The LMS selection mistakes that have nothing to do with features
The feature grid is rarely why an LMS selection goes wrong. It is who ran the evaluation, what the reference call would have shown, and what the signed contract actually says.

Contents 9 sections
For a lender preparing an LMS shortlist for its board, a feature grid is an incomplete decision record. It needs to show who tested the operating model, what the references established and how the lender would continue if the vendor failed. Those questions belong in the selection of a loan management system, before the commercial decision becomes difficult to unwind.
An LMS selection can fail even when the features fit: an all-IT buying committee with no collections or compliance voice, a contract sized for today’s book instead of the next two years, a reference call skipped in favor of a polished demo, an RFP response trusted more than the signed contract, and a roadmap slide bought as if it were a shipped feature.
None of this shows up on a feature comparison grid, which is exactly why it is so easy to miss. A grid rewards the vendor who answers “yes” the most times. It has no row for who was in the room, what a reference lender actually said, or what the contract’s exit clause reads like at 2am eighteen months from now.
Mistake one: the buying committee is all IT
IT can assess architecture, security posture and integration effort better than anyone else in the building. It cannot tell you whether the platform’s part-payment logic matches how your collections team actually negotiates a settlement, or whether your compliance team can pull the export a regulator will ask for on day one. Those are different questions, answered by different people, and a committee that skips them finds out the hard way after go-live.
The fix is not complicated: put collections, servicing and compliance in the room before the shortlist is final, not after the contract is signed. Each of them tests a different failure mode a purely technical evaluation will not surface, because what predicts fit is how a platform governs a change to a loan before it posts, not its architecture diagram.
Mistake two: sizing for today’s book, not the next two years
A contract priced against today’s loan count, product count and transaction volume may behave differently as the book grows. Model the effect of additional users, products and transactions before selecting a price. Establish which changes are covered by configuration and which need separately priced work.
The question to ask before signing is not “what does this cost today” but “what does this cost at twice the book, and does adding a fourth product require a new commercial conversation or a configuration change.” A vendor who cannot answer the second half of that question in specifics is telling you the contract was priced for the demo, not for growth.
Mistake three: skipping the reference call
A demo shows exactly what the vendor chose to show, in an environment the vendor controls completely. A reference call, arranged directly with a lender running a similar product mix and a similar book size, is the one part of the evaluation the vendor does not get to stage.
The questions worth asking on that call are specific: what actually broke in the first ninety days after go-live, how long support actually took to resolve it against what the contract promised, and whether the implementation timeline the vendor quoted matched the timeline the reference lender actually lived through. If a relevant reference is unavailable, record what remains unverified. Confidentiality or scope can limit the references a vendor can offer; the lender still needs a considered response to that evidence gap.
Mistake four: trusting the RFP response instead of the contract
An RFP answer records a vendor’s proposal. The executed agreement and its incorporated schedules should preserve the commitments the lender relied on. Have legal counsel check which documents form the agreement, how conflicts between them are resolved and what remedies apply. Do not leave an important service level or exit obligation only in a presentation.
Every specific promise made during evaluation, a support response time, an audit right, an exit timeline, a data-export format, needs to appear in the contract with a number and a remedy attached, not just in the deck the sales team presented. What RBI’s outsourcing rules actually require you to prove about audit rights and exit strategy is a useful checklist for exactly which promises need to survive the trip from RFP response to signature.
Mistake five: buying the roadmap slide
A roadmap item is a dependency to assess, not evidence that a capability is available today. If the operating plan needs it at launch, record the gap and the consequence of proceeding without it. Score the demonstrated current scope separately from proposed development.
The honest evaluation question is narrow: is this shipped, in production, on a live tenant today, or is it a roadmap item with no committed date? A vendor being honest about that distinction, rather than blurring it, is itself a signal worth weighing. Ours is one of them: Lokta lists Loan Origination as a roadmap capability, precisely so a buyer scoring today’s platform is not scoring tomorrow’s slide.
The board pack: three questions before you choose
A new lender may want its technology ready before its licence arrives. Use that time to define the book, compare options and test the operating handover. Technology preparation does not establish permission to commence lending. The board still needs a reasoned choice it can revisit when an assumption changes.
Start with the new-lender stack and cost worksheet for the case for a system. Then compare three shortlisted options against the same post-approval scope. Give each option an evidence reference, an accountable reviewer and any condition that must be met before signing or launch. An unanswered row stays open.
What does it run on?
Ask for a dated architecture and dependency map: the record that holds the loan, the services it depends on, the parties responsible for operation and the handover from your existing systems. Follow an approved loan into servicing, then a repayment, a correction and the accounting result. Record what was demonstrated and what still needs work.
For Lokta, the scope starts after approval. Your LOS, or your partner lender’s, hands the loan over through an API. Data is hosted in-region. Carry the lender’s required region and responsibilities into the evaluation and contracting discussion; the short hosting statement does not settle the whole architecture review.
Who else runs it?
Ask for a reference relevant to your products and operating scope, with permission to speak to the lender. Record what the reference actually used and when, rather than treating a logo as proof of every capability. Where a suitable reference is unavailable, record the evidence gap and decide what testing and contractual conditions would be needed to accept it.
Lokta is built by the team behind Apache Fineract. That heritage explains the team’s lending-infrastructure experience. It does not establish a named Lokta deployment or transfer an ecosystem result to this product.
Aggregate impact attributed to the Mifos/Fineract ecosystem, not Lokta directly.
What happens if the vendor goes away?
An export clause needs an operating counterpart. Ask for a sample export with its schema, the configuration and account history needed to interpret it, and a reconciliation that another team can reproduce. Name the person who can obtain it, the alternative arrangement and the support needed during transition. Assess service continuity and vendor failure alongside a planned commercial exit.
The RBI’s NBFC Managing Risks in Outsourcing Directions, 2025, issued on 28 November 2025, address continuity and exit in paragraphs 83-88. Their IT chapter applies to covered NBFCs in the Middle layer and above, subject to the material-outsourcing scope and exclusions in paragraphs 3 and 5. Have the compliance owner establish applicability to the actual arrangement. The deployment guide explores audit, continuity and exit evidence in more detail.
A recovery plan restores a service after disruption. An exit plan gets the lender to an alternative operating arrangement. Ask for evidence for both, and carry accepted obligations into the agreement. The lender remains accountable for its book.
| Board question | Evidence beside each option | Who reviews it |
|---|---|---|
| What does it run on? | Architecture, dependencies and a tested servicing handover | Technology and operations |
| Who else runs it? | Relevant reference notes, or a recorded gap and acceptance condition | Business sponsor and risk |
| What if the vendor fails? | Usable export, continuity evidence and a contractual transition plan | Operations, risk and legal |
Download the board-pack checklist (PDF). It has space for three options, evidence checks and the conditions behind a recommendation. Use the LMS RFP toolkit for the detailed requirements that sit behind it.
Run the evaluation you would defend to your own board
There are three honest ways to run this evaluation. Run it fast with the committee you already have, and accept the risk that a voice missing from the room becomes a problem discovered after go-live. Build the full cross-functional committee and add the reference calls and contract-language review, which costs weeks the procurement timeline may not have budgeted for. Or run a structured discovery phase with your top two vendors before the final decision, trading a longer evaluation for a contract that was actually tested against your book rather than a template.
A fuller evaluation takes time and still leaves judgement calls. Its value is a record of the evidence, the gaps and the risk the lender chose to accept. Bring one unresolved board question to a conversation with Lokta, together with the post-approval workflow it affects.
Frequently asked questions
What are the most common mistakes NBFCs make when selecting a loan management system?
Important selection mistakes can sit outside the feature list: letting IT run the evaluation without collections, servicing and compliance in the room, sizing the contract for today's book instead of the next two years, skipping a candid reference call with a lender running a similar book, trusting the RFP response deck instead of what the signed contract actually commits to, and buying a roadmap slide as if it were a shipped capability.
Should IT alone run an LMS vendor evaluation?
No. IT can assess architecture, security and integration fit, but cannot answer whether a platform's appropriation logic matches how collections actually works, or whether compliance can produce the exports a regulator will ask for. An evaluation committee without collections, servicing and compliance in the room risks accepting a platform whose demonstration does not reflect the operating work after go-live.
Why does a reference call matter more than a vendor demo?
A demo shows what the vendor chose to show. A reference call with a lender running a similar product mix and book size answers the question the vendor cannot: what actually broke after go-live, how support responded, and what the vendor's own timeline estimate turned out to be worth. Vendors control a demo completely. They do not control what a reference lender says on a call the lender arranged directly.
What is the difference between an RFP response and the signed contract?
An RFP response describes the vendor's proposal, while the executed agreement and incorporated schedules set out the agreed obligations. Ask legal counsel to check which evaluation documents are incorporated, which terms take precedence and what remedies apply. Carry material promises about support, audit access, export formats and exit assistance into that review. Do not assume that a statement in a demo has the same effect as a negotiated commitment, or that one document label alone decides enforceability.
What should an LMS selection board pack contain?
Compare the same operating scope across three options. For each, record what the platform runs on, the relevant reference evidence and how the lender would continue or exit if the vendor failed. Attach evidence references, owners and acceptance conditions to unresolved items. Include the required post-approval workflows and the handover from origination. A board pack summarises the decision and remaining risk; the detailed RFP, commercial model and legal review remain supporting work.
Does Apache Fineract heritage prove a Lokta customer deployment?
No. Lokta is built by the team behind Apache Fineract. That is a statement about the team's experience, not evidence that a particular lender uses Lokta or that an ecosystem outcome belongs to Lokta. Assess customer references separately, with permission and enough context to judge similarity of product mix and operating scope. Also test the current software, support arrangements and contract for the book you intend to operate. Heritage is one input to the evaluation.
Read next
- How long an LMS implementation actually takes: the timeline questions to ask before you sign, not after.
- On-prem, VPC or shared cloud: the audit, exit and disaster-recovery test that belongs in the contract itself.
- Loan management system cost: pricing models and TCO: the five-year cost sheet that catches mistake two before you sign.
Lokta editorial analysis by Chandramouli, co-founder and CEO.


