When Twine announced itself to the world exactly one year ago, it claimed to be “the first mainstream Semantic Web application”. However despite raising millions of dollars in its quest to bring the Semantic Web to the mainstream, Twine has been beset by usability and performance issues in its beta period. Our own Marshall Kirkpatrick wrote probably the most brutal review. The post title said it all: Twine Disappoints After Semantic Web Hype.
However Twine has just launched publicly, confident that it is ready for prime time. I spoke with Twine founder and Semantic Web proponent Nova Spivack today to find out what’s changed, who’s been using Twine up till now, and where the service is headed in the future.
Growing Pains
When I met Nova Spivack one year ago in San Francisco, Twine was still in a private beta. Spivack described Twine to us at that point as a “knowledge networking” application with aspects of social networking, wikis, blogging, knowledge management systems. All built with Semantic Web technologies. It sounded exciting, potentially revolutionary.
The problem though, as we found out when we actually got our hands on the product, was that the theory and hype hadn’t translated into a usable app. Marshall identified major shortcomings in usability and performance. For example he explained that the service “doesn’t consistently grab summary text or tags for pages you save in Twine, it doesn’t recognize article authors as relevant people and it often captures summary information about the domain you’re on instead of a particular page’s content.”
In a word, Twine felt “half baked” in March ’07 according to our review of the product.
That Was Then, This is Now…
Unsurprisingly, in our interview today Nova Spivack was at pains to say that both usability and speed of the system have been improved for the 1.0 launch. Spivack told us that the main focus of this release is usability. Twine claims it has implemented “major performance fixes”, which Spivack said has resulted in “very dramatic speed-ups”. He explained that Twine has implemented caching on a lot of features, except for some of the more personal unique ones such as ‘My Items’ and general search.
Twine is also being re-marketed now as an “interest network” and not just a social network or bookmarking system. Part of the 1.0 launch is a new feature called an “interest feed”. Spivack told us that “interest tracking” can start from importing your bookmarks (from e.g. delicious) and users (from e.g. Yahoo Mail). From there Twine’s recommendation algorithm will find more items and people of interest – it apparently looks at your social graph, which “twines” you join (i.e. groups of topical interest), and other semantic data that Twine can surface. From there Twine automatically creates “semantic tags” and mines data to expand your ‘interest network’.
Spivack said that in the next 2-3 weeks Twine will release “next gen crawling and mining”, which will allow the system to index “pockets of the web” for you. As an example, Twine will index all the links and data in the first page of a website that you add to the system.
The Numbers & Demographics
How has Twine performed in its 1-year closed beta? Here are some interesting details about usage so far:
We were told that Twine has had 500,000 unique visitors in its closed beta, of which 50,000 are currently “active”. I asked how they define active: in this case it means a user who visits the Twine site at least once per month.
There are currently 20,000 ‘twines’, with 1 million pieces of content having been added to the system. 50% of the twines are private – Spivack said that many of those belong to companies ranging from small businesses to large corporations. Of the
public twines, they range from people who use them for bookmarking and hobbies, to people who use twines as a kind of blog (e.g. Nova himself does this), to “cool hunters” (kind of like BoingBoing).
The most popular twines reflect the early adopter audience: cool stuff, semantic web, politics, web industry news, etc.
The average age of a Twine user is 30 yrs, and they tend to be young professionals with medium-to-high income and education. They’ve used the product for both professional and personal reasons. 50% of users come from outside the US, but the service is primarily english language. Spivack told us that currently they provide basic level support for other languages, but this will be enhanced over time.
User engagement is currently at 12 minutes per session, which Twine says is “trending upward” (it started out as 6 mins). By comparison we were told that Digg is 2.5 mins, StumbleUpon 5.5 mins and Facebook 15 minutes – all according to Compete.com data.
What’s Next?
In terms of Twine’s roadmap, in 2008 their focus is on usability – which Spivack says is “currently 80% there”. In
2009 they will focus more on “surfacing the semantics”, meaning improving recommendations, search, and adding support for more kinds of data (e.g. currently users can add YouTube videos, but not all video sites are supported).
Spivack said that in ’09 users will be able to bring RSS into Twine (creating “a Reader on steriods”). Also users will be able to import emails. Twine hopes to data mine all of this.
And in 2009 an API is coming to Twine.
Monetizing Twine – Beacon-Like System Coming
Perhaps most intriguingly, Twine is planning to implement a new type of monetization system in 2009. Spivack had a big claim for this: “Twine will be for marketing what google is for advertising” (!). He said it will be the semantic equivalent of Google’s Adwords, but for marketers.
The system he described sounded similar to Facebook’s Beacon, in that it will insert
marketing recommendations into the core content. Essentially marketers will be able to post things into Twine, targeted to users interests. Twine will make recommendations in users interest feeds and some of this will be sponsored.
Spivack says Twine has some patents around new metrics for this – they’ll be able to see (in aggregate only) what people are doing with their content.
From the description, this sounds very similar to Facebook’s controversial Beacon, which was panned for infringing on users privacy. However Spivack claims it is “quite different from Beacon”. He also said it will be a CPA not CPM model.
It remains to be seen how this system will work and whether users will have reason to be up in arms. But given that social networks are so hard to monetize and CPM is under pressure right now from the economic situation, the Web industry is in need of innovative solutions. So this will be something to watch closely.
So, Will You Use Twine?
I have been an irregular user of Twine since joining the beta earlier this year. I’ve saved some bookmarks into Twine as a substitute for Delicious, but I tend to go back to Delicious for most of my bookmarks. For my purposes, Delicious is simpler. One issue that Twine has is that it tries to do an awful lot, which has the risk of confusing users.
Nova Spivack is pretty direct when talking about competing with other products. About Delicious he claimed that the ROI of putting things in Twine is better than Delicious – and it will get higher over time. At various stages in the interview he also said that Twine “can do better than google reader” (as an RSS Reader) and it is better than StumbleUpon, Digg, Facebook.
Nova Spivack can talk a very good talk – and I admire his passion for the Semantic Web and vision for his product. But for Twine to succeed, it needs to do the core things well. Behind all the talk about Semantic Web, beating Adwords and being better than Delicious, Google Reader and so on… is a product that basically is a knowledge management application. Can Twine find enough mainstream users interested in that core functionality? Only time will tell.
I for one will be giving it another go, if only because if Twine does fulfil its hype and becomes the first mainstream Semantic Web app – then it would be embarassing if I missed out on it.
