Summary

A/B testing CTA button text alone can lift conversion rates by as much as 49% yet most organizations still guess at CTA copy, or worse, build and ship each option sequentially to find out. For India’s largest NBFC’s hyperlocal e-commerce marketplace which had scaled to over 50 categories, 40,000+ sellers, and 1.35 lakh+ SKUs that guesswork wasn’t an option on a high-value lead generation funnel where every development cycle competed for limited engineering capacity.

Bajaj Tech.AI used Adobe Target’s A/B testing capability to test four different CTA button texts in parallel, in production, on the platform’s Wheels Marketplace, a lead generation funnel for two-wheeler and four-wheeler products. Rather than building and deploying each option sequentially, all four ran simultaneously for a month, letting the team identify the highest-converting option with real data before committing any further development capacity to it.

Business Challenge

At the business' scale, every new feature and functionality requested by the business had to be built by the Bajaj Tech.AI development team and development capacity, however substantial, is always finite. Prioritizing which features to build first, out of a constant stream of requests, had created significant backlogs.

This challenge was especially acute on the Wheels Marketplace, a lead generation platform for two-wheeler and four-wheeler products. On the product description page, customers had to click a CTA button to proceed to the payment summary section and because both product categories carried high ticket values, the exact wording on that button carried outsized importance in determining whether a customer continued their journey or dropped off. The team faced a familiar but expensive problem: building and deploying each CTA option sequentially to test performance wasn’t a viable use of limited development capacity.

Summary: With development capacity constrained by backlogs and CTA wording having outsized impact on a high-ticket lead funnel, sequentially building and testing each option in production wasn’t viable, the team needed a way to test multiple options at once without committing development resources to each.

Solution Approach

Bajaj Tech.AI used Adobe Target’s A/B testing module to solve this without full development cycles for each variant:

• Front-end testing without full builds. Using Adobe Target, changes could be made to the platform’s front end and run in production by writing and executing scripts without requiring each variant to go through a full development and deployment cycle.
• Four CTA variants tested in parallel. The team ran four different CTA text options simultaneously in production, each shown to a different segment of the audience: “Book Now” (Pune), “Enquire Now” (Mumbai), “Explore” (Delhi), and “Interested” (Hyderabad). With four variants live at once, each individual option was seen by only 25% of the audience allowing a genuine like-for-like comparison rather than a sequential, time-staggered test.
• A full month of real production data. The test ran for one month before analysis, giving the team a large enough sample to identify meaningful differences in performance across variants rather than reacting to short-term noise.
• A data-driven development decision. Once the analysis identified the CTA text with the highest conversion rate from product description page to payment summary, that option and only that option was taken up for full development and deployed platform-wide.

Summary: Adobe Target let the team test four CTA variants simultaneously in live production over a full month, identifying the highest-converting option with real data before committing development capacity to build and deploy it.

Business Impact & Results

The A/B testing approach delivered value on two fronts conversion performance and development efficiency:

• A clear, data-backed winner emerged. The four CTA variants showed measurably different conversion rates from product description page to payment summary, giving the team clear, evidence-based direction rather than a subjective choice.
• The winning CTA now runs platform-wide, giving the NBFC consistent, optimized conversion performance across all customers rather than a guess that might have underperformed at scale.
• Development capacity was used far more efficiently. Testing four options in parallel, without building each one out fully first, meant the team could evaluate all four possibilities in the time it would have taken to develop and deploy a single untested option.
• The approach created a reusable model for future prioritization decisions — using Adobe Target to validate front-end changes with real production data before committing scarce development capacity, rather than relying on internal debate or sequential testing.

Summary: The test identified a clear conversion-rate winner among four CTA options, which now runs platform-wide, while saving substantial development capacity that would otherwise have gone into building and testing each option sequentially.

Pictures :

Case 1: “Book Now” in Pune

Case 2: “Enquire Now” in Mumbai

Case 3: “Explore” in Delhi

Case 4: “Interested” in Hyderabad

Key Takeaways

• On high-ticket, high-consideration purchase funnels, small wording changes to a single CTA can meaningfully shift conversion testing before building is far lower-risk than guessing and shipping.
• Tools like Adobe Target let front-end variants run in live production without full development cycles, which is what makes testing four options in parallel practical rather than prohibitively expensive.
• Splitting audience exposure evenly across variants (25% each, in this case) is what makes a multi-way test a fair, genuinely comparative experiment rather than a series of sequential guesses.
• Running a test for a full month, rather than a few days, gave the team enough real production data to trust the result rather than reacting to short-term fluctuations.
• Beyond this specific test, the approach created a repeatable model: validate with data before allocating development capacity, rather than prioritizing features by internal debate alone.

Conclusion

At the scale the business was operating, a marketplace spanning 50+ categories, 40,000+ sellers, and 1.35 lakh+ SKUs development capacity is one of the scarcest resources a platform has, and every feature built without data behind it is a bet against that scarcity. By using Adobe Target to test four CTA variants in parallel rather than sequentially, Bajaj Tech.AI helped the NBFC make a confident, data-backed decision on a high-ticket conversion funnel while conserving the development capacity that scarcity demanded. Any platform managing constrained development resources against a growing feature backlog faces the same underlying trade-off.

Looking to validate high-impact changes before committing development resources? Connect with our experts to build a testing framework that turns guesswork into data-backed decisions.