Carousell, a C2C marketplace, had successfully launched products that helped sellers promote their items in the marketplace. In this project we explored ways that would enable them to promote more than just an item — a collection of items or their entire profile.

Role

Product design, User research and testing

Result

Opened up a brand new way for users to promote their entire storefront in the marketplace. Increase in revenue and first time paying users

Introduction

Carousell, a C2C marketplace, had launched products that helped sellers promote an item — products like Spotlight and Bumps enabled them to gain additional visibility for one item in the marketplace. But, there was no way to promote multiple items or their entire profile in the marketplace. In this project we explored ways that would enable Carousell sellers to promote their profile along with a collection of items to gain additional visibility in the marketplace.

Finding the need through user research

At Carousell, product designers are expected to carry field research which involves recruiting the users, scheduling an interview and then conducting the interview at their preferred locations. If a prototype is ready then the designers are also expected to head out and do a quick usability testing if necessary. This allows the designers to hear what users think and any feedback they might have of the product—first hand.

Over the multiple user research that I conducted over the months, there was one recurring theme that most of the sellers touched upon. I heard them say things like: “Is there a way I can promote more than one item at once?” OR “How can I make more people look at my profile and the things that I am selling?” Digging deep through whys of this motivation we found that sellers wanted to be discovered by other buyers who weren’t quite sure what they were looking for. Sellers wanted to buyers to “window shop” hoping that they might find something “interesting.”

Heard during interviews

Sellers also believed that when buyers notice the breadth of items a seller has, they consider those sellers as “serious sellers” and they tend to trust them more than other casual sellers.

Sticky notes from one of the research

When we looked at offline marketplaces (a.k.a malls?), we saw that a need similar to this was already being taken care of. We saw billboards inviting people to a specific shop, placed strategically at multiple locations through out the malls. Our assumption was that sellers wanted to create an awareness around their shop when people visit these malls.

Spotted in the wild

The need was similar for sellers at Carousell, but with a twist of trust. Since Carousell is an open and free-for-all marketplace, sellers — especially merchants — are always looking for ways to establish their credibility over other sellers. Showcasing the breadth of inventory was one of the ways to do that.

Sellers wanted to showcase their breadth of inventory to establish their credibility in this open, free-for-all marketplace.

Jobs to be done

Based on the research findings, we were certain that there was a need to showcase sellers’ inventory in the marketplace. Coming up with jobs to be done stories helped us become clear about the sellers’ situations and motivations. We would use these primary and secondary JTBD stories to align team members and stakeholders throughout our discussions.

When sellers think about differentiating themselves from rest of the sellers, they want to showcase their breadth of products, so that they can establish themselves as “serious and trustworthy sellers” so that they can sell more and grow their business.


The Profile Promotion System

With the help of product manager and engineers we brainstormed an initial user flow to start with which comprised of two different parts:

  • Creation of profile promotion by sellers (initially from the profile)
  • Serving those created profile promotions to the right buyers at right moment

User flow - Take 0

Users would see a nudge on their profile that would motivate them to give this feature a try. After tapping they were required to select the items (up to 5) to be displayed when their profile is promoted in the promotion block. In the set up page, they can select the desired duration for which their profile will be promoted. They can also put a helpful title that would be displayed with the profile promotion block. If they don’t put it, we planned to show the snippet of their profile intro.

We were ready to test our first prototype and it looked something like this:

User Interactions - Take 0

We user tested this approach with a few sellers and found that the approach had opportunity to be revamped further. Based on the user research findings we were started to brainstorm in these directions:

  • Is this a profile promotion or collection promotion?
  • Why should a seller select the items to be included in the promotion? Sure this gives them a sense of control but will this necessarily mean an effective promotion?
  • How might we simplify the budget setting?
  • What are other ways we can display the profile promotion blocks that catch the required attention without hampering buyers’ tasks at hand?
  • What happens when one of the selected items get sold? Or is entirely different from the rest of the bunch? How do we prioritise?

Profile vs Collection Promotion

Post discussions with the team, we agreed that we can take better and more informed decisions by dynamically selecting the items to be displayed in the promotion block based on the browsing patterns of the buyer. This saved us from a lot of other complexities like what happens if the selected items get sold or if one of the items is different from rest of the items. This also gave us opportunity to improve on the buyers experience — buyers would only see items that they were currently looking for.

We also realised that in the current context, it will be easier for the user to complete the task if we simply focused on promoting their profile, instead of items. Sure, the profile is a collection of items but the main job of the the promotion block in the search and browse page was to direct perspective buyers to the promoted sellers profile to take a look at items that the seller is selling. We de-prioritised the promotion of a custom collection until we had creation of the collection in place and went ahead with the profile promotion.

The updated flow looked something like this:

User flow - Take 1

Based on the decisions taken and feedback received we tried an approach where we would auto-select the items to be included in the promotional block. We also iterated the design of the entire block and made it fit in one of the columns. We did this because the cost of paid products is loosely based on the real estate that they cover on the screen and we wanted to minimise the area to keep cost in control.

The previous design of the promotion block — an entire row — would have been extremely costly to buy for users since it took up the space for two listings. Making them fit a single column would exponentially reduce their price and hence had higher chances of adoption — an assumption that would be proved wrong when we consult the data.

In it’s entirety, the updated interactions looked something like this:

User Interactions - Take 1

Instead of showing the preview of the promotion upfront we moved it to a separate screen to lessen the visual cognitive overload for repeat users. We also provided a way to toggle between different layouts of the promotional block. While testing this approach internally, we found that the second layout had serious UX issues — it was hard to swipe to different items in that small area, resulting in unintended taps which took them to the next screen.

Entry for Profile Promotion

While we were working to simplify the entire flow for the product we were also testing out different design for the entry point on profile. We were testing different copy of the message and call to actions. Though it looks obvious in the hindsight but the design that had an outcome oriented messaging — what’s in it for me — worked better than the ones that were simply explaining the feature. We also learned that the prompt would work better if it looked like a part of the interface. Prompts looking too different from rest of the content could be easily dismissed as banners.

We learned that the prompt would work better if it looked like a part of the interface instead of standing out from it.

Here are some of the variations of the entry point that we considered:

Outcome oriented and explicit call to actions worked best

Profile Promotion Block

The latest flow was working better than all of the previous flows. But before we would further optimise the flow, we wanted to make sure that the design of the final result — the promotional block in the feed — was the most optimum solution. Few things that we considered while designing this block was:

  • It presents enough relevant information for buyers to motivate them to checkout the sellers’ profile
  • It doesn’t distract the buyers from the intended tasks — it’s easier to skip
  • It judiciously uses the real estate in the feed

To display the breadth of inventory in a limited space we also tried an approach where the items listed by the seller would auto scroll, generating enough information gap that encourages people to know more about the seller. The good thing about this approach was that it effectively communicates the breadth of the inventory a seller has while optimising the real estate.

Profile Promotion Block

Findings Answers Through Quantitative Data

While we were evaluating the best working solution among these profile promotional blocks, we felt we were simply debating our feelings and opinions — my product manger liked one variation while the iOS engineer preferred the other. We all had “feeling” that something was better than something else. At this juncture we started looking for quantifiable reasons to prefer one solution over another.

Click through rates and consistent health metrics were our core success metrics.

We went back to the original question that we usually ask ourselves before designing anything — what is the core goal of this design? How will we measure the success? And the answer came out to be this: click through rates (number of buyers visiting the promoted sellers’ profile) and the consistency in the usual health metrics. We looked at the difference between ad formats that were already in production and found that the double slots had higher CTRs and had no significant impact on the health metrics in the past six months.

Double panel slots had the higher CTRs without any effect on health metrics

Apparently, more successful design was the double panel slot. The ads in the double panel had higher click through rates (CTRs) and the health metrics were in check. Next, the challenge was to come up with a way that keeps the prices affordable. We did that by decreasing the height of the slot instead of letting it to span to full height of the columns.

Different approaches for Profile Promotion block


Payment options

After we were sure of the direction, we thought about simplifying the payment options. Spotlight was aimed at sellers relatively proficient in digital marketing and the workings of it. On the other hand Profile Promotion wasn’t. It was a brand new product and we expected it to be used by casual sellers too. Hence, instead of allowing them to set budget we packaged everything into three different duration based packages with flat pricing.

Moving from CPC to CPM for simplicity

Once the profile promotion was running…

Once the profile promotion was running, we needed to present the status of it. We wanted sellers to know that who it was performing and what else can be done to improve the overall exposure. We ran a quick survey to find out what sellers were interested in knowing and we find out these data points mattered most to the sellers:

  • Number of people checking out their profiles
  • Demographics of those people (male, female etc… )
  • Products that those people were most interested

Apart from including these data points, we also added the uplift in profile traffic after running the profile promotion. Our assumption was that these would give them a better sense of return on their investments. We have plans to include more data points such as demographics of the profile visitors etc… in the coming quarters. Once the promotion is complete, we show a snippet of what they gained in the recently ended promotional period.

Different states of stats


Shipped version


Hiccups

The way our ads backend is structured, adding a new ad format turned out to be harder than anticipated. At Carousell, when users browse, they see internal ads — ads served by Carousell sellers in the form of Spotlight and Bumps — and external ads — ads served by DoubleClick for Publishers (DFP). For every browse session, our engine performs a backend bidding with the contending ads to be shown and the one with the highest cost per click (CPC) wins and is assigned a placement in the feed. Up until, we kept the sizes of the ads — internal and external — similar to the size of listings. Either it fit in one column or the entire two columns. Because of this, even though there was a delay in loading of ads, users wouldn’t feel the content jumping.

We solved this by assigning dedicated slots and limiting the number of promotional blocks per users everyday. The promoted profile was supposed to be shown after each 40 listings and if we didn’t have a profile to show we could always display a prompt to promote your own profile of the same height. And since we limit the number per session, users wouldn’t see it that often.

Ideal vs Fallback

Learnings

The products and the solutions that we design will keep evolving but the learnings are something that stays with us and makes us better than what we were before starting the work.

Involving stakeholder and users early in the process is crucial and saves a lot of time later.

This project made me realise that apart from involving the stakeholders early in the process, it’s helpful to involve the users as well. Sometimes, I didn’t wait for the visual prototype to finish but went ahead and tested with paper prototypes (once, I asked a user to draw out what she imagines would her promoted profile would look like) and that gave me an idea of another direction that we could explore. Waiting for that perfect interaction to come through only sets me up for disappointment because the more effort I put in coming up with an idea the harder it was to discard that idea early phases in the process.

⦿