摩纳哥|官方

    1. <small id="oemes"><strong id="oemes"></strong></small>
        <meter id="oemes"><delect id="oemes"></delect></meter>
        <mark id="oemes"></mark>
              <meter id="oemes"><strong id="oemes"></strong></meter>
              
              
              <mark id="oemes"></mark>

              <input id="oemes"></input>
              <menu id="oemes"><u id="oemes"></u></menu>
            1. Showing posts with label collaborative filtering. Show all posts
              Showing posts with label collaborative filtering. Show all posts

              Friday, August 31, 2012

              Designing The Next-generation Review And Recommendation System


              It's unfortunate that despite of the popularity of social networks and plenty of other services that leverage network effects, the review and recommendation systems that are supposed to help users make the right decisions haven't changed much.

              Thumbs-up and thumbs-down or likes and unlikes signal two things: popularity and polarization. If a YouTube video has 400 thumbs-up and 500 thumbs-down it means that the video is popular as well as polarized, but it doesn't tell me whether I will like it or not. The star review system also signals two things - on average how good something is and whether it's significant or not. There are multiple problems with this approach. An item with 8 reviews, all 5 stars, could be really bad compared to an item that has 300 reviews with 3.5 stars. Star ratings alone, without associated descriptive reviews, wouldn't make much sense if there aren't enough people who have reviewed the item. Also, relying on an average rating alone could also be problematic since it lacks the polarization element. On top of it, the review and likes could be gamed.

              Pandora's as well as Netflix's recommendations are a good example of using collaborative filtering to fine tune recommendations based on user preferences. The system aggregates the overall likes and dislikes and combines that with your taste profile and a few killer algorithms to recommend what you might like. If designed well and if it has large user population, it does work. But, the challenges with such system are missing descriptive reviews and lack of ability to perform any analysis on it. If I dislike a song on Pandora, it doesn't mean the song is bad in the absolute sense. It simply means it doesn't match my taste profile. This isn't entirely true if I dislike a blender. In this case, a descriptive context is more meaningful such as I don't like this blender because it doesn't crush spinach well. People who care to make smoothies and crush ice may not care about this issue. But, these consumers have to wade through large number of reviews to determine the product fit.

              E-commerce sites review systems use the same descriptive as well as non-descriptive review systems, commonly used at all places on the internet, without any significant modifications, even if the expected investment of a user is much higher on their site. If I don't like a song, I can skip it. If I don't like a YouTube video, I can stop watching it and now if I don't like a movie I can stop streaming it. This does not apply in the traditional world of e-commerce. I absolutely need to make sure that I buy something that I like. Returning an item is a far more involved process than stop watching a movie. It's an exception, not a norm.

              Word of mouth and passive buying

              People shop in two ways: 1) they look for a specific product, research for it, and buy it. 2) they come across a product while not looking for it, like it, and buy it.

              The second way of shopping, passive buying, is as important as active buying. There are many companies with a business model built around this impulse or "serendipitous commerce", but they don't leverage collaborative filtering. I would happily read reviews of products written by my friends and people that I trust regardless of whether I'm looking for those products or not. Think of it as Disqus-style aggregated reviews by people that I trust in my social graph. This is like an online version of a cocktail party conversation where someone is raving about a new phone that he just bought. I'm not looking for a phone, but I might, in a few days. This could create new interest or expedite my decision process. This isn't done well in the online world.

              The word of mouth is still by far the best system for following recommendations. I invariably watch movies that my brother recommends to me and one of my friends will read all the books that I recommend to her. I have non-transactional relationship with my friends and family.

              Contextualized long tail 

              One of my favorite things, when I travel (leisure or business), is to try out at least one or two recommended Indian restaurants to see how Indian food compares from city to city and country to country (so far my vote for the best Indian food outside of India goes to London). While researching for a restaurant, I typically read all the reviews that I can find. Some reviewers are Indians and some are not. Also, for the reviews written by non-Indians, some are new to Indian food and some are not. In most cases people don't identify who they are and I end up guessing based on their username, description etc. These reviews, positive or negative, don't help me much to narrow down which restaurant I should try out.

              I have always found the best food at the most unusual places. All sophisticated recommendation systems would fall short of helping me find such an unusual place. These places are not the hits. They are the long tail. Getting to this long tail isn't an easy process - a lot of asking around, digging for reviews, trying out a few awful places etc.

              Privacy concerns and connected identities

              As the debate between anonymity and identity continues, there has been a little or no effort to get to the middle-ground, a connected identity. As a marketer I don't care who Jane is in its absolute sense but I am interested in what she likes and dislikes based on her collective and aggregated behavior across the Internet and beyond. This is not an easy system to build and consumers won't sign up for this unless there's a significant value for them. The popularity of social networks is an example where even if users are arguably upset about their privacy they still use it since the value that they receive far outweighs their concern. And remember the social networks follow the power laws. As more and more people use it the network becomes more and more valuable to the users.

              Why not design review and recommendation systems that are based on connected identities? Users don't want ads, the marketers do. If companies can focus on building good products, incentivize users to write reviews, and rely on great recommendation systems to connect the right users with right products they wouldn't need ads. The marketers are chasing the illusion of targeting the right users but the inconvenient truth is that it's incredibly hard to find those users and if they do find them, they don't really want ads. What they really want is value for their money. That is the inherent conflict between the marketers and end users.

              Using connected identities beyond reviews and recommendations

              Connected identities are also useful beyond reviews and recommendation systems. Comcast support is one of those examples where using connected identities could greatly improve their customer support.

              Comcast started using Twitter early on to respond to customers' support issues. It was a novel concept in the beginning and they really understood Twitter as an effective social media channel, but lately that model has turned out to be as bad as their phone customer support. When I tweet to @comcastcares someones gets back to me asking who I am and what issues I have. You follow me, I follow you, you DM me, I DM you my info, and after few minutes, we are nowhere close to resolving the issue. What if Comcast allowed me to attach my Twitter account to my Comcast profile? I will OAuth that, for sure. When I tweet, they exactly know who I am, what problem I am experiencing, and how they might be able to help me. This is an example of using a connected identity without compromising privacy. Comcast knows their customer's billing information; it's transactional information. But they attempt to use Twitter to communicate with you without connecting these two identities.

              I don't want to "like" Comcast or "follow" Comcast to be a victim of their spam and indifference. Comcast is easy to pick on, but there are plenty of other examples where connected identities could be useful.

              Users don't like to be sold at, but they do want to buy. Let's build the next-generation review and recommendation system to help them.

              Monday, May 21, 2012

              Data Is More Important Than Algorithms


              Netflix Similarity Map

              In 2006 Netflix offered to pay a million dollar, popularly known as the Netflix Prize, to whoever could help Netflix improve their recommendation system by at least 10%. A year later Korbel team won the Progress Prize by improving Netflix's recommendation system by 8.43%. They also gave the source code to Netflix of their 107 algorithms and 2000 hours of work. Netflix looked at these algorithms and decided to implement two main algorithms out of it to improve their recommendation system. Netflix did face some challenges but they managed to deploy these algorithms into their production system.

              Two years later Netflix awarded the grand prize of $1 million to the work that involved hundreds of predictive models and algorithms. They evaluated these new methods and decided not to implement them. This is what they had to say:
              "We evaluated some of the new methods offline but the additional accuracy gains that we measured did not seem to justify the engineering effort needed to bring them into a production environment. Also, our focus on improving Netflix personalization had shifted to the next level by then."
              This appears to be strange on the surface but when you examine the details it totally makes sense.

              The cost to implement algorithms to achieve incremental improvement isn't simply justifiable. While the researchers worked hard on innovating the algorithms Netflix's business as well as their customers' behavior changed. Netflix saw more and more devices being used by their users to stream movies as opposed to get a DVD in mail. The main intent behind the million dollar prize for Netflix was to perfect their recommendation system for their DVD subscription plan since those subscribers carefully picked the DVDs recommended to them as it would take some time to receive those titles in mail. Customers wanted to make sure that they don't end up with lousy movies. Netflix didn't get any feedback regarding those titles until after their customers had viewed them and decided to share their ratings.

              This customer behavior changed drastically when customers started following recommendations in realtime for their streaming subscription. They could instantaneously try out the recommended movies and if they didn't like them they tried something else. The barrier to get to the next movie that the customers might like significantly went down. Netflix also started to receive feedback in realtime while customers watched the movies. This was a big shift in user behavior and hence in recommendation system as customers moved from DVD to streaming.

              What does this mean to the companies venturing into Big Data?

              Algorithms are certainly important but they only provide incremental value on your existing business model. They are very difficult to innovate and way more expensive to implement. Netflix had a million dollar prize to attract the best talent, your organization probably doesn't. Your organization is also less likely to open up your private data into the public domain to discover new algorithms. I do encourage to be absolutely data-driven and do everything that you can to have data as your corporate strategy including hiring a data a scientist. But, most importantly, you should focus on your changing business — disruption and rapidly changing customer behavior — and data and not on algorithms. One of the promises of Big Data is to leave no data source behind. Your data is your business and your business is your data. Don't lose sight of it. Invest in technology and more importantly in people who have skills to stay on top of changing business models and unearth insights from data to strengthen and grow business. Algorithms are cool but the data is much cooler.

              Wednesday, March 21, 2012

              Learning From Elevators To Design Dynamic Systems


              Elevators suck. They are not smart enough to know which floor you might want to go. They aren't designed to avoid crowding in single elevator. And they make people press buttons twice, once to call an elevator and then to let it know which floor you want to go to. This all changed during my recent trip to Brazil when I saw the newer kind of elevators.

              These elevators have a common button panel outside in the lobby area of a high rise building. All people are required to enter their respective floor numbers and the machine will display a specific elevator number that they should get into. Once you enter into an elevator you don't press any numbers. In fact the elevators have no buttons at all. The elevator would highlight the floor numbers that it would stop at. That's it! I love this redesigned experience of elevators. It solves a numbers of problems. The old style elevators could not predict the demand. Now the system exactly knows how many people are waiting at what floors wanting to go where. This allows the system to optimize the elevator experience based on several variables and criteria such as speed, priority, even distribution, power conservation etc. This also means an opportunity to write interesting algorithms for these elevators.

              This is how I want ALL the systems to be - smart, adaptive, and dynamic. Just like this elevator I would like to see the systems, especially the cloud and the analytics, to anticipate the needs of the end users as opposed to following their commands. The context is the key to the success of delivering what users would expect. If the systems are designed to inquire about the context — directly or indirectly, just like asking people to push buttons before they get into an elevator — they would perform more intelligently. Some location-based systems have started to explore this idea, but it's just the beginning. This also has significant impact on designing collaborative recommendation systems that could help the end users find the right signal in the ever increasing noise of social media.

              The very idea of the cloud started with the mission to help users with elasticity of the commodity resources without having users to learn a different interface by giving them a unified abstraction. If you had two elevators in a lobby, you wouldn't use this. But, for a high rise with a few elevators, the opportunities are in abundance to optimize the system to use the available resources to provide the best experience to the people, the end users.

              Self-configuring and self-healing dynamic systems have been a fantasy, but as the cloud becomes more mature, dynamic capabilities to anticipate the needs of an application and its users are not far fetched. Computing and storage are commodity on the cloud. I see them as resources just like elevators. Instead of people pushing buttons at the eleventh hour I would prefer the cloud take a driver's seat and becomes much smarter at anticipating and managing applications, platforms, and mixed workload. I want the cloud to take this experience to the next level by helping developers develop such adaptive and dynamic applications. I almost see it as a scale issue, at system as well as at human level. If the cloud does promise scale I expect it to go beyond the commodity computing. This is why PaaS excites me more than anything else. That's a real deal to make a difference.

              Friday, April 10, 2009

              Amazon's Re-designed Review System Generates More Revenue But Has Plenty Of Untapped Potential

              Amazon's design tweaks to its review system has resulted into $2.7 billion of new revenue argues Jared Spool. Other people have also picked up this story with their analysis. I am wary of absolute revenue numbers tied to a feature to derive lost opportunity cost since a variety of other things could have driven the sale. It is wrong to assume that people would not have bought the products had the feature not existed. However I do believe it is a great step in the direction of making the review system more useful and drive more clickthroughs and conversions. Simply the presence of the reviews, magic number 20 in this case, motivates consumers to drill down into the details of a product and its reviews.

              Amazon has made significant progress in collaborative filtering through their review system and it is an exemplary of a long tail business model. It has helped consumers to gain transparency and has also helped expose issues with the products. This is not enough. As an e-commerce market leader I would want Amazon to continue innovating around their review system. This is what I specifically would like to see in Amazon's review system:

              Mining social media channels: Amazon.com is not the only place where consumers talk about the products. Consumers discuss product features and frustrations on Facebook, Twitter, and other social media outlets. Amazon has an opportunity to provide unified product review experience, a tool similar to ConvoTrack, by tapping into these social media channels for all the product conversations.

              Tag cloud as a visual filter: One of the ways to make sense out of large number of reviews is to generate a tag cloud from the raw text of the reviews. A tag cloud acts as a great visual filter to narrow down the reviews that the consumers are looking for e.g looking only at rebooting issues and not anything else while buying a router.

              Provide diverse search options: I want to search for the routers that have 4 or 5 stars ratings in the last 6 months. I cannot do that today. This search criteria makes sense. Manufacturers fix defects via firmware updates and models tend to improve as they mature. If the item had many negative reviews early on there is no way to find out without reading the other positive reviews whether the issues have been fixed or not. Higher recent ratings tend to correlate with mature product and satisfied customers.

              Re-think one-size-fits-all format: All the products sold on Amazon ranging from a book to a TV has the exact same review format. It does not have to be that way. The book reviews tend to be more subjective and philosophical where the gadget reviews are generally more fact-based e.g watch out this monitor does not come with a DVI cable. Re-thinking the format for the types of products being sold make sense e.g pros and cons section for the gadgets, similar books to the one that I am reviewing etc.

              Incentivise people to write reviews: Few days after consumers receive a product ask them whether they are satisfied with their purchase or not. Incentivize them to write reviews on the product; not only this helps generating more reviews per product but it also brings people back to Amazon to make more purchases. Make promotional email personal and relevant e.g.

              How are you liking the "Tipping Point"? Malcom Gladwell has authored his latest book called "Outliers" and we are positive you will enjoy that as well. Would you mind writing a brief review of "Tipping Point" and we will discount the Outliers for you by 5%.


              Closed-loop feedback channel: The current comments structure does not allow the manufacturers, authors, and the publishers to identify themselves and clarify the features, issues, and respond to consumers' concerns. The reviews are a great platform and a closed-loop feedback channel for the vendors to converse with the consumers. Amazon could certainly extend the review system to help create a dialogue between the consumers and the manufacturers.

              Monday, December 1, 2008

              Does Cloud Computing Help Create Network Effect To Support Crowdsourcing And Collaborative Filtering?

              Nick has a long post about Tim O'Reilly not getting the cloud. He questions Tim's assumptions on Web 2.0, network effects, power laws, and cloud computing. Both of them have good points.

              O'Reilly comments on the cloud in the context of network effects:

              "Cloud computing, at least in the sense that Hugh seems to be using the term, as a synonym for the infrastructure level of the cloud as best exemplified by Amazon S3 and EC2, doesn't have this kind of dynamic."

              Nick argues:

              "The network effect is indeed an important force shaping business online, and O'Reilly is right to remind us of that fact. But he's wrong to suggest that the network effect is the only or the most powerful means of achieving superior market share or profitability online or that it will be the defining formative factor for cloud computing."

              Both of them also argue about applying power laws to the cloud computing. I am with Nick on the power laws but strongly disagree with him on his view of cloud computing and network effects. The cloud at the infrastructure level will still follow the power laws due to the inherent capital intensive requirements of a data center and the tools on the cloud would help create network effects. Let's make sure we all understand what the powers laws are:

              "In systems where many people are free to choose between many options, a small subset of the whole will get a disproportionate amount of traffic (or attention, or income), even if no members of the system actively work towards such an outcome. This has nothing to do with moral weakness, selling out, or any other psychological explanation. The very act of choosing, spread widely enough and freely enough, creates a power law distribution."

              Any network effect starts with a small set of something and it eventually grows bigger and bigger - users, content etc. The cloud makes it a great platform for such systems that demand this kind of growth. The adoption barrier is close to zero for the companies whose business model actually depends upon creating these effects. They can provision their users, applications, and content on the cloud and be up and running in minutes and can grow as the user base and the content grows. This actually shifts the power to the smaller players and help them compete with the big cloud players and yet allow them to create network effects.

              The big cloud players, that are currently on the supply side of this utility mode, have few options on the table. They either can keep themselves to the infrastructure business and I would wear my skeptic hat and agree with a lot of people on the poor viability of this capital intensive business model that has very high operational cost. This option alone does not make sense and the big companies have to have a strategic intent behind such large investment.

              The strategic intent could be to SaaS up their tools and applications on the cloud. The investment and control over the infrastructure would provide a head start. They can also bring in partner ecosystem and crowdsource large user community to create a network effect of social innovation that is based on collective intelligence which in turn would make the tools better. One of the challenges with the recommendation systems that uses collaborative filtering is to be able to mine massive information that includes users' data and behavior and compute the correlation by linking it with massive information from other sources. The cloud makes a good platform for such requirements due to its inherent ability to store vast amount of information and perform massive parallel processing across heterogeneous sources. There are obvious privacy and security issues with this kind of approach but they are not impossible to resolve.

              Google, Amazon, and Microsoft are the supply side cloud infrastructure players that are already moving in the demand side of the tools business though I would not call them the equal players exploring all the opportunities.

              And last but not the least, there is a sustainability angle around the cloud providers. They can help consolidate thousands of data centers into few hundreds based on the geographical coverage, availability of water, energy, and dark fiber etc. This is similar to consolidating hundreds of dirty coal plants into few non-coal green power plants that can produce clean energy with efficient transmission and distribution system.
              1. <small id="oemes"><strong id="oemes"></strong></small>
                  <meter id="oemes"><delect id="oemes"></delect></meter>
                  <mark id="oemes"></mark>
                        <meter id="oemes"><strong id="oemes"></strong></meter>
                        
                        
                        <mark id="oemes"></mark>

                        <input id="oemes"></input>
                        <menu id="oemes"><u id="oemes"></u></menu>
                      1. 豪胜彩票平台

                        www.w88优德.com

                        龙8娱乐手机pt官方网站

                        网信的彩主页

                        大发大发888娱乐游戏

                        tt百家乐

                        网上在线打鱼

                        大乐购彩票注册网址

                        澳门网上在线赌博平台