Amazon Polls

A user-generated product comparison tool intended to make shopping on amazon.com more delightful, community-driven, and engaging rather than purely transactional.

Overview

Polls were confined to Spark, Amazon’s experimental social platform. They had a high click-through rate (CTR), but a loosely defined value prop and fuzzy long-term vision.

The original ask was to identify and design new ways for polls to appear across the retail website. I took a step back and reframed the opportunity.

Opportunity

How can we proliferate polls throughout the retail website to increase engagement?

How do polls benefit customers, and where on the site can we best leverage their value?

What are polls?

User-generated content type allowing customers to compare 2 Amazon products and ask a question about them. As shown in the screenshots, those questions could be anything - resulting in a slew of disparate uses.

How are polls perceived?

I did some guerrilla testing with a few employees and non-employees to quickly surface pain points when interacting with a poll and to compile qualitative data around what people thought polls were for.

Example questions asked

  1. Do you use the Amazon shopping app?

  2. Do you know what Spark is?

  3. [Navigate to Spark until you find a poll]
    What is your initial reaction to seeing this type of content?

  4. [Go ahead an interact with the poll, and please narrate your thoughts as you do so]

  5. Can you think of instances when you’d want to publish a poll?
    If yes — when and why?
    If no — why not?

  6. Can you think of instances when you’d want to vote on a poll?
    If yes — when?
    If no — why not?

Insights

  1. Customers interact with (vote on) polls because it’s quick and easy.

  2. In practice, customers often have no connection to the poll content and therefore don’t find it valuable.

  3. In theory, customers see polls as a way to help make purchasing decisions (for themselves & others).

The conflict between #2 and #3 highlighted a tension I wanted to explore further.

Where do polls fit into the shopping journey?

I recognized a disconnect between the last insight around polls’ role in purchasing decisions and the current way polls were being used (as a quick, lightweight community engagement tool meant to inspire rather than aid in a purchase).

I started to segment polls’ purpose into 2 branches:

  1. Polls to inspire (lightweight/upper-funnel)

  2. Polls to aid in purchasing decisions (lower-funnel)

Lean UX Exercise with Product & Engineering

I led a brainstorming session with cross-departmental stakeholders to align on a prioritized list of hypotheses to test regarding polls, based on the highest risk assumptions with the least information. We examined assumptions around polls’ benefit to customers and polls’ value to Amazon as a business.

The outcome of the exercise illuminated conflicting assumptions around polls.

The big question: do polls help with purchasing decisions? If they don’t today, can they in the future?

Digging Deeper

The team agreed that polls should help with purchasing decisions. But did they?

I pulled data on the last 100 polls posted on the retail site to get a snapshot of how they were being used in practice. The majority fell under “general preference” - often these questions weren’t about the specific Amazon products themselves. Only 14/100, or 14%, of polls seemed to be about a purchase decision.

I also thought about what parts on the website were closest to a point of purchase, and about what kinds of questions customers might be asking.

A Path Forward

Proposal: put polls on product detail pages, where people are almost ready to buy.

Rather than have polls solely exist on Amazon’s social media feeds, I advocated for catching customers when they were already invested in a product, one click away from adding it to their cart. We could measure success by monitoring # polls published and # purchases made as a result, in addition to engagement (e.g., click-through rates, comments).

Detailing the Experience

Polls on a detail page

The basics
Polls are situated after reviews. A clear call to action, such as “Can’t decide? Ask the community” shows the intent is about asking for help. The page also pre-populates products rather than asking the customer to come up with them. “A” is the product they’re currently viewing. “B” is a suggested similar product.

Expanded functionality
What if a customer doesn’t like “B” and wants to choose another product? The platform should surface something more useful, such as an item from a past search or saved list.

Leveraging the “compare with similar items” table (future iteration)

The “compare with similar items” table appears on some product detail pages. It is already meant to help customers make a purchase decision, but often times the items appear similar: high star ratings, similar shipping, similar functionality, etc.

What if polls could serve as a tiebreaker?

Reflection

TAKEAWAYS

  • Dig deep to define the true customer problem. Often it must be reframed.

  • UX is more than tangible deliverables. It’s a tool / process for gaining alignment.

  • The power of qualitative data can help uncover the why in addition to the what.

IMPACT

  • Aligned team on a prioritized list of specific hypotheses to test and delivered a research plan.

  • Informed feature roadmap and overall product vision.

  • Documented findings and next steps.
    (Project was paused due to reorgs and resource constraints. Spark, the social discovery platform, was ultimately sunset.)