I had a really interesting workshop/discussion the other day at work, discussing how to integrate tags as a way to visualize a vast selection of services we offer. The more I thought about it, the more I came to the conclusion that building the perfect service is a two way rocket, first you find ways to create a lot of content/instances of services that is interesting to your customers. By doing this you will grow for sure, as long as the content you provide, or enables users to provide, is of interest and good quality. People are dragged to good content like flies to …, yeah you know. But what happens when you have 3 billion units of your content, how do you make sure that your customers can actually find all the good content you (or someone else) provide. It is a big risk that they only find the top layer and you as a service/content-provider do not offer as good service as you could if you could show all the things that is a possible match to your customer. In order to actually make use of all your content, you must make it findable. How do you do that? Maybe you just build your website according to Google guidelines and hope that the traffic you build is enough for actually getting both the “premium keyword”-traffic and the “long-tail”-traffic. But what about the people that is actually already converted and fancies your service/content? How do you go about building a service that helps your customers finding the bits and pieces that they never will find using Google, but that would serve you a great value as service provider if they did? I do not think user generated content is the killer application until search has become an even better killer application. You must be able to find all the content that is generated, and in this article I will share my ideas on what I think will make search the ultra killer application used right and enable you to actually make use of ALL the content you have. Search, findability and visualization of data is the number one competitive advantage when you act on a market where the competition is fierce and the content items comes in large numbers.
Tags
Tags are widely used today in social media and user generated content websites. Tags can be used for an individual to store content logically according to tags the user finds suitable for the content. The cool thing is that the more tags content get from different users, the more diversified and findable it gets because you can use these tags to create navigational solutions such as tag clouds, recommendations based on your most used tags and pushing otherwise hidden content to users that may be interested of things tagged in a special way. Tags are a blessing for finding content, but not always that easy to use, the user actually has to tag the content, that could be a hurdle for a lot of users, it is therefore important to make it easy for the user to tag content, the effort has to be the bare minimum. In order to make this hurdle even smaller, I suggest that you show examples on what other people tagged the content as and possible tags that may suit the content (based on software estimation)
System Tagging
When adding content to your system, making it available for your users to interact with, why not automatically tag it and enable a better findability for your content to your users. Typical methods that could be used for extracting these tags could be term extraction, that is extracting entities from the content such as places, names and things that occur in the content. By doing this you get the advantages of tagged content, but you do not only have to trust your users to add the tags, the system takes care of some of the work. This way you can offer a solution to tagged content without having an enormous user base that takes care of your content and tagging it. The good thing with tags is that you do not have to know where you are going with your system when you introduce tags, tags are chaotic and there are absolutely no problem adding more type of tags while the product evolves. If you work with categories or other single-man-made taxonomy you may run into trouble when your content grows and the needs for re-categorizing is increasing.
User Tagging
Let the user tag everything, it is good for the user, other users and for your business. It must however, more or less, be effortless to add tags and you have to, preferably intuitively, show the user why tagging is a good thing for him/her. If the user sees no reason for tagging, why would she? A typical way of using tags is to making them clickable in order to find related content. So in order to give the user the incentive to add tags, show how the content is tagged and add an option for the user to add more tags describing the content. Give the user help while they adding tags, maybe remembering old tags the user has added.
Ratings
Rating content is an interesting topic. You have probably seen everything from small forms for rating to just a thumbs up. Depending on what type of content you have and what type of search you would like to offer to your customer, different types of ratings may be more or less suitable to your solution. It is like setting up a democratic system, depending on how you ask and setup the rules for voting, you will have different outcomes. And depending on outcome you can use the result differently. A system based on the Facebook-like-functionality only indicates whether or not an item has a value to one or many users, while more sophisticated rating-systems can be used to find exactly what you are looking for in a large set of data (products in example). As with tagging, in order to make a rating system valuable for finding new products, the efforts of rating the product must be just enough, or if somewhat time consuming very rewarding to the end user. If the efforts needed for rating are too high, you will have lower frequency on voting users and therefore also a worse product than if the rating system were a tad bit easier.
Long-tail
If you have a large set of data or content that you are offering to your customers, it is of great interest to offer solutions to your customers that makes them not only find the most popular stuff, but also the more very specific items that may be of interest to them. This type of content consumption is called long-tail because there are few items bought of many instances. Companies such as Amazon found out that their long-tail approximately stands for x% of their earnings. The good thing with long-tail offerings is that being online lets you scale cheaply, and having an extra item for sale does not imply the same extra costs as it would in real life with bigger warehouses. So when you have the big set of content to offer to your customer, you should find ways to let them find even the not-so-popular items in your data sets as long as it suits their needs. Tools such as tags, recommendations and user-to-user can help you achieve this long-tail findability online. A complete user-centric search solution must be designed in such a way that it makes it easy to find content that are just not the most popular. You, as a user, must be able to find things that are what you are looking for, even though not so popular. Long-tail findability will create an explorative user experience where users actually feel like they stumble upon content they feel they would never have found if you have not handed them the solution to do so.
Semantic and Personalized
What does the user actually mean when he searches for something? When he writes “Thailand Weather” is the user doing it because he/she is:
- going on vacation to Thailand
- doing research on how the weather is changing due to global warming in eastern Asia?
- has just woken up in a hotel room in Bangkok, and needs to know the weather in detail for the day.
In order to offer the best possible search solution to your customers you should be able to better understand what the user actually means when he/she tries to find answers to their problems in your content. This is a major task, and often the meaning of something is closely connected to the person it self. In the above case, things such as search history, personal profile and geographic location could act as entities making the search a little more semantic and personalized for the user. There is probably a thin line to walk here as something that is really personalized may trigger personal integrity questions for the user. I would suggest that you carefully measure and ask your customers on the personalized features you add to your search in order to not break the thin line were it gets scary for the user, when they realize how much you actually know about them. Even if Internet is not anonymous, maybe it is important to serve that feeling to the user to some extent. Another suggestion is to make these features possible to turn off in some way, or at least describe why and how you know all these things about the customer and how it benefits the customer.
Real-time
Twitter took “real-time” to search. Now you do not have to wait 15 minutes or more for the big newsdesks to give you the latest info on what is happening. The problem is that rumors spreads faster and that you have a hard time verifying the truth factor of the things that actually happens real-time. So in order to offer real-time search and give value to the customer, you should have a product that has content constantly added and you feel that it is possible for you to actually find things that may suit your users as they happen. If you offer real-time search and the things you throw at your users do not match their expectations, don’t do it. Content Portals, Gaming Platforms, Newspapers and likes are products that would benefit most from having high quality real time search, that way they can sell whatever is created directly and hopefully increase the chance of creating a better margin on your revenue.
Recommendations
People like to buy stuff other people have bought. People like to buy stuff other people like them have bought. People like to buy stuff that people they look up to have bought. People like to buy things that are similar to things they already like. People like to discover things that are similar to things they already bought and liked. People like to get pointed in a direction when the supply of choices are huge. There are some companies that focus a lot on recommendations when it comes to help the user find more products in their database and to up-sell. Two companies that do this really good is Netflix and Amazon. I believe that this is the next commodity-functionality we will see across almost all online products that offer a big range of products to their end user. The value to the customer and thereby the daily incomes are far to big to not act upon. Today we see recommendations for books, music and movies, but no doubt that the algorithms for selecting products that fits the end user will grow in a lot of different product areas. Today it is very easy to start using recommendation on your online product, there exists different open source libraries that lets you integrate a powerful recommendation engine into your existing product database. An example is the powerful library Voogo running on PHP that is very easy to setup. Today we see websites that are built upon the concept of recommendation; Last.fm and another good example is the swedish site http://iglaset.se, where people can rate and get recommendations for beers, wine and liquors. Spotify has a lot of clickability to thank for the recommended similar artists. ITunes uses Genius to help people find new music based on their music library, and of course to up-sell. Recommendations are really a win-win solution for businesses and customers. If you look at social networks such as Facebook and LinkedIn they use recommendations as well. They ask if you maybe know this person. And they recommend on further actions you can take to further enhance the user experience of the service. There is no end to how you could utilize such a powerful tool as recommendations for your online service.
User-to-user
If all systems fail delivering the answer to a question, why not use other persons to help finding the right answers. This way you can take more detailed and complex questions to your system and have real persons answer them. Today there are some services that offer this kind of help (answers.yahoo.com, answers.com) and of course a lot of big brands offer forums and communities. Another part of user-to-user interaction could be crowdsourcing. Crowdsourcing is the combination of the words crowd, a lot of people, and outsourcing, letting someone else do the job. Lets say you would like to find a book on Amazon that has unicorn on its cover but not in the title or description. Why you would want such a thing is strange, but never the less, how would you go about finding it? It is a time consuming job trying to find it but if 10,000 people did the same search somewhat structured you could probably have your unicorn cover in less than an hour. In order to make this work, you got to give the user incentives to help, either money, status or fun. Google make use of crowdsourcing for labelling images in their image search. They have set up a game (http://images.google.com/imagelabeler/), two persons get shown the same image and they start to tag it. When they tag the image with the same tag, they are shown a new image to tag. This way, people get to play and Google gets better meta-data for their images.
Customer Support
Sometimes when you are lost, human help is the only thing that will actually help you find answers to your problems. Why not offer your customer a channel where they can communicate and find what they are looking for by communicating with real people? People cost more than software solutions, but, maybe there are problems that are worth finding solutions to guided by a dedicated customer support. What values do a Customer Support channel serve that software solutions for search do not?
Authority
Customer Support offers authority, they are the company, and the answers you get from them are official answers from the company.
Trust
Sometimes it is difficult to trust the information you get by searching and finding solutions yourself and contacting the official Customer Support offers you as a customer a channel where trust is the key thing they deliver. If Customer Support says that the plane takes off at 1900 on monday, then I trust them more than the table that the company has on their website. Why? Because a human said so.
Integrity
For some people it is easier to go through personal information where integrity is key, together with dedicated personell. Some people still go to banks for some errands even though it is possible to solve these things interactively online. Why? Maybe because they believe that some things are to sensitive to handle for computers and software (even if the customer support probably uses the same tools as are available online), and the feeling of having another official person handling it makes the customer feel more secure.
Verification & Sanity Check
Sometimes the answers you get from the software solutions available doesnt make sense, maybe you get different information from different sources and you as a customer really feel that you would like to know what is the correct data. Then Customer support can act as a channel for verification of data/sanity check. They may not have the answer immediately, but they can take on the role of finding the solution for you.
Conclusion
The more content you have to offer your customer, the better functionality you need to give them for finding the information/products/content that fits their need. In this article I have discussed and shared some ideas on how you can apply search in order to give your users a better user experience.
Andrew B. King also focus on optimization and tuning of your website, but the focus is on SEO, PPC-campaigns and Conversion Rate Optimization. He also tries to write some chapters on performance tuning, but when it comes to frontend performance tuning I recommend Steve Souders book instead. This book is really good for people who would like to learn more about optimizing PPC-campaigns and start tracking and incread conversion rates. I recommend this book for those chapters, which are really good and have help me increase incomes on the websites I have PPC-campaigns for.
Toby Segaran guides us in building smarter web applications based on Collective Intelligence. The book mixes theory and practice and shows how “easy” it is to actually create applications with community based product recommendation system and other intelligent solutions. I found this book interesting as a starting point on how to technically implement some of the web 2.0 features users have become used to use online.
Wikinomics is the book I have just started reading and it focus on how the community trend will change the enterprises. The focus is on how the masses create something of big value through collaboration. An interesting thing of course is the fact that a lot of people interact in larger communities building a greater value (Wikipedia, Youtube, Linux etc, etc) without getting paid, still the quality of the outcome often exceeds the alternatives built traditionally with labour and investment banks.
. For online stores such as Amazon a big part of their income is from hard-to-find items. Because of the size of their inventory they are going to attract a huge number of customers searching for such items. The long tail turns out to be as important as the sales of best-sellers and other items that are of interest to bigger audience. So Long Tail refers to targeting smaller niche markets often. When a lot of smaller niche markets gets targeted, as in Amazons case, this tends to create a big impact on sales.