The Number That Started It
There is a number that I return to whenever someone asks me about CarInfo. It is not 45 million — the monthly active users we eventually reached. It is not 680 times — the insurance revenue multiplier we achieved over 18 months. It is not even 1,200% — the headline revenue growth figure that ends up in presentations.
The number I return to is 0.007%.
That was the conversion rate on insurance products when I joined CarInfo as Senior Product Manager in 2021. Seven out of every hundred thousand users who opened the app were purchasing an insurance product. The company was generating millions of sessions per month from a product that had genuine and obvious insurance monetization potential, and converting less than one in ten thousand users on that potential.
When I first saw this number in our analytics dashboard, I spent a long time looking at it. Not because I couldn't believe it, but because I was trying to understand exactly how a product with such clear intent signals — people were actively searching for their vehicle insurance information — could be performing so catastrophically on conversion.
The answer to that question, and the process of addressing it, is what this story is about.
"0.007%. Seven out of every hundred thousand users. The potential was obvious. The gap was enormous. The question was: what exactly was breaking?"
India's Vehicle Information Desert
To understand CarInfo, you need to understand the problem it was solving. India had, as of 2021, approximately 295 million registered vehicles. Cars, two-wheelers, commercial vehicles, three-wheelers — the full spectrum. Every single one of those vehicles had data attached to it: registration information, owner history, insurance status, fitness certification, challan (traffic violation) records, emission testing dates.
Accessing that data, for a vehicle owner, was historically a nightmare. It was scattered across state transport authority databases, insurance company systems, traffic police records, and RTO (Regional Transport Office) registries. None of these systems talked to each other. None of them were designed with the end user in mind. The RTOs themselves were notoriously offline, paper-heavy bureaucracies where getting a simple vehicle ownership transfer could take months.
The consequence of this information fragmentation was significant. A used car buyer couldn't easily verify if the vehicle they were considering purchasing had a clean title and no hidden loans against it. A vehicle owner couldn't easily check if their insurance was still active or when it expired. A traffic police officer couldn't instantly verify insurance validity for a vehicle stopped at a checkpoint. This created a multi-sided problem: friction, opacity, compliance risk, and fraud.
The insurance situation was particularly striking. Approximately 62% of registered vehicles in India were either uninsured or operating with lapsed insurance policies. This wasn't primarily because people couldn't afford insurance — for a two-wheeler, basic third-party insurance cost as little as ₹700–900 per year. It was because the renewal process was friction-heavy, the awareness of lapsed status was low, and the consequence (traffic fines, legal liability in an accident) felt abstract until it wasn't.
CarInfo had identified this market correctly. The question was whether they could convert that market insight into product outcomes.
How CarInfo Started
CarInfo began as a simple utility: enter a vehicle's registration number, get back the vehicle details pulled from government databases. RC (Registration Certificate) check was the core use case. Is this car what the seller says it is? Who is the registered owner? Is it a stolen vehicle? Is there a loan against it?
This utility was genuinely valuable — valuable enough that the product grew to 3.8 million monthly active users organically, largely through word of mouth from the used car market. Anyone who was buying or selling a vehicle needed CarInfo, or something like it.
The founding insight was sound. The execution had created real traction. But the product had two fundamental problems when I arrived: it had no clear growth engine beyond organic word-of-mouth, and it had almost no monetization despite enormous intent signals in its user base.
What I Inherited: The Full Audit
My first thirty days at CarInfo were entirely diagnostic. I built a complete picture of where the product stood before proposing any solutions. Here is what that picture showed:
Traffic and Acquisition
The product had meaningful organic traffic driven by one thing: insurance expiry and renewal searches. When people discovered their vehicle insurance had lapsed (often after a traffic stop), or when insurance companies sent renewal reminders, they searched for their vehicle's insurance status online. CarInfo ranked for many of these queries. But the SEO was shallow — a handful of generic pages, not a structured content architecture.
User Behavior Inside the App
Using session recording and funnel analysis, I mapped what users actually did inside the app after completing an RC check:
| Action After RC Check | % of Users | Insight |
|---|---|---|
| Left immediately | 51.3% | Used it as a lookup tool, got what they needed |
| Explored vehicle details further | 28.7% | Higher engagement, but no clear next step |
| Clicked on insurance section | 14.2% | Clear intent signal — they noticed their insurance status |
| Interacted with insurance CTA | 3.8% | Showed interest in buying/renewing |
| Started insurance form | 1.1% | Form start was deeply in the funnel |
| Completed insurance purchase | 0.007% | The disaster number |
The funnel analysis was illuminating. Of the 14.2% of users who clicked on the insurance section after seeing their vehicle's insurance status, only 0.05% eventually purchased. That's the conversion rate within an already-interested cohort — and it was catastrophically low. Something was happening between "interested user sees insurance section" and "user buys insurance" that was killing almost every potential transaction.
The 0.007% Revelation: Mapping the Seven Leak Points
I spent two weeks on nothing but the insurance funnel. Session recordings, heatmaps, drop-off analysis at every step, and qualitative interviews with users who had started but not completed insurance purchases. What I found was not one problem. It was seven distinct problems, layered on top of each other, each doing its share of damage.
Leak 1: Wrong Placement, Wrong Timing
The insurance CTA appeared on the RC check results page — the first page a user saw after entering their registration number. This sounds right. It isn't. At the moment a user sees their RC check results, they are in information-gathering mode. They've just learned whether the vehicle has a clear title, who owns it, what its specifications are. Their mind is processing that information. They are not in purchase mode.
Inserting an insurance purchase CTA at this exact moment is asking for a context switch — from "understanding my vehicle" to "making a financial decision about insurance." The CTA appeared at the wrong moment in the cognitive journey.
Leak 2: The Trust Deficit
CarInfo was known as an RC check tool. Insurance was a financial product with meaningful risk implications. Users who encountered the insurance offering were, in many cases, encountering a financial services product from a provider they associated with a completely different category. The trust transfer wasn't happening automatically.
In user interviews, this came through clearly: "I know CarInfo for checking vehicle details. I wouldn't buy insurance from them without knowing more about who's underwriting it." This is a rational response. Insurance is a category where trust matters enormously because the product only delivers value in the event of a claim — a moment that, by definition, involves some of the most stressful circumstances in a person's life.
Leak 3: The Form from Hell
The insurance application form was 23 fields across 6 steps. Twenty-three. For a product renewal that insurance companies could, in many cases, pre-fill using vehicle registration data they already had. I timed the form completion on a standard Android phone: 11 minutes if you moved quickly and had all information at hand. More typically 18–22 minutes for someone not familiar with the process.
For comparison: the leading direct insurance apps (Digit, Acko) had moved to 3–4 field renewal flows with UPI payment integration. The CarInfo form was a relic of an earlier era of web-based insurance forms translated directly to mobile without mobile-first rethinking.
Leak 4: No Payment Method Alignment
The insurance payment options at the time were credit card and net banking. UPI was not supported. This is hard to overstate in the context of 2021 India: UPI was the payment method for the demographics most likely to buy vehicle insurance through a mobile app. People who were buying two-wheeler insurance for ₹700 were not likely to have a credit card. Net banking was cumbersome on mobile. This single gap was eliminating a significant portion of the purchase-intent population.
Leak 5: The Price Confusion
Insurance quotes appeared in the app without context. A quote of ₹4,200 for comprehensive car insurance might be a good price, a bad price, or an average price — the user had no way to know. There was no comparison, no context, no explanation of what the coverage included. Users who reached the quote stage frequently abandoned at this point to "check prices elsewhere." Many of them never came back.
Leak 6: No Urgency Signal
If a user's vehicle insurance had expired six months ago, the app treated this the same way it treated an insurance expiry next week. No urgency messaging. No "you're driving without insurance, which is illegal and risky" framing. No incentive to act now rather than later. The product was presenting a time-sensitive decision as if it were a timeless one.
Leak 7: No Follow-Through
The biggest leak of all: users who saw their insurance status, seemed interested, but didn't purchase on that session received no follow-up. No notification, no email, no reminder. The session ended and the opportunity evaporated. Given that insurance purchase decisions often require some deliberation — comparing options, checking bank balance, discussing with a family member — the absence of any follow-through mechanism meant that a huge proportion of warm leads were going cold through pure neglect.
Each individual leak reduced conversion by 60–80%. Compound seven of these and you get 0.007%. The lesson: catastrophic conversion rates rarely have a single cause. They are usually the product of multiple compounding failures that are individually addressable but collectively devastating.
The Insurance Opportunity — Properly Sized
Before moving to solutions, I needed to size the opportunity accurately. The team had a sense that insurance monetization was important, but nobody had put a rigorous number on what it could actually be worth.
Here is how I approached the sizing:
Addressable user base: Of the 3.8M MAU, approximately 68% were returning users — people who had checked their vehicle on CarInfo at least once before. Of returning users, approximately 74% had vehicles whose insurance status was expired or within 90 days of expiry when they checked. That's roughly 1.9 million monthly users with a demonstrable and time-sensitive insurance need.
Conversion assumption: If we could fix the funnel to industry-standard levels (comparable direct insurers were converting at 2–4% on mobile), even a 2% conversion of the 1.9M addressable users was 38,000 policies per month.
Revenue per policy: Average commission on a two-wheeler policy was approximately ₹180. On a car policy, approximately ₹650. Blended average, given our vehicle mix, approximately ₹290.
Monthly revenue potential: 38,000 × ₹290 = ₹1.1 crore per month, or approximately ₹13 crore annually. From roughly ₹0 at the time of the audit.
This number focused the entire team. Insurance wasn't a side project — it was potentially the largest revenue line in the business. Fixing the funnel wasn't optimization work. It was company-defining work.
The SEO Architecture Decision
The insurance funnel fix was the immediate priority. But parallel to it, I was thinking about a different problem: CarInfo's growth engine was weak. At 3.8M MAU, the app had good traction but no mechanism that would drive it to 10M, 20M, or 40M users without proportional increases in marketing spend.
The insight that changed my thinking: CarInfo's core value proposition — vehicle information — was inherently searchable. When someone wanted to know their vehicle's insurance status, they searched for it. When someone was researching a used car, they searched for the model's reliability, problems, recall history. When someone wanted to check their RC status, they searched for that too. Every use case of CarInfo's product was preceded, in a significant proportion of cases, by an organic search query.
This meant that SEO wasn't a marketing channel for CarInfo — it was a growth engine. But the SEO architecture needed to be built to serve it.
The SEO Audit
When I audited CarInfo's SEO position, I found:
| SEO Dimension | Current State | Opportunity |
|---|---|---|
| Total indexed pages | ~340 | Potentially 5,000–8,000 |
| Ranking keywords | ~1,200 | Estimated 40,000+ addressable |
| Vehicle model pages | None | 300+ makes × models |
| State/city RTO pages | None | 1,500+ RTO offices |
| Insurance how-to content | None | High-volume informational queries |
| Page load speed (mobile) | 4.8s avg | Industry best: under 2s |
The opportunity was enormous. CarInfo had 340 indexed pages competing for a keyword universe of tens of thousands of highly relevant queries. The solution was programmatic SEO: building a content architecture that could generate thousands of relevant, high-quality pages from structured data.
The Programmatic SEO Architecture
We designed a content architecture with four main content types:
Vehicle Model Pages: A page for every major vehicle sold in India, covering specifications, common problems, insurance costs, RTO registration fees, known recalls, owner reviews, resale value trends. For popular models, this content was manually researched and written. For the long tail, we built templates that pulled from structured data sources.
RTO Location Pages: A page for every Regional Transport Office in India, covering services offered, contact information, address, hours, process guides for common transactions (ownership transfer, hypothecation removal, fitness certificate renewal). There are over 1,500 RTOs in India. We built 1,400 pages.
Insurance Informational Pages: How-to content covering every insurance-related query we could identify: "how to check vehicle insurance status," "what to do if insurance lapses," "difference between comprehensive and third-party insurance," "how to file a vehicle insurance claim," "how to transfer vehicle insurance after resale." These were genuine informational queries from real people facing real situations.
RC Check Guides: State-specific and use-case-specific guides for using the RC check function, targeting the long tail of RC-check-related queries.
Building this architecture took five months of coordinated work — engineering, content, and SEO. The results took longer to materialize (SEO always does) but when they did, they were transformative.
The Trust Problem in Financial Products
Let me come back to the insurance funnel, specifically to Leak 2: the trust deficit. Solving the trust problem was the most interesting and nuanced part of the funnel optimization work, and it's the piece that most product teams underinvest in.
Trust in a financial product is not conveyed by saying "we are trustworthy." It is conveyed by signals — specific, legible, contextually relevant signals that the user's pattern-recognition systems process, often unconsciously, when evaluating whether to proceed with a transaction.
What Insurance Trust Signals Look Like
Through user research — a combination of qualitative interviews and A/B-testable trust signal experiments — we identified the signals that actually moved conversion in the CarInfo context:
| Trust Signal | Type | Conversion Impact |
|---|---|---|
| Insurer brand visibility (Bajaj, Tata, HDFC) | Institutional credibility | +34% on quote acceptance |
| IRDAI regulatory disclosure | Regulatory legitimacy | +18% on form start |
| "1.2 lakh policies sold" claim | Social proof | +22% on quote click |
| Claim settlement ratio visibility | Product quality signal | +29% on insurer selection |
| Instant policy issuance guarantee | Process reassurance | +41% on checkout start |
| Data security / SSL visible badge | Safety signal | +11% on form completion |
The biggest impact came from a combination of insurer brand visibility (showing the logos of recognized insurance brands prominently, not just CarInfo's brand) and the claim settlement ratio. The latter deserves special attention.
Claim settlement ratio — the percentage of claims that an insurer actually pays out — is a public IRDAI-mandated disclosure. It ranges from about 95% to 99.5% across major insurers. The difference between a 95% and 99% claim settlement ratio might seem small. But in user research, when we explained what it meant — "if you make 100 claims, this insurer pays 95 of them vs 99 of them" — the higher ratio had significant influence on insurer selection and willingness to complete purchase.
We built claim settlement ratio comparison directly into the product, so users could see it next to the price quote. The market had been treating insurance as a pure-price commodity at the point of purchase. We introduced quality signals alongside price, and conversion improved significantly.
The Borrowed Trust Architecture
One structural decision that solved the trust problem was building what I called a "borrowed trust architecture." Instead of trying to convince users that CarInfo was trustworthy in insurance (a heavy lift, given that we were new to the category), we leveraged the existing trust that users had in the underlying insurers.
CarInfo became a transparent marketplace. When you bought through CarInfo, you were explicitly buying a Bajaj Allianz policy, or a Digit policy, or an Acko policy. CarInfo's role was explicitly framed as "finding you the best price from trusted insurers" rather than "providing insurance." This framing reduced the trust barrier significantly — users already trusted Bajaj Allianz. They just needed to trust that CarInfo was showing them legitimate offers from Bajaj Allianz.
You don't always need to build trust from scratch. Sometimes you can borrow it from brands your users already trust. The design challenge is making that borrowing explicit and credible.
Rebuilding the Insurance Funnel
With the leak analysis complete and the trust architecture designed, we rebuilt the insurance funnel from scratch. Not patched — rebuilt. The old funnel was abandoned entirely. Here is what the new funnel looked like and why each decision was made:
Step 1: Redesigned Entry Point
We moved the insurance offer from the RC check results page to a dedicated "Insurance Status" tab that users navigated to intentionally. Instead of ambushing users with an insurance CTA while they were processing vehicle information, we created a specific context — "Your Insurance" — where the intent was clearly about insurance and the cognitive mode was appropriate for a purchase decision.
The Insurance Status tab showed:
- Current insurance status (active/expired/expiring soon) with a color-coded indicator
- Days remaining until expiry (or days since expiry, with urgency framing)
- Current insurer name and policy number (pulled from government databases where available)
- Estimated renewal cost range (using historical pricing data by vehicle type)
- A prominent but contextually appropriate "Renew Now" CTA
Step 2: Urgency Calibration
The urgency message was calibrated to the actual expiry situation:
| Days to/from Expiry | Messaging Tone | CTA Copy |
|---|---|---|
| 61+ days before expiry | Informational | "Your insurance is active. Renew early for best rates." |
| 31–60 days before expiry | Proactive | "Renew in the next 30 days to avoid any lapse." |
| 8–30 days before expiry | Urgent | "Your insurance expires in X days. Renew to stay protected." |
| 1–7 days before expiry | High urgency | "⚠ Insurance expires in X days. Driving without insurance is illegal." |
| 1–30 days after expiry | Compliance + risk | "Your insurance expired X days ago. You're driving illegally." |
| 31+ days after expiry | Recovery | "Renew insurance to protect yourself from ₹2,000+ fines and legal liability." |
This sounds like basic personalization. It is. But it was completely absent in the original product, which used the same generic CTA regardless of insurance status. The difference in conversion across these urgency bands was significant — expired insurance users (the highest urgency state) converted at 4.8× the rate of users with 60+ days remaining.
Step 3: Form Compression
The 23-field, 6-step form was reduced to 4 fields and 1 step for the majority of renewals:
- Vehicle registration number (already known from the RC check)
- Current insurance status (populated from our data)
- Mobile number (for policy delivery)
- Coverage type preference (comprehensive vs third-party)
Everything else was either pre-populated from the vehicle registration data or collected progressively after the user had committed to a quote. The principle: only ask for what you need to generate a quote. Collect additional information after purchase intent is established.
Step 4: UPI Integration
UPI payment was integrated as the primary payment option. Implementation took 6 weeks due to integration with our insurance partner APIs and compliance requirements. In the first month after launch, 73% of insurance purchases were completed via UPI. The removal of the payment method gap alone was estimated to account for approximately 18% of the total conversion improvement.
Step 5: The Notification Cadence
We built an insurance renewal notification system that contacted users at four touchpoints before and after insurance expiry. This required building out push notification infrastructure (previously under-used) and a permission request flow that contextualized why the notifications were valuable.
The notification cadence:
- T-45 days: "Your insurance renews in 45 days. Lock in your rate now."
- T-15 days: "15 days to renewal. Compare quotes in 2 minutes."
- T-3 days: "Insurance expires in 3 days. Renew now to avoid driving uninsured."
- T+1 day (if not renewed): "Your insurance expired yesterday. Renew immediately."
Open rates on these notifications were significantly higher than the app's baseline push notification open rate, because the content was personally relevant and time-sensitive. The T+1 notification had the highest urgency and the highest conversion rate.
CAC: The Attribution Discovery
While the insurance funnel was being rebuilt, a parallel problem was surfacing in the growth data: CarInfo was spending on user acquisition without any real understanding of which acquisition channels were generating users who actually monetized.
This sounds like a basic attribution problem. It was, technically. But the reason it hadn't been solved was more interesting than the technical gap: the team was measuring acquisition success by cost per install (CPI) — a metric that tells you nothing about the downstream quality of those installs.
Building Attribution Infrastructure
The first step was connecting acquisition source to downstream behavior. This required implementing proper UTM tracking and attribution logic that could follow a user from the ad they clicked through install, first session, first RC check, and ultimately first insurance purchase.
When we had this infrastructure in place and ran the analysis on four months of data, the results were startling. The cost per install by channel varied enormously, as expected. But the cost per insurance purchase — what we actually cared about — varied far more:
| Channel | CPI | Insurance CPP | CPP / CPI Multiple | ARPU (3-month) |
|---|---|---|---|---|
| Google Search (insurance intent) | ₹28 | ₹420 | 15× | ₹280 |
| Google Search (RC check) | ₹34 | ₹680 | 20× | ₹195 |
| Facebook — interest targeting | ₹18 | ₹2,240 | 124× | ₹55 |
| Facebook — lookalike | ₹22 | ₹890 | 40× | ₹140 |
| YouTube pre-roll | ₹41 | ₹3,100 | 76× | ₹48 |
| Organic (SEO) | ₹0 (no paid cost) | ₹190 | — | ₹310 |
The Facebook interest targeting channel was spending significantly. On a CPI basis, it looked reasonable. On a cost-per-insurance-purchase basis, it was 5× more expensive than Google Search intent keywords and 12× more expensive than organic traffic. The users it was delivering were people who were in the broad "vehicle owner" interest category but had no immediate intent or need — they were window shoppers who would check the app once and never return.
The organic traffic, by contrast, was converting at the lowest cost precisely because it was capturing users who had searched for their vehicle information because they had an immediate need. Intent-driven traffic converted. Interest-driven traffic didn't.
CPI is a measure of acquisition cost. What you actually need to measure is the cost of acquiring someone who will give you the outcome you want. The gap between these two numbers is where most acquisition budgets get wasted.
Reallocation
The reallocation was straightforward once the data was clear: shift budget from interest-targeted social to intent-targeted search, and invest in SEO to expand organic. The organic investment was validated by the lowest CPP in the table — every incremental SEO user who converted was effectively a ₹190 insurance sale with zero acquisition cost allocated to it.
We reduced Facebook interest targeting from 40% of the paid budget to 12% and moved the freed budget to Google Search. Within three months, blended CPP had dropped from approximately ₹1,100 to approximately ₹590 — a 46% improvement — without reducing total insurance conversion volume, because the higher-quality traffic converted at higher rates.
Engineering Growth Loops
By month six, the insurance funnel was working, the CAC was improving, and the SEO architecture was in place. The growth rate was accelerating. But I was increasingly focused on a question that felt more important than any of the tactical optimizations: what was the growth loop?
A growth loop is a mechanism by which growth generates more growth. Unlike a paid acquisition funnel (which requires continued investment to sustain), a growth loop compounds. The businesses that reach extraordinary scale — hundreds of millions of users — almost always have at least one strong growth loop at their core.
CarInfo's potential growth loops were hidden in the product's usage patterns. Here's how I identified and built them:
Loop 1: The Marketplace Referral Loop
Vehicle information is inherently shareable in the context of buying and selling. When a used car buyer uses CarInfo to check a vehicle's history and finds something concerning (an undisclosed loan, a title issue, a challan history), they often share that finding with others in the transaction — the seller, a family member advising on the purchase, a friend who's also buying a car.
We made this sharing deliberate and structured. The "Share Vehicle Report" feature allowed users to generate a formatted, shareable PDF of the vehicle's CarInfo report. The shared PDF included CarInfo branding and a prompt to download the app. Each share that led to a new install was a referral with zero paid acquisition cost.
Within four months of launching this feature, shared reports were responsible for 12.3% of new installs — a meaningful non-paid acquisition channel that grew organically as the sharing behavior became habitual.
Loop 2: The Notification Re-Engagement Loop
The insurance renewal notification system wasn't just a conversion tool — it was a re-engagement mechanism. Users who had checked their vehicle once and left were brought back by a timely, relevant notification. When they returned for insurance renewal, they re-encountered the full product, often discovering features they hadn't explored on first use.
The average user who received and engaged with an insurance renewal notification showed a 2.4× increase in 30-day retention compared to users who checked insurance renewal on their own without a notification trigger. The notification created a habit touchpoint that wouldn't have existed otherwise.
Loop 3: The SEO Compounding Loop
The programmatic SEO architecture was itself a growth loop: more pages → more search rankings → more organic traffic → more user-generated signals (session time, engagement) that improved existing rankings → even more organic traffic. This is the standard SEO compounding effect, but the rate of compounding depended heavily on the quality and relevance of the content.
The critical investment was in making the programmatic content genuinely good — not templated boilerplate that search engines would eventually penalize, but real information that vehicle owners found useful. We hired a small editorial team to review and enrich the highest-traffic content pages. The quality investment paid off in ranking stability when algorithm updates hit competitors with lower-quality programmatic content.
18 Months of Building: The Full Execution Timeline
From the initial audit to the point where the growth rate was self-sustaining, the work spanned approximately 18 months. Here is an honest account of that timeline:
Months 1–2: Diagnosis and Architecture
Full funnel audit. Underwriter interviews (adapted for insurance context — we talked to insurance agents and adjusters rather than bank underwriters). SEO audit. Attribution infrastructure setup. Competitive analysis of Digit, Acko, Policybazaar mobile flows. No new features shipped — pure analysis and planning.
Months 3–5: Insurance Funnel Rebuild
New insurance entry point. Urgency-calibrated messaging. Form compression to 4 fields. UPI integration (this was the hardest technical lift — insurance partner API integration was complex). Trust signal architecture. Initial notification system. Soft launch to 20% of users. Early conversion data was positive but noisy due to small sample size.
Months 6–9: SEO Architecture
Vehicle model page buildout began. RTO location pages built. Insurance informational content published. Technical SEO improvements (page speed, structured data, mobile optimization). Zero ranking improvement visible for first 6 weeks — this was expected but tested organizational patience.
Months 10–12: Growth Loop Activation
Share Vehicle Report feature launched. Attribution data analysis and budget reallocation. Lookalike audience rebuild using high-LTV users (insurance purchasers) rather than generic vehicle owners. Notification cadence refined based on open rate and conversion data.
Months 13–15: Scale and Refinement
SEO traffic began accelerating meaningfully. Monthly organic installs crossing 200K for the first time. Insurance conversion rate reached 3.2% — a 457× improvement from the starting 0.007%. Insurance monthly revenue crossed ₹1 crore for the first time.
Months 16–18: Expansion
New insurance product categories (two-wheeler, commercial vehicles, previously only doing passenger cars). New insurer integrations (added two additional insurer partners for price competition). Geographic expansion of paid acquisition to Tier 2 and 3 cities. Monthly active users crossed 20M.
The Mistakes That Hurt
No 18-month product journey of this complexity runs without significant failures. Here are the ones that cost us most:
Failure 1: The SEO Content Quality Mistake
In months 7–8, under pressure to build page count quickly, we reduced editorial review on the RTO location pages and allowed a larger proportion of low-quality templated content to go live. We paid for this three months later when a Google algorithm update significantly reduced rankings for several hundred of these pages.
Recovering from this required rewriting approximately 400 pages — more total effort than doing it right the first time would have cost. The lesson: programmatic SEO is only sustainable when the content is genuinely useful. Volume without quality is a short-term gain and a long-term liability.
Failure 2: The Wrong Feature Priority
In month 11, I prioritized building a "vehicle service history" feature — a way for users to log their vehicle's maintenance records in the app. The premise was that users who maintained service records would be more engaged and would be a natural audience for extended warranty products.
The feature was built and launched. Adoption was low — about 2.8% of users engaged with it. The maintenance logging behavior required repeated, deliberate user action and provided no immediate value without historical data. We had overestimated how much users cared about maintaining digital service records.
Six months after launch, we deprecated the feature. The engineering time would have been better spent on improving the notification system, which had clear return on investment. I learned from this to be more disciplined about what the data actually showed versus what I intuited users might want.
Failure 3: Moving Upmarket Too Early
Around month 14, encouraged by the strong insurance performance, we attempted to add health insurance as a product category. The thesis: CarInfo had 20M vehicle-owning users who were financially active and might be interested in health insurance for themselves and their families.
The thesis was correct but the execution was premature. Health insurance is a fundamentally different product than vehicle insurance — more complex, longer consideration cycle, higher value, different trust requirements. We didn't have the product infrastructure to handle health insurance claims inquiries properly. Our customer support team was trained for vehicle-related questions, not health coverage questions.
We launched, got some early sales (the CarInfo brand carried some trust from the vehicle insurance success), and then received enough support failures and user complaints that we had to pull the product six weeks later. The damage to trust from a failed product launch, even a limited one, took months to repair.
Expand into adjacent categories only after you have the infrastructure to support them end-to-end, including post-purchase. A failed adjacency launch damages the core product's trust more than the incremental revenue of a successful one would be worth.
Results: The Full Picture
At the 18-month mark, here is where CarInfo stood:
Monthly Active Users: 3.8M to 45M
The 11.8× growth in MAU over 18 months came from three sources in roughly equal proportion: organic search (driven by the SEO architecture), referral/sharing (driven by the Share Report feature and word-of-mouth from insurance satisfaction), and paid acquisition (now buying higher-quality intent-driven traffic at lower effective CPP).
Insurance Revenue: 1200% Growth
The 1200% revenue growth on insurance represents the compound effect of three simultaneous improvements: more users (11.8× growth in addressable base), higher conversion rate (457× improvement from 0.007% to 3.2%), and expanded product range (two-wheeler and commercial vehicle added). The math actually understates the improvement in the core vehicle insurance product — the blended number is diluted by the newer product categories that were still ramping.
On a like-for-like basis (passenger vehicle insurance only), the revenue improvement was approximately 2,400% from baseline — the 680× figure that appears in summary decks refers to a specific insurance revenue line, not the blended total.
CAC: -30% Improvement
The cost per acquired user (properly defined as cost per user who completes at least one insurance purchase) dropped 30% over the same period. Given that total user acquisition volume increased significantly (we were acquiring far more users), the total acquisition spend increased substantially — but the efficiency improved materially. Each rupee of acquisition spend was generating 1.43× more insurance revenue at the end of the period than at the start.
What 45M MAU Actually Means
Forty-five million monthly active users is a large number. It's easy to say and harder to visualize. Let me try to give it some texture.
India has 295 million registered vehicles. CarInfo, at 45M MAU, was reaching approximately 15% of that total vehicle fleet every month — either directly or through someone who owned or was researching a vehicle. In many cities, CarInfo had become the default first check for anyone involved in a vehicle transaction. The question "did you CarInfo it?" was entering colloquial usage in automotive marketplaces, much like "did you Google it?" had entered general usage a decade earlier.
From a market access perspective, the insurance conversion at 3.2% on a 45M MAU base meant approximately 1.44 million insurance transactions per month being processed through or influenced by CarInfo. At average policy values around ₹3,500–4,200, that represented hundreds of crores of gross premium value monthly passing through the CarInfo ecosystem.
The 62% uninsured vehicle statistic that I cited at the beginning of this story — India's enormous insurance penetration gap — hadn't been fully closed. But it was being addressed, vehicle by vehicle, renewal by renewal, through a product that had learned to meet users at exactly the moment they had a real need and a real reason to act.
At scale, every product decision is a policy decision. When 45M people use your product monthly, the way you design it determines outcomes at a level that has real-world consequences for real people. That's the weight of scale. It's also the opportunity.
What CarInfo Taught Me
The CarInfo experience compressed a decade of product lessons into 18 months. Here are the ones I carry most consistently:
1. Intent-Driven Traffic is a Different Asset Class
There is a categorical difference between a user who arrives because they searched for something specific they need right now, and a user who arrived because they saw an ad and were vaguely interested. The first user will convert. The second user mostly won't. Build your growth strategy around capturing the first type. SEO, when done with genuine intent-alignment, is one of the most powerful tools for this.
2. Compound Failure is Stealthy
The 0.007% conversion rate was not produced by one bad decision. It was produced by seven moderate decisions compounding against each other. Each individual failure was survivable. Together, they were catastrophic. The lesson: when you see a deeply poor outcome, your instinct might be to look for the one thing that broke. Often you need to look for the seven things that are each 60% wrong.
3. Trust is the Infrastructure of Financial Products
You can optimize every other variable in a financial product — form length, price display, urgency messaging — and still fail if the trust infrastructure isn't right. Trust is a precondition, not a feature. You must establish it before you can sell. And trust is established through signals, not through claims. Design your trust signals explicitly.
4. The Moment of Intent is Everything
Meeting a user at the exact moment they have a need — not before, not after — is the highest-leverage intervention in conversion optimization. Insurance messaging to someone whose insurance expired three days ago converts at 4.8× the rate of messaging to someone with 60 days remaining. The same product, the same copy, the same CTA — but the moment makes all the difference. Design for the moment.
5. SEO is a Growth Strategy, Not a Tactic
When a product's use cases are inherently searchable, SEO is not a marketing channel — it's a growth engine. CarInfo's SEO architecture, properly built, became the company's most efficient user acquisition mechanism. The investment took months to manifest but, once established, compounded in a way that paid acquisition never could. If your product solves problems that people search for, build the SEO infrastructure first.