Running IPO Season at Scale — Bulk Applications for Broking Desks
Collecting bids from thousands of clients across a branch network is a data problem before it is a market problem. Here is where desks lose applications, and how the workflow should be structured.
For an individual, an IPO application is a three-minute task. For a desk placing three thousand across forty branches in a three-day window, it is a data pipeline with a hard deadline and no retry after the close.
The failure modes are the same ones individual investors hit. What changes is that they arrive in batches, from a single upstream cause.
The shape of the problem
A retail investor with a wrong DP ID loses one application. A desk with a stale client master file loses every affected client at once — and finds out on allotment day, when nothing can be fixed.
That asymmetry is the whole argument for front-loading validation.
Before the issue opens
Reconcile client master data against depository records. DP ID, client ID, PAN and account status. Anything that has changed since the last reconciliation, or any account that has gone dormant or failed re-KYC, is a rejection waiting to happen.
Check PAN-to-demat mapping. The registrar matches these exactly. A client whose PAN was updated in one system and not the other will fail.
Flag inactive accounts. Dormant demat accounts and incomplete re-KYC block the credit even when everything else is correct.
Confirm bank and UPI mappings are in the client’s own name. Third-party funding is rejected on detection, and joint-family arrangements are where this most often goes wrong.
This work is not issue-specific. Done once a quarter, it removes most of the rejection surface for every issue that follows.
During the bidding window
Collect early. The last afternoon is when mandates fail, platforms slow and banks time out. A desk that has 80 percent of its bids in by the end of day two has room to fix problems; one that collects on the final afternoon does not.
Deduplicate on PAN, centrally. Two branches submitting for the same client causes both applications to be rejected at the registrar. This must be caught before upload, because after it nothing can be done.
Validate category and quantity at entry. Applications above two lakh rupees are not retail. Quantities that are not whole lots are invalid. Both are cheap to catch at the point of capture and impossible to fix later.
Track mandate status as a live worklist. This is the highest-value operational activity in the window. Every bid sits in one of three states — mandate not sent, sent and unapproved, approved and blocked — and only the third is a real application. The second is where applications die.
Working the mandate list
The client’s experience is that they applied. The system’s view is that nothing is funded. Closing that gap is a phone-call operation, and it needs to be structured:
- Pull unapproved mandates at least twice a day, more on the final day
- Sort by application value, so the largest exposures are worked first
- Tell clients to look in the pending mandates section of the UPI app, not the notification tray
- Confirm the client has the full cap-price amount free, not the amount they expect to pay
- Where mandates repeatedly fail, move the client to bank ASBA rather than retrying
Reconciliation before the close
An hour before your internal cut-off, three numbers should agree: bids collected, bids uploaded to the exchange, and mandates approved. Any gap between the first two is an upload failure. Any gap between the second and third is an unfunded application.
Both are fixable at that moment and not afterwards.
Allotment day
Check in bulk, not one PAN at a time. Registrar portals are built for individuals and are under heavy load on allotment evening. A desk needs a list of PANs in and a report out.
Read the rejections as a group. Individually a rejection looks like bad luck. In aggregate, ten clients rejected for the same reason is a data defect in your master file, and it will repeat on the next issue unless you fix the cause.
Reconcile unblocking. Clients notice a lien that has not been released, and it lands on your desk regardless of whose fault it is. Knowing which accounts are still blocked before the client calls is the difference between a support case and a complaint.
What good looks like
| Stage | Manual desk | Structured desk |
|---|---|---|
| Pre-issue | Client data assumed current | Reconciled against depository records |
| Collection | Forms and messages per branch | Single capture with validation at entry |
| Deduplication | Discovered at the registrar | Enforced on PAN before upload |
| Mandates | Client is told to approve | Live unapproved list, worked twice daily |
| Allotment | One PAN at a time | Bulk check, rejection reasons grouped |
| Root cause | Rediscovered every issue | Fixed once in the client master |
None of this affects allotment probability — that is arithmetic the desk cannot influence. All of it affects how many of your clients are actually in the draw, which is the only part within your control.
Frequently asked questions
What is the single biggest cause of lost applications at scale?
Unapproved UPI mandates, by a wide margin. The bid is on the exchange, the client believes they have applied, and nothing is blocked. At scale this is a tracking problem: you need a live list of unapproved mandates and someone working it.
How do we stop duplicate applications across branches?
Deduplicate on PAN centrally before upload. Deduplication at the registrar rejects both applications, including the one that would have been allotted, so catching it on your side is the only useful place to catch it.
Can we submit applications on behalf of clients?
You can collect and upload bids as a syndicate or registered broker, but funds must be blocked against an account in the client's own name. Third-party funding is prohibited and is rejected on detection.