS within a defined window. The agent picks up where they left off, asks only the remaining questions, and submits via API. If they don't reply, an outbound call follows. The web form is untouched.

🔒

Operational, not promotional

Every message is framed as a follow-up on an incomplete process, not marketing outreach. The agent is informational only -- it never advises on rates, recommends a lender, or crosses into financial guidance. We start by directing users back to the website to complete; conversational completion is gated behind Credible's compliance review.

"The beauty of Lorikeet is you can just outbound and say hey, we noticed you got stuck. What help do you need? Let them tell you, and then the AI goes back and forth with them to solve it."

-- How Lorikeet thinks about outbound recovery

How it works

User abandons → wait window → SMS nudge → optional call → back to form, or complete in-conversation (subject to compliance sign-off).

Outbound recovery flow

Triggered when partial-form data is captured but the PQ isn't completed within a configurable window.

Cadence

All timing below is a starting hypothesis -- final cadence is set with Credible's compliance and ops teams during Phase 1.

Step	Channel	Timing (to validate)	Message
1. Initial nudge	Operational SMS	Same day	"Hi [name], you're almost done checking your personalized loan rates on Credible. Tap this link to pick up where you left off -- or reply here and I can help."
2. Follow-up call	Outbound voice	+24h if no reply	Voice agent introduces itself as Credible, acknowledges they started checking rates, and offers a deep-link back to the form. Conversational completion is enabled only after Credible compliance sign-off.
3. Final reminder	Operational SMS	+48h	"No rush -- your progress is saved. When you're ready, pick up right where you left off and see your rates in under 2 minutes."

📊

Why this differs from a generic reminder email

Standard re-engagement sends a one-shot link back to the form. Lorikeet has a back-and-forth conversation, and once compliance allows it, can complete the PQ in-channel -- skipping the questions you already have. A 12-step form becomes a 3-question completion. Credible has already had hands on the prototype and confirmed the experience feels right -- the question now is whether it lifts completed PQs at scale.

Out of scope for the pilot

Two adjacent ideas surfaced on Feb 25 -- an in-form chat widget for live assistance at high-friction steps (SSN, account creation), and bottom-of-funnel re-engagement for users who got rate offers but didn't pick a lender. Both build on the same Lorikeet workflow and are natural Phase 2+ extensions once the outbound recovery loop is live and measured.

Pilot plan

Three phases. Instrument, run, decide.

A design-partner pilot designed to be low-cost to start and quick to learn from -- matching what Credible has asked for: validate impact before scaling.

Phase 1 -- Instrument

Baseline + integration

Stand up step-by-step analytics on the PQ form to confirm the actual drop-off curve, agree the user cohort, and connect the trigger from Credible's system to Lorikeet (capture phone earlier in the form, or use the partial record's contact details once available). Lock down the operational message templates with Credible compliance.

Phase 2 -- Run

Live outbound recovery on a slice of traffic

Deploy the SMS + voice recovery loop on a measurable share of abandons. Crawl phase: agent directs back to the saved form. Walk phase: opens up conversational Q&A on objections (security, freelance income, rate confusion). Run phase: conversational completion of remaining PQ fields, gated on Credible compliance approval.

Phase 3 -- Decide

Read the lift, scale or stop

Compare completion and revenue-per-start for treated vs. control cohorts. Tune the SMS/voice mix and timing. Decide whether to scale across personals, extend to student loans (cosigner prompt), or layer in bottom-of-funnel re-engagement for users who saw rates but didn't pick a lender.

Measurement

What we're testing for

We don't want to put fake numbers on the page. Instead -- the questions Phase 1 + 2 should answer, and the metrics we'll watch.

Question	Metric	Baseline	How we'll measure
Where do users actually drop?	Completion rate per form step	TBD in Phase 1	Step-by-step PQ analytics
Do they engage when we reach out?	SMS reply rate; call connect rate	n/a (no recovery today)	Lorikeet conversation analytics
Do engaged users finish?	Completion rate of recovered cohort	n/a	PQ completion attributed to outbound trigger
Does it move the bottom line?	Incremental completed PQs per 1,000 abandons; revenue per recovered PQ	TBD with Credible	Treated vs. control cohort, agreed in Phase 1
Where does compliance let us go?	% of recovered PQs completed in-conversation vs. via deep link back to form	0% in crawl phase	Channel attribution; Credible legal sign-off log

📈

Why this is worth running, even small

Hundreds of thousands of users abandon the form every month and none are touched today. A pilot on a single-digit-percent slice of that volume is enough to read whether outbound recovery moves completed PQs without committing engineering or marketing budget upfront. Lorikeet absorbs the build cost; Credible commits compliance review and a measurement window.