A 45% improvement in campaign conversion rate is not a creative or channel optimisation — it is a matching improvement. The same campaign, the same budget, the same channel mix, the same creative: the only change is that each borrower receives the product offer they are most likely to want at the moment they are most likely to want it, computed from 30 signals, rather than a generic offer sent to everyone on a single send date. The Personalisation Agent AI does not make campaigns more expensive — it makes them more accurate. And accuracy, in a market where the cost per funded loan determines whether the marketing function is profitable, is the only metric that matters.
What the 45% improvement is measuring — and why it is conservative
The 45% conversion rate improvement is measured from generic campaign to fully personalised campaign, using the same audience, the same budget, and the same primary channel (WhatsApp and email), across a 90-day period. The conversion event is a funded loan — not an open, a click, or an application. The generic campaign converted 2.8% of the contacted audience to funded loans. The personalised campaign converted 4.1%. The absolute difference is 1.3 percentage points. Applied to 6,841 eligible borrowers, it represents 89 additional funded loans from the same ₹0 incremental budget (personalisation replaces content that was already being sent, not additional spend).
The improvement is conservative in the sense that it does not account for the quality improvement in funded loans — the personalised campaign's funded loans show a 4.2% early DPD rate against the generic campaign's 8.7%, suggesting that personalised offers select for better-quality conversions as well as more conversions. When the credit cost improvement is included, the net economic benefit of personalisation is 2 to 2.5 times the gross conversion rate improvement.
The 5 mechanisms that produce the 45% lift
Each borrower receives only the product offer they are eligible for and suited to — no mismatched offers
In the generic campaign, 28% of the audience received an offer for a product they were ineligible for at the time of sending — either FOIR was too high to add an EMI, their CIBIL was below the product threshold, or their loan vintage was insufficient. These borrowers cannot convert regardless of how good the offer is. Removing them from the audience and replacing them with eligible borrowers raises the conversion denominator quality. Among eligible borrowers, offering the product that best fits their signal profile (rather than defaulting to the institution's highest-margin product) raises the acceptance rate within the eligible audience.
→ Eligibility gate removes 28% of generic audience · Remaining 72% are genuine candidates · Offer matched to best-fit product within eligible setEach offer is sent at the moment when the borrower's individual signals indicate peak receptivity
The generic campaign sends all messages on a single date — November 1. The personalised campaign sends each message on the borrower's individual optimal date, determined by their payment cycle position, income event proximity, engagement window, and suppress windows. The result: 841 of 6,841 messages are sent in the first 7 days because those borrowers have active engagement signals. The remaining 6,000 are distributed across November 3–30, each timed to the borrower's specific financial calendar. The improvement in acceptance rate from timing alone — sending at the borrower's moment rather than the institution's schedule — accounts for approximately 12 percentage points of the 45% total lift.
→ No single campaign date · Each offer sent at borrower's individual optimal moment · Suppress windows prevent contact during distress periodsThe offer message references the specific signals that make the offer relevant — creating recognition rather than solicitation
The generic offer says "You are pre-approved for an MSME working capital loan." The personalised offer says "Your Tamil Nadu expansion is underway — we've pre-approved ₹17.4 lakh in working capital at 13.5%, disbursed in 5 days." The borrower who receives the second message recognises their situation in it — and recognition is the single strongest predictor of response. The 8 percentage point contribution from message personalisation comes from the difference between messages that feel like solicitations and messages that feel like useful financial advice.
→ Message references 2–3 specific borrower signals · Offer amount is the FOIR-constrained exact figure · Language matched to segment (MSME vs salaried vs SE)Each borrower receives the offer on the channel where they have highest historical response rate
WhatsApp has a higher open rate for MSME borrowers; email has a higher click-to-conversion rate for salaried home loan borrowers; some borrowers respond to both equally. The Personalisation Agent AI checks each borrower's 6-month communication response history and selects the primary channel accordingly. For borrowers who respond to both, it sends a WhatsApp first and follows with email 48 hours later if no response — a sequenced two-channel approach rather than a simultaneous blast that trains borrowers to ignore redundant messages.
→ Primary channel: highest 6-month response rate · Secondary channel: 48-hour follow-up if no primary response · No simultaneous multi-channel blastOffers that are genuinely pre-approved (FOIR and CIBIL already verified) convert better than offers that say "you may be eligible"
A borrower who receives an offer with specific locked terms (₹17.4 lakh, 13.5%, 36 months, valid until November 30) can make a binary decision: yes or no. A borrower who receives "you may qualify for up to ₹20 lakh" must apply to find out whether and what they actually receive — an uncertain process that many borrowers defer indefinitely. The Personalisation Agent AI generates only genuine pre-approvals — where the eligibility check, FOIR calculation, and product-fit scoring have already been completed. The specific terms in the message are the terms the borrower will receive, not a starting point for negotiation.
→ All offers are genuine pre-approvals · FOIR verified · CIBIL checked · Specific terms locked · Valid date stated · No "subject to credit assessment" languageThe conversion comparison: generic vs personalised (same 6,841 borrowers)
The 45% conversion improvement is not a projection — it is the measured outcome of sending the right offer to the right person at the right moment, at scale
The generic campaign's 2.81% conversion rate is not low because the institution's products are unattractive or the creative is poor. It is low because 28% of the audience cannot convert (wrong eligibility moment), the remaining 72% received the offer on a day that was right for the institution but may not have been right for them, the message did not reflect their specific situation, and the offer terms were a range rather than a commitment. Each of these four failures costs conversion rate. The Personalisation Agent AI corrects all four simultaneously — eligibility gate, timing, message, and genuine pre-approval — and the result is a conversion rate that is 46% higher, a cost per funded loan that is 45% lower, an early DPD rate that is half the generic rate, and an unsubscribe rate that is 15 times lower. Personalisation at scale is not a technology advantage — it is what the borrower relationship looks like when the institution actually uses the data it already has.