I Tried Shopping on Facebook Messenger. It Didn’t Go Well.
The Emperor can’t sell clothes.
Yesterday Facebook unveiled what — if you’re in the tech world — everyone has been talking about non-stop for the last two months: Bots for Messenger. For the uninitiated, bots are simple: machines you can send messages to that can understand your message, and reply to it. If you’re my age, you may remember SmarterChild, the AIM bot from the early 2000’s to whom you could ask for movie times, the current weather, or simply have silly conversation.
As a founder of a mobile commerce company, I’ve been exposed to another level of this conversation, the concept of “conversational commerce”. Conversational commerce is the result of a shopping bot that, at its theoretical zenith, would understand who you are, what you like, and take any simple query (“I need some new shoes”) and return exactly the products you’re looking for, ready for instantaneous purchase. The reality is much less ideal, but Facebook is so jazzed about this idea that two of the four bots they highlighted yesterday were dedicated to buying things: 1–800-Flowers, and the shopping app Spring. (Spring, if you’re not familiar, is a well-hyped startup that has raised over $30 million to focus on mobile commerce.)
Facebook’s demo of Spring was very slick, so I decided to try these shopping bots myself. What follows is the summary of my time with two shopping bots on Facebook Messenger: Spring and Operator. Needless to say, the experience was not as advertised.
Facebook’s VP of Messenger made a point of saying that Facebook believes that UI (buttons, basically) combined with chat is how they foresee conversational commerce working in reality, and you can see that the initial experience with Spring’s bot isn’t very different from a simple category tree: gender > product type > clothing category > secondary category > price. This type of rigid structure isn’t necessarily ideal (there’s no easy way to find sales, or shop by size, for instance), but it’s easy enough to navigate. Upon making my (required) price range selection, I’m presented with five products. I’m not given any option to narrow by color, or fabric, or size. But fine. I swipe through each pair, one by one, and find one I like. I tap on “see more like this.”
Oof. Let’s count the things that went wrong here:
- I had to be switched to a human to answer a sizing question.
- Said human (!) didn’t understand my request for both a waist and inseam of 36", and had to be corrected.
- Upon being asked specifically for a reccomendation that would fit (my admittedly unusual dimensions), the human provided a single option that also didn’t have my size available (the closest was 34x36).
That I needed to be switched to a human to answer a sizing question was unfortunate, although I know first hand that sizing data between brands is outrageously inconsistent, if it exists at all. But it’s also an important thing you need to nail if you want to make shopping via chat at all easy for both customer and business. That shopping by size was so difficult, even for someone as tall as me, is a bad sign for the future of Spring’s bot.
But let’s keep going:
I believe my exact words were “You’ve got to be kidding me.”
- The human sent me another single recommendation for jeans that fit me that didn’t have sizing information on inseam, which is pretty silly for a customer that clearly cares about size.
- The jeans were outside of my originally selected price range ($75–$250).
- Had I chosen to buy these, I would have had to go through the same checkout experience as their mobile website.
While the 3rd sizing snafu in a row was definitely pretty ridiculous, what really gets me is that once I got to checkout the most dominant feeling I had was that I would’ve been better off just shopping on their mobile website. This is almost exactly what a bot experience shouldn’t be.
So sure, Spring was the marquee example in Facebook’s keynote, but maybe another shopping bot could do better? So I tried Operator, another hyped e-commerce startup that has raised $10 million to focus solely on conversational commerce. Surely their Messenger bot would be good!
Again, we see a more button-focused interface with Operator. I choose to forgo the emoji shopping at first to check out the 5 different sections they present in the carousel: This Week’s Picks, Mother’s Day Gifts, A Chill Night In, Health & Wellness, Fun in the Sun, and Home Essentials. Kind of a generic mix of products, but fine.
Once you’re in a section you’re presented with three products at a time. Tap “Like! See more” and you’re given three more products. Tap it again and you’re given another three and then the option to download their app. Let me say that again: I was shown all of nine products before I was directed into an app. Maybe bots aren’t the new apps as much as bots are the new app install ads.
But hey, they have what could be a nifty emoji shopping option, let’s try that:
Not only do they not have recommendations for literally any of emoji you might expect to use when shopping, but the one they do have is a product that 1) doesn’t make sense, and 2) is ugly.
And just to add insult to injury, when I finally, actually type words into my messaging window (a novel idea, apparently?) the bot doesn’t know how to respond.
It’s fair to say that Operator’s Messenger bot isn’t so much of a bot as a shitty interface to shop generic categories of products, three at a time, with a useless emoji gimmick. That just happens to live in Facebook Messenger. Sigh.
It goes without saying that shopping bots — and “conversational commerce” — have a long way to go if they are to even be proven useful at all. However I think that these two spectacular failures of execution can teach us a few lessons on how we might be able to convince people to shop on mobile:
Chat Is the Default Interface. Spring got this (kind of) right, Operator failed miserably. A user simply must be able to get an answer for any shopping related query, quickly. If you’re just going to put a bunch of buttons into a Facebook chat, you’re better off just sending people to your website.
Know Thy User. Chat gives you a unique opportunity to learn about your customer and their preferences. Don’t let that go to waste by not paying attention to their gender, their size, and the things they care about when shopping and chatting (hint: what people reject is just as important as what they like).
Show Me The Products. First and foremost, a shopping experience is defined by the products. The less information a bot has about my search, the more products it should be showing me to help me narrow it. A bot should show me as many products as it can, as soon as it can.
Checkout Must Be Seamless. I suspect this is primarily Facebook’s fault, but payment information should flow through Messenger (enter it once, and never again), and you should be able to pull my billing and shipping address from my Facebook profile. This is so obvious I can only guess that Facebook couldn’t ship this in time.
Overall, I remain unconvinced that shopping via chat makes sense. Shopping, especially for clothes and lifestyle products, is as much an emotional experience as it is an information problem, and I believe that people are usually better at picking what they want from a large assortment of products than they are at describing exactly what it is they’re looking for. But mobile shopping is still an unsolved use case (though one I’m working very hard to solve at Shophood), and chat may still have a part to play.
What’s clear is that if chat is going to help people shop, we need to do much, much better.