Intent, Discovery & Bots

Bridging the gap between consumers and brands. Moving purchase up in a consumer journey. Creating a remedy for home screen bloat. These are just a few of the problems Messenger bots are trying to solve for. We’re very familiar with these problems at Button, and the way to start thinking about them is around intent.

When there is context to determine a user’s intent on making a purchase that context can be used to narrow the inventory relevant to the user, and when served up to the user earlier will more likely result in a purchase. Context provides proof of intent.

When a user enters a messenger app with a bot, the bot only has a user’s input (and in the future their purchase history) as context. If the user’s input is specific — I need a bouquet of roses for less than $50 — and the bot has relevant inventory, then a purchase is highly likely. They’ve proven intent through greater context—the specificity of their request.

But let’s look at a different scenario. I’m a gal that could really use some new shoes, but I don’t have a specific style I’m looking for. I don’t have an occasion or outfit I’m trying to match and I don’t necessarily know what’s in style with the season change, but I’m positive I’ll need new wedges or heels in the near future. I have no idea what I’m looking for and know that I won’t make a purchase until I’ve figured that out, which feels pretty far away at the moment. I have low intent.

I open Messenger, find Shop Spring and get started:

  • What are you looking for today? Men’s or Women’s items
  • What kind of Women’s items? Clothing, Shoes or Accessories
  • What kind of shoes? Flats, Heels & Wedges or Sneakers
  • What’s your price range for women’s heels & wedges? Under $75, $75 to $250 or over $250
  • Here are five items we think you’ll dig…

At this point, the bot doesn’t know anything about me, only that I’m looking for Heels & Wedges for Under $75. If you’ve ever browsed women’s shoes, particularly on Spring, you know how many types let alone very particular shoes can fall under this selection, from bohemian chic lace-up sandals to ‘I’m going to Vegas this weekend and would like to be mistaken for Britney’ heels. But we’re now being shown five items that they think I’ll dig. I scroll through and on each individual item I can: “Buy,” “See more like this,” or “Ask a question.”

I don’t like any of them. Given the range of shoes that can fall under my selection, I’d love to see the numbers on how likely someone would be to like any of them. I don’t want to see more like any of them. So I say to this bot “I don’t like these.”

Bot: “We got your note. We’ll be with you shortly”

After 30 minutes: “Oh hi! Yep, it’s us again. Let us know if you’d like to continue shopping. If not, no worries, and we’ll look forward to future shopping adventures :smile:”

Well that was unsuccessful.

Dealing with a large set of inventory is difficult. Commerce businesses have dealt with this for ages, but for a user that starts with low intent, the discovery process is pivotal — it’s what shifts the level of intent. In the past 5 years, solutions have been focused on curation, without limiting access to the broader product offering. This bot is trying to curate while aggressively limiting. If a user doesn’t see something they’re interested in right away they’ll drop off. Discovery is all about converting low intent users, through either an immediate purchase or by generating interest in the brand and a future purchase. Prove to me you have something I’ll want, otherwise, I will move on (this is New York there are hot dogs on every corner for god’s sake.)

It’s a risky move. Spring is a fantastic app, that I’m fully behind. What they do best is to provide a curated sales experience. By limiting the product offering through the bot Messenger they’re risking the loss of converting a first-time user to a loyal Spring customer by saying “This is all we have to offer.”

Spring has had great execution thus far, and I’m confident they have a strong motive for their Messenger bot. But I’m not confident in this iteration’s ability to convert a low intent user. Linking existing users’ interests and habits sooner could help. They would need to make the connection to a user’s existing account prior to the curation process, which could be a friction point, but if they’ve already logged in through Facebook on Spring that connection should pre-exist…

I don’t have the information to even begin to solve the problem for their specific case, and there are lots of ways to think about how to generate context in the Messenger format. But limiting inventory without the context to do it intelligently is a dangerous move, especially for brands that need to be focused on growth.