Text to Messages — Improving WhatsApp Search in 4 Steps
Texting, as often as we do it has become a journal into our lives. Our texts tell a story of our progress. They hold many of our secrets, negotiated jobs, planned events, and all the I love you’s and I’m sorry’s in between. They are more than just texts. They are living memories and messages and messengers of moments… and I believe more meaningful search results could be crafted with better options to filter conversations with our favorite messaging apps we use and depend on on a daily basis. Here’s how I improved a product of one of the world's largest messaging apps in 4 steps. This is the first post of a monthy series in improving products during challenges I outline below.
TDLR: As much as I love my dear WhatsApp, its search feature needs some work. Therefore, I was inspired to improve it and used a 4-step process (1. interview the users, 2. do a micro usability test, 3. prototype a solution, 4. do a preference test) to do so. Using this process I improved my favorite app’s search feature which was ultimately preferred by the majority of long-time, frequent users in a preference test.
Towards the end of the summer of 2020, I was headed back stateside. For the better part of the year, I had been riding out the coronavirus in Madrid, Spain thousands of miles away from family. I was making the most of a blossoming relationship threatened by my career change in the middle of a pandemic and covid-19 putting a shear halt to traveling as we know it. A one month visit turned into nearly six. We had been laying in bed one evening with phone lit eyes thinking back on the memories we made, flipping through our WhatsApp thread, re-reading our messages to each other since we met nearly a year earlier. We knew the reality of me leaving didn’t guarantee that I would be coming back. We were holding on to everything.
We struggled to look through our thread, with some laughs, but still annoying. We couldn’t find our first messages exchanged, or the name she saved me under because she had no intention of learning my real name (American Ozuna). American Ozuna though?! It was hard trying to find when we shared our first kiss. WhatsApp knew the when but trying to filter through thousands of messages to find the when to read the emotions and relive the feeling of falling in love was frustrating. We could only search by word match and searching with “kiss” gave us a gazillion results… no thanks. This was a problem for me, for us. This was the beginning of this WhatsApp project.
Using WhatsApp’s word match search feature is a pain in the ass and inefficient. I assumed others felt the same (to some degree) as me. In this project, I hypothesized that:
frequent and long-term WhatsApp users would be more successful in finding content using my prototyped search compared to the current search feature. I’ll know this is true when +60% prefer the prototype.
I enlisted a 4-step process that ultimately improved WhatsApp’s search efficiency by talking with current users to figure out how they currently use the search feature, prototype a solution, and let the users determine which one they prefer to use.
1. Interview Current WhatsApp Users
This step is important because it gave me a chance to put my initial assumptions to the test. What I was trying to discover during these initial interviews was how satisfied/dissatisfied were users with the current search feature. I wanted to know all the ways they search through their message threads. I definitely thought they would be just as annoyed as I am. However, I was mistaken, initially. Here’s what happened:
The criteria were having to use WhatsApp daily or at least 4–5x’s a week (frequent) and have been a WhatsApp user for at least 4 years (long-term). One user was Spanish (P1 — participant 1), where 85% of Spanish people use WhatsApp exclusively (88% market penetration) and the other was American (P2 — participant 2) wherein the U.S., WhatsApp is relatively underutilized in comparison (20% market penetration).
I initially thought that WhatsApp’s limited search abilities would be a hindrance, especially for threads that span years but this was not the case, rather P1 and P2 worked through its limitations. They both used the photos and media section as a way to navigate to certain points in threads. P1 even stated that she enjoyed the challenge of searching through her WhatsApp threads and “playing detective”.
Using the photo workaround told me that there was some acknowledgment on their end of the limitations of the search feature — otherwise, they wouldn’t need to use a workaround, right? But it’s not enough of a limitation that they considered it a problem. P2 also said that he didn’t see the search feature being a problem because if something were that important that he needed to recall or find it, he would know exactly where to go and with whom he had the conversion.
Half-way through the interview, the problems I had with WhatsApp’s search feature were starting to feel like a “me-problem”… until the next step.
2. Micro Usability Test
The thing about doing interviews and asking people questions is sometimes (oftentimes) 1, they’re not honest, 2, the right questions weren’t asked, or 3, they aren’t aware of the problem. This is where usability tests are so wonderful.
Towards the end of the interviews with P1 and P2, I wanted to see them use the WhatsApp search feature IRL and show me that indeed there was no problem with the search feature. Here’s where things got interesting.
During the remote moderated micro usability test for P2, the visibility of the search feature on iOS was a critical incident. He couldn’t find the damn thing (not a problem, huh?).
I had assumed that long-time users knew where the search feature was seeing as they were long-time users and had adopted a workaround for navigating message threads. I was wrong to assume that and had written the tasks based on this assumption. Shame on me.
When we finally got into the search feature, P1 and P2 both felt limited by only being able to use word matches to complete the tasks and thought having the ability to narrow down conversations by dates would be more helpful. Because P1’s threads are so long (10 years worth of message threads and conversations), she said she gets frustrated when she can’t find what she’s looking for when she knows it’s there in the message (this is the same one who liked to “play detective”).
3. Prototype a Solution
With the information I gained from the interviews and micro usability test, I thought:
how can I improve the visibility and functionality of the WhatsApp search feature to improve frequent and long-term user’s ability to efficiently find content in their message threads?
This is where I used Figma to craft a solution I thought would address their needs. If the prototype were to be adopted by target users, I would need to avoid:
- Making the search process difficult or complicated to use.
- Eliminating the ability to use media/pictures as a way to filter through message threads.
- Compromise security and privacy.
These would serve as my functional constraints. My functional requirements needed to address in some way:
- Increased visibility of the search feature.
- Allow more options to search through message threads besides word match.
4. Do a Preference Test
Lastly, I needed to tie up some loose ends with this project and put everything together. Till this point, I had been taking the word of 2 WhatsApp users in terms of what worked well and what didn’t with the search feature. But how do I know their sentiments are shared by other WhatsApp users on a larger scale or are they the 2-off that want more options to search their message threads and like to “play detective”? Additionally, how can I determine if the solution I’ve prototyped addresses their needs? Conducting a preference test was a way to get these questions answered.
Preference tests can be used instead of A/B testing and in this situation was more useful because this was a mobile prototype and I got to ask participants directly which version they preferred. I also used the preference test to determine if the needs of P1 and P2 were shared on a larger scale. The current search feature and the prototype were both measured on users’ visibility satisfaction, ease of use, clarity, usefulness, and of course, their preference. Before I get into the results (rubs hands together), here’s how I recruited and structured the test:
- I recruited 30 of my friends (n = 29) in my WhatsApp network via text message and WhatsApp status with a link to the survey they can quickly follow (who knew more people used WhatsApp stories than Snapchat).
- Collected some relevant demographics (nationality, frequency of use, and frequency of use with the search feature).
- For the first 13 participants, they saw a short (~1 min) video of how to use the current WhatsApp search feature first (I couldn’t assume they knew where/what it was anymore, remember?)
- Then I showed them a click-through prototyped solution made with Figma.
- Ask them which feature (current or prototype) they preferred.
- Ask them why they preferred one over the other.
For the next 16 participants, they saw the prototype click through first, then the short video of how to use the current search feature. This flipping of the solutions was done to reduce recency bias as participants tend to select a preference for the solution they see last.
The Results
At last! Which search feature did the people prefer? Did they even have a problem with the current search features' visibility? And did they prefer my prototype or even find it useful?
By my standards, the prototype was a success, but this brought me to some other questions: are the users not using the search feature because 1, they can’t find it and/or it’s not useful to them, or because the visibility of it has been unsatisfactory and they can’t find it (~17%) so they haven’t used it and have adopted the workaround, such as using photos and media to navigate their threads. On the other hand, some who were indifferent stated that they were “use to it (the way the current feature is)” and it didn’t really matter about the search options.
While a large majority of the participants preferred the prototype (62.1%), what makes this useful is ~52% reporting to always or using the searching feature fairly often. Humans are creatures of habits. We talk and text in our own unique lingo. Therefore when searching for keywords, especially commonly used words, searching by keyword only would yield a lot of results.
2021 had arrived in Europe hours before it reached California. In came the flood of “Happy New Years” texts from friends — but thus far nothing new has come with the new year. We’re still holding on to hope for our relationship. Since I left Madrid in October our texts are filled with more I miss you’s than you’d want between two lovers — messages of loneliness and fear that this nightmare won’t end. In a way, we’ve grown numb to reading them, unfortunately. Quarantine has a way of mushing everything together and the days are all on repeat. We’re relying more than ever on our texts to become messages of patience and hope as we wait until it's safe to see each other again. Communication is all we have until then.
Check out the full report on dandramoss.com for more details.
