Become.com was in a different situation, with no mobile product launched yet. I suppose it was a tough call to decide to launch an iPad application as a first mobile product, but it certainly helped them to use the full potential of the device. What makes this application such interesting? It forgets the standard iPad / iPhone GUI patterns, takes benefit of animation and touch (like pinch and zoom) and for a long time makes a shopping engine fun and enjoyable to use!

When launching the app, you are welcomed with a list of shopping domains illustrated with big photo-realistic images ; you can scroll the screen to discover the various domains. In my opinion, the domain pages and browse experience can be improved; but the search is simply great: the result pages is displayed as a mosaic of products, displayed in a variety of sizes (randomly it seems). You can pinch and zoom product images, bookmark favored products by using drag & drop… ShopTop is an option to visually compare bookmarked products, or put them above a picture chosen from user’s images : a first glimpse of what could become a very powerful tool to see how a product could fit on you or in your appartement. You have to try it to see how good it is !

I was happily surprised to see such a good application from Become.com, as I’m not a big fan of the user experience at the main Become.com site. For me, this iPad version is far more polished than the main site; and hopefully this is a sign there are changes going on at Become.com. Congrats Become team for such a great app!

Homepage of the iPad application

Search result page

Bookmarking products

Product page

[Nicolas] Hi Jakub, could you tell me why have you built Periskop and what your goal is?

[Jakub] We began with typical customer problem – wanted to find good photo-camera for holidays. So looked for e-shop and each producer said that his products are the best, with which we didn’t agree. So we looked at CSEs and found that they do not provide characteristics of products but focus only on prices. Then we found some product reviews, but reading through them was difficult, because many of them are biased by writer and were to long to get precise information for our needs.

So we set up a company with 3 mathematicians and spend 9 months trying to figure out what algorithms could analyze content of product reviews to give people answer to such problems. Than we developed feature-based-analysis algorithm with our semantic ontology for e-commerce. In some time ago we have been backed by VC fund which believes that new CSEs will need to be more at product reviews and more in helping customers to make right decisions. And we are here.

[Nicolas] That sounds promising. What is the difference of approach compared to other CSEs for the consumers?

[Jakub] We started with a vision that any future CSE would be a place where customers decide which product to buy. So not only price will be important but also other parameters of products and users’ opinions. For example if we go to a shop we usually look for good product not the cheapest one. And the same story is on-line, but it’s much more difficult to present benefits of products in a concise way, so that customers can compare products easily. Nobody is reading 10-15 reviews per product (if there are so many) and then check them product by product. So we decided to do semantic analysis of content of products’ reviews to simplify this procedure. With our algorithms which analyze user reviews we give customers an easy way to compare products or to choose which one is better than others.

Based on the above we developed semantic natural language query for e-commerce. It was tough, but now one can search “good mobile phone” or “mobile phone with huge battery and cheap” or “windows mobile phone with good internet connection” etc. The same is ready for many other categories.

[Nicolas] Could you tell me more about your technology breakthroughs / innovation?

[Jakub] We aggregate product reviews for analysis purposes from almost 1700 sites. We are searching towards English and Polish reviews, and then apply our semantic analysis model based on feature based analysis of specific products’ features. In short it means that we could analyze what is written in product reviews and align these to specific products’ features. So in simple words, if somebody says about a photo-camera that is good, we get it. If someone else adds that it’s superb, we get it, and combine this information with previous ones. So for some statistic we know what people think about some characteristics of product e.g. battery for iPhone is weak instead of user interface which people said that it’s great. And so on, for up to 30 different features per category.

Based on these features of products we offer semantic natural language query for e-commerce, which simplifies product search. According to Jupiter Research from 2009 85% of users told that product search need to be redone, as more than 53% of customers prefer to search than browse for product discovery. So it’s huge market, and we have almost 2 years experience exactly in this direction.

We are working on similar model of visualization as Wize.com which was acquired by Nextag last month. So it would be nicer for users to get e.g. “diswasher for noice level” than do it with faceted way where user has to set decibels, etc.

[Nicolas] Thanks Jakub for those explanations, and thank you for your time.

Here is my take on the front-end experience of the Peryskop site (below are screenshots translated in English):

• Product pages with “Benefits / Disadvantages” section are really great, and it is clearly a nice benefit over the classic 5-stars pattern seen on most sites. It gives a feeling of professional review, while being built from automatic analysis (which is really elegant :) ).
• Category pages show a horizontal faceted navigation, with the classic feature-oriented facets; but the facet that is opened by default contains the soft attributes extracted from product reviews. In my opinion, the implementation could be perfected: the long drop-down list is not really convenient and may require editorial work to be really usable.
• However, the real power, like Jakub explained, lies in the search engine: having such “natural language” labels attached to products (like “easy to use”, “small size”, “good zoom”, “smile detection”) enables a better understanding of user queries. I would suggest to put more examples of such “natural” search queries directly on the homepage, to advertise this way of searching.

Congrats to the Peryskop team: it shows there is still room for innovation in the CSE world, and it’s a product I will follow carefully in the near future.

Category page:

Product page:

Multiple-image browsing at Zappos

It’s common to see multiple-image browsing on product pages, but it’s pretty rare on search result pages. And it always comes with more UI weight (icon to open the browsing widget, or prev/next icons to browse icons…). Zappos has implemented a very nice solution that adds no extra UI weight: when rolling over the image from left to right, the image changes, and you can discover up to 6 photos per product directly on the search result page! Sure, at first glance, this is not easy to understand: it took me a few seconds to understand how to move the mouse over (top to bottom? squared zones?…); but once you’ve got it, it’s a pleasure to use!

Usage-oriented facets at Amazon

In the CSE world, we often oppose hard goods and soft goods. It’s usually easy to build a product database for hard goods (electronics goods, domestic appliances, cars…), and then let users browse products, click on “compare prices” and land on product pages. But for soft goods (fashion, health & beauty, flower…), the number of references makes it a daunting task and so CSEs usually let users browse offers rather than products. In short, hard goods = easy to structure; soft goods = the mess to organize.

When it comes to facets, most CSEs and retailer sites implement feature-oriented facets and rarely usage-oriented facets… Using the analogy “hard goods / soft goods”, I’m tempted to use the terms “hard facets” for feature-oriented facets and “soft facets” for usage-oriented facets. Does it make sense?

Anyway, Amazon is experimenting with soft facets on its “Women’s denim” category in the US. When landing on the homepage of the category, you can choose “Modern chic”, “Timeless style”, “Everyday casual”, “Fresh & forward” styles. Basically those values act as shortcuts in the navigation (1 option in the soft facets = multiple options in the hard facets selected), and they give a kind of “buying guides embedded into faceted navigation” behavior that I really like.

Measy is currently limited to a few categories (Digital cameras, DSLRs, E-Readers, HDTVs, Netbooks, Smartphones). Quick review of the DSLR category :)

• The first question is about price, and relies on slider to define max price. Classic.
• The next questions are written with the “how important is…” pattern. How is important is brand? How important is size? How important is ease of use? … For each question, four values are available: “Don’t care”, “Slightly important”, “Important”, “Very important”.
• Choosing “Slightly important”, “Important” or “Very important”, you get complementary information or question… The “How important is brand?” option lets you choose your favorite brands using checkboxes. Behind the “How important is ease of use” question, you can define your level of expertise (from beginner to expert) using a slider…
• When you are done filling the quiz, the best product match is displayed, with a global score + scores for each question. Other results can be browsed ; and feature-per-feature comparison tables, along with price comparison and reviews services, are available.

Overall, I like the quiz idea and the natural speech used in the quiz (vs. the wording usually found in faceted navigation to describe facets). For sure, it’s difficult to please experts with such quiz. As such, I think that classic faceted navigation on large range of filters is complementary for experts. An interesting UX challenge would be to mix the quiz idea, faceted navigation and to some extent buying guides into a compelling user interface that could serve the needs from beginners to experts inside a product category.

The "how important is brand" question, with checkboxes to choose prefered brands

The "how important is ease of user" question, see how the sentences about Ergonomics and Menu system are written: natural speech improves empathy with the site.

The "how important is shooting in different situations" question, with checkboxes to choose common usages.

Result page

• A strong tagline: “Hunch personalizes the internet” with an asterisk that refers to an explanation text to have a better understanding and to explain the promise of Hunch…
• In the explanation text, note the “How do we personalize the internet? (glad you asked!)”… I like the ton of voice: casual, a bit humoristic, it develops the empathy with the site… Great!
• Visual representation of what the tagline means, with hot colors.
• A proeminent (yet pretty low in the page) yellow box with the call to action: “Answer 20 questions to build a taste profile” and Twitter/Facebook login options (but no custom Hunch option: this one is hidden behind the “Sign in > Create account” links; strange).
• Then there is a huge gap with other links displayed at the bottom of the page. I suppose it means: no distraction on the page, scroll down if you want more; and if you want more, you are not the average user.
• In the “help” section of those bottom links, there is a “serendipy” link which leads to a list of searches. Sounds like a SEO trick, but again I like the “serendipity” anchor text: it simply feels “human” !