Redesigning how CarWale presents variant level differences for any given model, so buyers can confidently pick the right configuration without having to scroll through dense spec tables that test their patience instead of supporting their decision.
Every car listed on CarWale has multiple variants, often four to six, and the differences between them are typically small individually but consequential when stacked. A base variant versus a top variant might differ on 30 to 40 individual specifications. Most of those differences are irrelevant to most buyers, but the ones that matter to any given buyer are scattered across the list and easy to miss.
Our existing variant view solved this by dumping a full spec table on the user and trusting them to find the differences themselves. Session recordings revealed the cost of that trust. Buyers were scrolling up and down the table repeatedly, opening two and three browser tabs to compare variants side by side, and frequently abandoning the variant decision altogether before completing it. Several recordings showed users scroll the table twice, sigh visibly, and switch to a competitor's site to make the same decision there.
The deeper issue was that humans are bad at finding differences in tables of identical-looking rows. The 30-to-40 spec table format respects the data but disrespects the user. We were testing buyer patience instead of helping them decide. The framing question I wrote on a whiteboard at the start of the project and kept coming back to was: what is a comparison page actually for? The honest answer was that it is a decision-support tool, not a data display, and our existing page was solving the wrong problem.
I ran a research pass in two phases. The first was a session recording review across two weeks of variant-page traffic, sampled across hatchback, sedan, and SUV models to make sure the patterns generalised. The recordings produced a clear behavioural signature: scroll, scroll, scroll, scroll, abandon. Time on page was high not because users were engaged but because they were lost.
The second phase was a moderated study with eight buyers who were actively in the market for a car in the next 60 days. I asked them to use the existing variant page to pick a variant and narrate their reasoning out loud. The verbal data was as useful as the behavioural data. The phrases that came up repeatedly were 'I don't know what I'm looking at', 'this is too much', and 'I'll just talk to the dealer about this'. The last phrase was particularly telling, because handing the decision to a dealer is almost always worse for the buyer than making it themselves with good information.
The research output was a decision framework that mapped the buyer's actual decision process onto the spec data, identifying which spec categories carried decision weight at different stages and which were noise. That framework became the spine of the redesign.
The redesign reframed the page around a single question: what is meaningfully different between these variants? Everything that variants share is, by definition, not part of the variant decision and does not need to be on screen during that decision.
The new design defaults to a diff-style view that surfaces only the rows where variants disagree, with a clear toggle to reveal the full spec table for buyers who want it. Power users keep their data. New buyers get clarity. The split serves both groups without compromising either.
Within the diff view, differences are visually weighted by category. Safety differences and powertrain differences carry visual emphasis appropriate to a buyer's decision. Cosmetic differences like alloy wheel design or upholstery colour are present but visually quieter. This required a taxonomy of spec categories that the content team co-owned, which itself became a reusable asset for other pages on the site.
The mobile design drove the desktop layout, which was a deliberate inversion of how CarWale had historically worked. Mobile is where most of our traffic lives, and designing for the small screen forced a clarity that the larger canvas would have let us avoid. The column-swap pattern we landed on for mobile turned out to be a better desktop experience than the side-by-side variant layout the team had originally proposed.
Diff view as default, full table as opt-in. Power users wanted everything visible at once. New buyers wanted clarity. Defaulting to diff serves the majority while leaving a single-click escape hatch for power users. This was the highest-leverage decision in the project, and it was the one with the most internal resistance because it required trusting that the default would be used by most users (it was) rather than ignored.
Visual hierarchy by decision impact, not by data type. The standard pattern is to group spec rows by category (engine, dimensions, features, safety) and treat each category equally. We changed this to weight rows by decision impact, with safety and powertrain differences visually loudest and cosmetic differences visually quietest. This required real conversations with content and product about what 'decision impact' meant for different buyer segments.
Mobile drives the design. Most of CarWale's traffic is mobile. Designing the mobile experience first and treating desktop as the wider format of the same idea was unpopular at first because the team was used to working on the bigger canvas. After a few rounds of mobile prototypes, the discipline of designing for the small screen actually clarified the desktop layout. We were no longer adding bells and whistles that the small screen had to amputate.
A taxonomy that travels. The spec category weighting was a reusable asset. Once we had a shared vocabulary for which specs mattered most to buying decisions, the comparison page, the variant page, and the model overview page all benefited. The taxonomy outlived the project.
Diff view with optional full spec table
Time to selecting a variant came down noticeably in moderated post-launch tests, although the harder-to-measure outcome was the qualitative shift in how buyers talked about the page. Where the old page produced phrases like 'overwhelming' and 'I gave up', the new page produced 'helpful' and 'I knew what I was looking at'. That qualitative shift mattered more than any quantitative number we could report internally, because the page had stopped fighting its users.
The taxonomy of spec categories became a reusable internal asset, used by the comparison page, the model overview pages, and several editorial features. The output of one project turned into shared infrastructure for adjacent work, which is the kind of compounding return that justifies the upfront investment in research and structure.
The hardest mental shift for the team was accepting that a comparison page is not primarily a place to display data. It is a place to support a decision. Once that framing locked in, every layout question became easier to answer because we had a clear test: does this help someone decide.
I would push that framing harder upfront next time, because we spent more time than necessary in early reviews arguing about which spec fields to include rather than which decisions to support. The first question is rabbit-holes. The second question converges quickly. Future projects in this category will start with the decisions, not the data.