Financial software should make its reasoning available to the institution that relies on it.
Leadership paper / 01
Reconciliation engineering and ambiguity6 min read
A financial result needs an explanation
Abstract
A high matching rate can conceal the records for which several relationships remain plausible. Good financial software preserves the evidence needed to decide among them.
Aggregate throughput does not reveal which customer is waiting. Scheduling rules should express an agreed service policy rather than the influence of the largest import or loudest escalation.
03
Portability of financial-workflow evidence6 min read
A complete row count does not prove that a receiving team can understand the cases it inherits. Exit testing should examine meaning, evidence and authorised scope outside the supplier's application.
I lead a financial-software engineering team that must explain its decisions to the institutions buying the product and to the people who inherit its records.
I am CTO of Orevathi Settlement Systems in Surat. We build reconciliation and exception-workflow software for banks, lenders and other financial institutions. We do not hold their customers' money or make their settlement decisions. Our 28 institutional clients use the product within their own responsibilities. I lead 86 engineers through five heads and administer a ₹36 crore annual engineering budget. At March 2026 the company had 310 employees, with FY 2026 revenue of ₹182 crore.
I entered engineering through document-imaging equipment in 1995. Working on the gap between a page and its digital representation taught me to distrust an apparently complete result whose source could not be examined. I joined Orevathi in 2001, moved through integration and product architecture, and became CTO in 2020. The work changed from connecting a file to a system to deciding which assumptions a product may safely carry across many institutions.
The decisions on this site concern ambiguous matching, a large client's effect on shared processing, an export that initially omitted usable evidence and a proposed customer-specific branch of the product. They show where engineering had to challenge a commercially attractive answer. Some choices reduced a headline performance measure or required expenditure that sales had not planned. I want the explanation to retain those consequences, because they are part of the job rather than footnotes to it.
The attractive match rate was insufficient
A proposed matching release was tested against 8,200 previously adjudicated record pairs. Fifty-four suggested pairings were plausible but wrong. Excluding ambiguous same-value groups reduced the simulated automatic-closure rate from 91% to 76%. We accepted the lower rate and retained human review rather than present a faster result as a better financial conclusion.
One busy client changed nine other mornings
A large import consumed 62% of a shared worker pool during the observed window. Nine other clients' median completion time rose from 21 to 56 minutes. We introduced bounded allocations and measured each client's completion experience, accepting that the largest job would finish later than under unrestricted use.
An export needed its supporting evidence
A customer exit rehearsal covered 36,000 exception cases. The data export contained 2,400 attachment references whose files were not included in a usable package. We funded ₹18 lakh of additional engineering work to produce an authorised export with the relevant evidence and access mapping; a row count alone had not established portability.
Current focus
I am strengthening the way product proposals describe their assumptions before a customer contract makes those assumptions difficult to change. The work includes matching evidence, capacity allocation and what an institution can independently understand from its own exported records.
Conversations I welcome
I would consider a first non-executive role in an enterprise-software or specialised service company with a substantive product-governance agenda. I can provisionally allow twenty-two days a year, subject to employer consent and conflict checks. I would help a board examine the relationship between engineering choices, commercial promises and continuing customer responsibility.
Beyond the page
The record behind the perspective.
Formation
Bachelor of Engineering in ElectronicsSavrenil Institute of Engineering · 1995
Since becoming CEO in 2019, I have relied on Farhad to explain the commitment behind a proposed delivery date. He will bring an alternative and its cost, including the possibility that the customer may decline it, instead of hiding the difficult part inside an engineering estimate.
Meher DalsaviCEO, Orevathi Settlement Systems; executive colleague since 2019The full record ↗
Contact
I would consider a first non-executive role in an enterprise-software or specialised service company with a substantive product-governance agenda. I can provisionally allow twenty-two days a year, subject to employer consent and conflict checks. I would help a board examine the relationship between engineering choices, commercial promises and continuing customer responsibility.
