Every education-loan lead Nbyula has assigned to Dimple Soni or Mahek Sharma since the 11 Aug handover, and exactly what has happened to each one since. Yash Naidu's book is tracked separately — nothing here is his. This is a days-old book, so it is a running record of how fast the new advisors pick leads up, not a verdict on their effort.
Method note: call effort is attributed by the advisor's phone
number, not the crm_user field — that field is stamped on fewer than half these calls (92 of 203)
and filtering on it undercounted Dimple and Mahek roughly 4× in the version of this page published on 14 Aug.
In the 36 days since the handover the two advisors have placed 368 dials across 183 of 212 leads, reached 122, and held 62 conversations of five minutes or more — 11h 14m of talk time on a book with a median age of 23 days. 87 leads have already been moved along UniCreds' own pipeline. Only 28 leads have had no contact in either direction. The one thing to fix: 21 inbound calls from leads went unanswered — those are people trying to reach you.
Dials, connects and talk time are counted from calls through the Nbyula dialer, attributed by the
advisor's phone number — the crm_user field is stamped on only a fraction of rows and would
undercount Dimple and Mahek by roughly 4×. Calls UniCreds places from its own systems are still invisible, so
"never dialled" means "no attempt visible in the shared system".
The full twelve-step funnel — including each advisor's own book and each verdict tag — is in The full funnel, assignment to disbursal further down. This section covers what UniCreds' pipeline stages say about the book right now.
212 assigned → 183 dialled → 122 reached, on a book with a median age of 23 days. That is a fast start, and the funnel below is a running record rather than a verdict. "At a lender stage" counts UniCreds' own Login / Docs Pending / Sanctioned tags, not just the legacy conversion tag.
These six stages went live on 11 Aug and Dimple and Mahek are actively setting them — 87 of 212 leads carry one. This is the most reliable progress signal on this dashboard, and the first one we can read directly instead of transcribing it off UniCreds' dashboard by hand.
Everything below follows the same 212 leads all the way into UniCreds' own lender pipeline, split
by the three loan verdict tags. Apart from Logged in, the last stages do not exist in the Nbyula
CRM and are read from UniCreds' dashboard export.
You asked how many of these are Pro / Premium / Ultimate leads so they could be excluded.
Checked against entitlements.activePlan.planName, joining lead → gamePlayer via the id
embedded in each lead's phone.<id>@mail.nbyula.com address, and by real e-mail for the seven
that have one. All 213 resolved — none was left unmatched.
Your premise holds almost exactly. Of the 213 leads on the
Dimple/Mahek book, exactly one carries a Pro / Premium / Ultimate plan, and it is
Shouvik Mitra on an internal @navigus.in address — a test record, not a customer. The
other 212 have no entitlement document at all. These are genuinely loan-funnel leads and effectively nothing
needed removing.
Reaching people is not the constraint — 122 leads connect (58%) and 62 hold a conversation of five minutes or more. Persistence is. Just 46 leads (22%) are ever dialled more than twice and 7 more than five times. Nearly every lead who missed the first two calls was never called again.
At the far end 12 leads reached Logged in, 5 Sanctioned, 3 PF Confirmed and 2 Disbursed — 1% end-to-end on a book whose median lead is 24 days old.
Width is proportional to lead count, so the taper is the leak. >2 attempts means 3 or more dials, >5 means 6 or more, >7 means 8 or more. The last four steps are UniCreds' own pipeline (Logged in → Sanctioned → PF Paid → Disbursed), counted cumulatively — a disbursed lead is counted at all four — and read from their dashboard, because those stages do not exist in the Nbyula CRM.
Read the blue and amber phases as parallel, not sequential. A lead can connect on dial one and never reach ">2 attempts", so the attempt steps and the connect steps are two different axes over the same 212 leads. The strictly-ordered version is the terminal-state chart below.
Split into the three chains that are genuine subset relations. Each starts from all 212 assigned leads, because "Connected" is a subset of assigned, not of ">7 attempts" — a lead can connect on the first dial, and one lead reached Logged in with no dial in our system at all. The highlighted row is the biggest drop within each chain.
| Transition | From | To | Carried | Lost | Leak | Of cohort |
|---|---|---|---|---|---|---|
| Dialling effort | ||||||
| Assigned → Attempted | 212 | 183 | 86% | −29 | 14% | 86% |
| Attempted → >2 attempts | 183 | 46 | 25% | −137 | 75% | 22% |
| >2 attempts → >5 attempts | 46 | 7 | 15% | −39 | 85% | 3% |
| >5 attempts → >7 attempts | 7 | 2 | 29% | −5 | 71% | 1% |
| Conversation | ||||||
| Assigned → Connected | 212 | 122 | 58% | −90 | 42% | 58% |
| Connected → Connected >1 min | 122 | 88 | 72% | −34 | 28% | 42% |
| Connected >1 min → Connected >5 min | 88 | 62 | 70% | −26 | 30% | 29% |
| UniCreds' own pipeline | ||||||
| Assigned → Logged in | 212 | 12 | 6% | −200 | 94% | 6% |
| Logged in → Sanctioned | 12 | 5 | 42% | −7 | 58% | 2% |
| Sanctioned → PF Confirmed | 5 | 3 | 60% | −2 | 40% | 1% |
| PF Confirmed → Disbursed | 3 | 2 | 67% | −1 | 33% | 1% |
The three loan tags you asked for, plus everything carrying none of them — Not Interested, Human Intervention, and leads the classifier has not reached yet. Watch the fourth card: it is the second-largest group on the book and it is producing outcomes.
Count, and share of that bucket's assigned leads. Shading follows the percentage.
| Bucket | Assigned | Attempted | >2 attempts | >5 attempts | >7 attempts | Connected | Connected >1 min | Connected >5 min | Logged in | Sanctioned | PF Confirmed | Disbursed |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Doable / Fast-Track | 23 100% | 20 87% | 2 9% | 0 0% | 0 0% | 17 74% | 12 52% | 9 39% | 4 17% | 2 9% | 2 9% | 2 9% |
| Borderline | 79 100% | 63 80% | 11 14% | 2 3% | 0 0% | 39 49% | 27 34% | 24 30% | 3 4% | 1 1% | 0 0% | 0 0% |
| Not Doable | 51 100% | 45 88% | 9 18% | 2 4% | 0 0% | 35 69% | 27 53% | 13 25% | 0 0% | 0 0% | 0 0% | 0 0% |
| Other / unclassified | 59 100% | 55 93% | 24 41% | 3 5% | 2 3% | 31 53% | 22 37% | 16 27% | 5 8% | 2 3% | 1 2% | 0 0% |
Strictly ordered, so every lead appears exactly once and the bars sum to the cohort.
The effort axis on its own. This is where the two advisors genuinely differ.
Dimple works the book deeper; Mahek works it once. 35 of Dimple's leads get three or more dials against 11 of Mahek's — yet the two reach an almost identical number of leads (62 vs 60) and hold the same number of deep conversations (31 vs 31). Mahek reaches the same result for far less dialling; Dimple's extra dials are going to leads who do not pick up.
The same twelve steps, one card per tag, per advisor. The drop column is blank on Connected and Logged with bank on purpose — both are subsets of assigned, not of the row above them, so a drop figure there would be meaningless.
Every Dimple/Mahek lead that reached UniCreds' pipeline, from their 16-Sep export. "Reached" is cumulative; "here now" is the current status.
A funnel built only on the Unicreds - * stages in the Nbyula CRM would show one
sanction and nothing else. Two of these leads carry no stage at all on our side.
Ajay Sakthivel reached Logged in with zero dials in the Nbyula dialer — a direct example of the standing caveat: calls placed from UniCreds' own systems are invisible here.
| Lead | Advisor (ours) | Bucket | UniCreds dashboard | Our CRM stage | Dials |
|---|---|---|---|---|---|
| Amol Dixit | Mahek | Doable / Fast-Track | Disbursed | Sanctioned | 1 |
| Jeeshan Shaikh | Dimple | Doable / Fast-Track | Disbursed | Processing | 1 |
| Ehtesham Ali Khan | Mahek | Other / unclassified | PF Confirmed | Login | 2 |
| Fida ur rehman Khan | Mahek | Other / unclassified | Sanctioned | no stage at all | 2 |
| Prayag | Dimple | Borderline | Sanctioned | no stage at all | 1 |
| Ajay Sakthivel | Dimple | Doable / Fast-Track | Logged in | no stage at all | 0 |
| Ashutosh Waghmare | Dimple | Borderline | Logged in | no stage at all | 1 |
| Devkinandan Mishra | Mahek | Borderline | Logged in | Contacted | 3 |
| Dhruv salwan | Dimple | Doable / Fast-Track | Logged in | no stage at all | 1 |
| Nilesh GUpta | Dimple | Other / unclassified | Logged in | no stage at all | 5 |
| ROHAN UPADESHE | Dimple | Other / unclassified | Logged in | no stage at all | 2 |
| Sheryl | Mahek | Other / unclassified | Logged in | Processing | 1 |
UniCreds' dashboard also credits a different agent on three of these (Devkinandan to "Sujal Thakur"; Rohan and Nilesh to Mahek where our CRM has Dimple). Cohort membership here always follows our assignment, which is what every other number on this page uses.
1. The book is not the constraint; the second call is. 58% of leads connect at least once, but 78% never receive a third dial. The funnel leaks at follow-up, not at reachability.
2. Doable / Fast-Track is small and is being worked properly. 23 leads, 20 attempted, 9 in a five-minute-plus conversation, 4 logged in with the lender, and both of the book's disbursals. The tag is doing its job.
3. Borderline holds the volume and the loss. 79 leads — the largest bucket — produced 3 logins, 1 sanction and no disbursal, and 16 of them have never been dialled at all.
4. Not Doable is absorbing real effort for nothing. 45 of 51 Not-Doable leads were dialled and 13 got a five-minute-plus conversation — more deep conversations than Doable/Fast-Track received — for zero lender outcomes. That is the clearest reallocation available.
5. Untagged leads are converting, and nobody is prioritising them. The 59 leads carrying none of the three tags produced 5 logins, 2 sanctioned and 1 PF-confirmed lead. Any prioritisation built on the verdict tags is blind to them.
Source: Nbyula CRM (argonaut_counselor_lead_relation, argonaut_leads_info,
argonaut_counsellors_call_history, entitlements), snapshot 16 Sep 2026, plus UniCreds'
own dashboard export of the same date for Logged in / Sanctioned / PF Confirmed / Disbursed. Call attribution is
by agent_number and was fully audited on 16 Sep 2026 — no UniCreds call is unattributed. Calls placed from
UniCreds' own systems never reach the Nbyula dialer, so "never attempted" means "no attempt visible in the shared
system".
| 0 | 1 | 2 | 3 | 4 | 5 | 6–10 | 11+ | Total | |
|---|---|---|---|---|---|---|---|---|---|
| Dimple Soni | 14 | 35 | 23 | 11 | 14 | 4 | 6 | — | 107 |
| Mahek Sharma | 15 | 59 | 20 | 8 | 2 | — | 1 | — | 105 |
| All | 29 | 94 | 43 | 19 | 16 | 4 | 7 | 0 | 212 |
The loan agent classifies each lead it reaches. "Never dialled" in the last column is the number in that bucket no UniCreds caller has ever attempted.
| AI loan verdict | Leads | Dialled | Reached | 5-min+ human call | 5-min+ AI call | Never dialled |
|---|---|---|---|---|---|---|
| Doable / Fast-Track | 23 | 20 | 17 | 9 | 23 | 3 |
| Borderline | 79 | 63 | 39 | 24 | 50 | 16 |
| Human Intervention | 7 | 6 | 3 | 1 | 4 | 1 |
| Not Doable | 51 | 45 | 35 | 13 | 17 | 6 |
| Not Interested | 8 | 8 | 3 | 1 | 3 | 0 |
| No AI verdict yet | 44 | 41 | 25 | 14 | 6 | 3 |
The classifier only went live in June 2026. The 44 unclassified leads are a median 30 days old — they are the legacy book, not rejects. They all arrived through the loan-eligibility form and carry a full underwriting questionnaire.
| Leads with a loan verdict the AI produced but nobody at UniCreds ever dialled | 26 |
| Doable / Borderline leads never reached by a UniCreds caller | 46 |
| Leads the AI held a 5-min+ loan conversation with, that UniCreds never connected to | 40 |
| Leads assigned 3+ days ago with no visible dial | 16 |
| Leads moved in UniCreds' pipeline but never dialled through the shared dialer | 2 |
| Leads where the lead rang in and the call went unanswered | 17 |
| Leads with zero visible contact in either direction | 28 |
| Leads dialled exactly once and dropped | 94 |
Sorted so the qualified, under-worked leads come first. Every row opens its Nbyula Space.