A lead-by-lead effort audit of the Zolve loan advisors — every dial, every connect, how long they actually talked, what was discussed, and the red and green flags on each case. Built to answer one question management asked: leads assigned are many, connected calls are few — why?
When any of the three picks up the phone they connect at a similar rate — John 38%, Swatika 40%, Tarun 44% per dial. The difference is how often they dial. John makes 3.4 calls per lead (median 2) and has reached 60% of his book; Swatika makes 1.5 (median 1) — she rings most leads once and moves on — and has reached only 44%.
Tarun is the standout. On a smaller book (161 leads) he dials 2.4×/lead, has reached 66% — the highest of the three — and holds 47 five-minute conversations (29% of his leads, also the highest). His calls log to the CRM without his name (they show as "Unknown"); the figures here are corrected by matching his dialler line. The clear gap now is Swatika's dialling volume, not skill.
Top of the funnel (assigned → connected) is from the Nbyula CRM; the bottom (logged with bank → disbursed) is joined from Zolve's shared tracker by phone. Split into the AI's three tags — Doable/Fast-Track, Borderline, Not Doable — plus untagged, per advisor.
268 leads · reached 162 · logged 12
263 leads · reached 115 · logged 10
161 leads · reached 107 · logged 0 — calls log as “Unknown”, corrected by dialler line
Read straight off the funnel above. Each leak pairs the discussion to raise on the call with the resolution to agree. Anchor number: from 692 shared leads, 1 disbursed — the money is being left in leaks 1 & 2.
663 were dialled at least once, but only 231 got more than 2 attempts (60 got >5, 27 got >7). So ~430 leads got just 1–2 rings and were effectively parked. Swatika is the sharpest case (median 1 dial / lead).
What's the minimum-attempt rule per lead before it's set aside?
Agree a minimum cadence — 5–6 attempts across different days and time-slots before a lead is dropped. Let the AI run the persistence tier (it dials ~10× and reaches ~78%) and hand RMs the leads that actually pick up.
The biggest leak in the funnel: 110 leads had a genuine 5-min+ conversation, only 22 reached “logged with a bank” (−80%). The reason is overwhelmingly documents — 63 of 111 worked leads sit at “Doc Pending”, waiting on the student.
What happens to a lead between a good call and a logged file? Where do the 88 that talked-but-didn't-log go?
Put automated document collection + daily follow-up in the lead's Space (Nbyula's loan-docs flow) so the checklist goes out the moment the call ends and chases the student until complete — the RM steps in only when docs are in. Set an SLA: file logged within X days of the qualifying call.
By tag: Doable / Fast-Track — 72 assigned → only 3 logged; Borderline — 169 assigned → 7 logged. These are the leads most likely to clear a lender, and they're not reaching one.
Are Doable / Fast-Track leads being prioritised, or worked in the same undifferentiated queue as everything else?
Work the funnel by tag — Doable / Fast-Track first, same-day, target logging every one within 48h. Nbyula can push these to the top of each RM's list daily.
Tarun: 42 Not-Doable leads assigned, 34 reached, 21 held 5-min+ conversations. That's deep effort on leads unlikely to clear a lender.
Should Not-Doable leads get full-length calls, or a quick check + polite close?
Agree a light-touch path for Not-Doable (one verification call, then park / recycle) so that time redeploys onto Doable / Borderline. Use the AI tag to route effort, not spread it evenly.
Zolve's tracker jumps straight from “Logged” → “Disbursed” with nothing in between — no sanctions, PF-confirmations, or where files stall with the lender. Only 1 disbursement from 692 shared leads, and 28 sitting at “Login” with no visibility into their status.
Where are those logged files actually stuck — lender side, or documents?
Add “Sanctioned” and “PF Confirmed” columns to the shared tracker so the deep funnel is measurable, and we can jointly chase the logged files.
Green = the best of the three on that metric, red = the weakest. The hero tile is dials-per-lead — the effort lever.
Of every lead assigned, how many survive to each depth of contact. The drop from "assigned" to "dialled" is pure effort; the drop after "reached" is conversation quality.
One block = one lead. The redder the block, the less contact ever happened.
The very top of the funnel — call attempts per lead, however far it later got. This is the purest read on effort: a lead that was dialled once and dropped had almost no chance to convert.
| Advisor | 0 | 1 | 2 | 3 | 4 | 5 | 6–10 | 11+ | Leads |
|---|---|---|---|---|---|---|---|---|---|
| John Stanley | 11 | 70 | 54 | 35 | 25 | 21 | 48 | 4 | 268 |
| Swatika Reddy | 16 | 163 | 53 | 21 | 6 | 1 | 3 | 0 | 263 |
| Tarun Tej | 2 | 46 | 46 | 43 | 11 | 8 | 5 | 0 | 161 |
John left 81 of his 268 leads (30%) with one dial or none, Tarun 2 of 161 (1%) — but Swatika left 179 of 263 (68%) with a single dial or none, and her median lead gets just one ring. A single unanswered ring is not a worked lead; the AI-agent comparison shows what happens when every lead instead gets dialled ~10 times: reach jumps to ~78%.
| No activity for N days | 576 |
| Only ever reached for under a minute | 131 |
| First real contact took N days | 106 |
| Tagged Not Doable (lender fit unlikely) | 70 |
| N attempts, never once answered | 61 |
| No attempt to close or ask for commitment | 60 |
| Never dialled once since assignment | 29 |
| Tagged Not Interested | 22 |
Click any lead to open its effort log, what was discussed on the best call, and its flags. Sorted by advisor, then longest conversation first.