SYNTHETIC DATA · NOT REAL RESULTS

Harborline Home Loans · Account health

420 simulated lead conversations, 2026-09-28 to 2026-10-25. Generated by a seeded script with planted patterns, to show how I'd run a weekly account review: read the numbers, then decide what to change in the agent.

Leads

420

4 weeks

Replied

70%

295 of 420

Qualified

51%

150 of 295 who replied

Transferred

90%

135 of 150 qualified

Loan officer connected

88%

of live transfers

Opted out

10%

of those who replied

Speed to lead

20s

median first text

3 recommended agent changes

Reframe the credit range question

14% of borrowers asked about credit range stopped replying, vs 4% at a typical question.

29 of 201 borrowers stopped at this question (14%).
Median drop-off at the other 6 questions: 4%.
Credit is the most sensitive question we ask; borrowers often assume it triggers a credit pull.

Change: Ask credit last, lead with "no credit check, just your best guess," and make "not sure" an easy answer. Allow a transfer without it, since the loan officer will review credit anyway.

Config change

qualification:
  fields:
    # move credit_band after occupancy and timeline
    - id: credit_band
      required: false   # loan officer confirms credit
      ask: >-
        Lead with "this isn't a credit check, just your best guess," and offer
        "not sure" as an answer.

How I'd validate: Re-run the eval suite (recall and borrower experience must hold), then watch credit range drop-off weekly. Target: at or below 6%.

If drop-off here matched a typical question, about 18 more qualified leads over these 4 weeks (synthetic estimate).

Stop promising a 5-minute call on weekends

Only 60% of weekend transfers reached a loan officer, vs 96% on weekdays.

12 of 30 weekend transfers did not connect within the promised window.
75% of borrowers whose transfer didn't connect ended the conversation frustrated.
The agent's transfer message says "usually within 5 minutes," which is untrue on weekends.

Change: On weekends, offer a scheduled call instead of a live transfer, or tell the borrower the honest wait. Share the connect-rate gap with the lender's sales manager: it may be a staffing fix, not an agent fix.

Config change

handoff:
  live_transfer_available: business_hours
  business_hours: { days: [mon, tue, wed, thu, fri], start: "08:00", end: "21:00" }
  after_hours_guidance: >-
    Offer a call at a time that works; don't promise an immediate callback.

How I'd validate: Add a weekend persona to the eval suite (expected: callback, clear next step). Then track weekend connect rate and frustrated endings weekly.

About 3 fewer broken "5-minute" promises per week (synthetic estimate).

Use a source-aware opener for comparison site leads

22% of comparison site leads opted out, vs 6% from the lender's own web form.

30% of these leads asked some form of "how did you get my info?" (web form: 5%).
They came through a third-party site, so a Harborline text feels unexpected. The opener references "the inquiry you just submitted" without saying where.
Qualification rate for this source: 43% (web form: 51%).

Change: Pass the lead source to the agent and name it in the first message ("you requested refinance quotes on [partner site] a moment ago"). Keep the STOP reminder.

Config change

channel:
  trigger_by_source:
    rate_comparison_partner: >-
      The borrower requested refinance quotes on a rate-comparison site a moment
      ago and chose Harborline as one of the lenders to contact them. Say so.

How I'd validate: Add a partner-sourced persona to the eval suite; then watch opt-out and "how did you get my info?" rates for this source weekly.

If opt-outs fell to the web-form rate, about 13 more conversations kept over these 4 weeks (synthetic estimate).

The evidence

Qualification funnel

Share of borrowers who replied, by how far they got

Replied295 · 100%

Loan goal236 · 80%

Home value222 · 75%

Loan balance213 · 72%

Current rate201 · 68%

Credit range172 · 58%

Occupancy168 · 57%

Timeline166 · 56%

Qualified150 · 51%

Transferred135 · 46%

Officer connected119 · 40%

Show data table

Replied	295	100%
Answered: Loan goal	236	80%
Answered: Home value	222	75%
Answered: Loan balance	213	72%
Answered: Current rate	201	68%
Answered: Credit range	172	58%
Answered: Occupancy	168	57%
Answered: Timeline	166	56%
Qualified	150	51%
Transferred	135	46%
Officer connected	119	40%

Drop-off by question

Of borrowers asked each question, the share who stopped replying

Loan goal4%

Home value6%

Loan balance4%

Current rate6%

Credit range14%

Occupancy2%

Timeline1%

Show data table

Loan goal	9 / 245	4%
Home value	14 / 236	6%
Loan balance	9 / 222	4%
Current rate	12 / 213	6%
Credit range	29 / 201	14%
Occupancy	4 / 172	2%
Timeline	2 / 168	1%

Live transfers that reached a loan officer

Within the promised 5 minutes

Weekdays96%

Weekends60%

Objections raised

Share of conversations; hover for qualification impact

Just shopping rates27%

Busy right now12%

How did you get my info?11%

Already have a lender7%

Show data table

Objection	Share	Qualified if raised	If not
Just shopping rates	27%	37%	56%
Busy right now	12%	17%	56%
How did you get my info?	11%	27%	54%
Already have a lender	7%	33%	52%

Opt-out rate by lead source

Of borrowers who replied

Own web form6%

Comparison site22%

Refinance ad2%

"How did you get my info?" by lead source

Of borrowers who replied

Own web form5%

Comparison site30%

Refinance ad2%

Show data table

Source	Leads	Opted out	Asked	Qualified
Own web form	235	6%	5%	51%
Comparison site	120	22%	30%	43%
Refinance ad	65	2%	2%	65%

Download the synthetic dataset: CSV. Every record is flagged synthetic.