Tuesday, 24 August 2010

About Facebook Places and why Foursquare, Gowalla, Qype and Yelp will continue to strive

Most likely, you've read about Facebook Places one way or another in the last couple days, even if you aren't a techgeek. After all the service is ultimately going to become available to about 8% of the world's population. So quite hard to avoid. But if you missed it, GigaOm did a good debrief here.

For now it's only available in the US, so I've only managed to get second hand experiences so far, but that didn't stop me from pondering whether and how I will use this new service, the impact it may have on other swimmingly similar services such as Yelp, Qype, Foursquare, Gowalla and the others and most importantly, what it means for businesses.

How will I use it?

I've been an early adopter of check-in based services, from always-on Google Latitude to action based Foursquare and Qype. I discovered that the service I used and the reasons for using it was heavily dependant on two things: the graph model and the graph members. Let me explain what I mean here.
  • The relevance of the model: Facebook is by default mostly closed to the world. So things I post on my stream, I expect only my Facebook friends to see. On the other side of the spectrum is Twitter. Things I post on Twitter, I expect the entire webuniverse to see. In the middle, you have Latitude, Foursquare, Gowalla, Qype, Yelp and the others. There are some check-ins I will share only with my family or friends, some other that I will share with business contacts and some I don't mind sharing with the world at large. Different use cases, different models required.
  • The graph members: I am also quite conscious that I don't want to pollute people screen estate with information that isn't relevant to them. As an example, when I was earlier this year at SXSW Interactive talking about how LikeCube powered recommendations for locations, Foursquare and Gowalla made a lot of sense to use, because 1-I needed an open model where I could find people easily and they could find me, 2-I needed a graph that was mostly business oriented and relevant with my time in Austin, TX (my business contacts from my days in IT services couldn't care less that I attended a panel on the future of geolocation). Foursquare and Gowalla adressed these needs well because most of the people there used one or the other service and these graph members were most not present in my Facebook or twitter graphs... or Linkedin graph for the matter.
Are the other services affected?
Well is it quite early to say, but in my personal case, because of the way I use my Facebook graph, I will continue to use Qype (for getting personalized recommendations to discover new places), Foursquare and Gowalla (for staying in touch with my SXSW extended crowd) and hopefully one day Linkedin. But maybe overtime, someone will solve the issue of how to manage all those graphs with respect to the business cases we use them for and I might revert to a single service.

So what does it mean for businesses?
The big opportunity is of course related to cracking to local advertising space. With the dismiss of the print Yellow Pages, a void was left. Google, Yelp and Qype have built their business on this. Now Foursquare and Gowalla are getting there through game mechanics and the check-in revolution. Now Facebook is clearly set on taking it to another level. With an unrivalled reach of 500m users, it will be interesting to see how things will evolve. You can find some advices on what to do with Facebook Places here.

Personally, I think that until such time one of these players manages to get the graphs playing nice with the business cases so that everyone adopt only one, you'll still have to consider what each player can do for you individually. Each is specific and can help you reach a different audience. So know your targets!

Tuesday, 2 March 2010

Is personalization the solution to the hotel reviews problem?

It isn't news that anonymous reviews are a problem - though it's now potentially turning into a serious legal battle as pointed out in the blog post: Hospitality industry rages at power of TripAdvisor's anonymous internet reviews.

Bob Cotton, chief executive of the British Hospitality Association, said hotels across Europe were seeking to persuade the EU Commission to overhaul the rules governing website reviews to ensure that they have been posted by genuine guests and not by rivals or people simply out to cause mischief.

The article goes on to state:

Industry leaders called for the rest of Europe to adopt the same standards of authentication that are already in operation in Germany.

While requiring authenticated users is a good way forward (potentially reducing the amount of spam, both positive and negative), it won't fundamentally change the problem: creating authenticated reviews won't prevent people from writing negative comments about hotels if their experience was truly poor; Tripadvisor will still be able to create a Top 10 list of dirtiest hotels; and some hotels will still feel unfairly represented (though one might argue that this is an opportunity for the named hotels to benefit from a surge in publicity).

The very idea of a global Top 10 is the problem.

The question is, do you align yourself with the 80% of negative reviews or the 20% of positive ones? Without knowing the individuals writing those reviews, how can you be sure? We often default for the majority, but that doesn't mean it's the best option.

One of the main complaints about Tripadvisor is that it offers no way to easily know if you can relate to the people writing reviews: Is your taste similar? Do you care about the same things in rating your hotel stay? What if comments are unfair, unjustified or plain unreasonable?

TripAdvisor would be far more useful if they offered personalized content, a personalized Top 10, with tailored reviews from people with similar taste to your own.

Monday, 7 September 2009

Who do you trust?

There has been a few reports that caught my attention over the summer. All have to do with trust online and the impact of social media on travel sites.

One of them was the Nielsen Global Online Consumer Survey, which includes an interesting measure of the forms of advertising most trusted by consumers.

90% trust recommendations from people they directly know, 70% trust consumer opinions online and brands. 69% trust editorial content.

Below is the full diagram and the press release can be found here.

"Recommendations from people known" stands very much alone at the top.

Not so surprising. We are far more inclined to listen to these than to those we don't. People we know have a great advantage as information providers. We are instantly able to filter what they say, because we know their taste. We know when to trust their judgment and when to discard their recommendations.

But that doesn't necessarily mean that the information delivered by them is better. I'm not saying recommendations from our social graph (people we know) aren't good, but they are usually too few to be really useful. A few simple reasons for this.
* We need information from people with taste somewhat similar to ours. After all, would you trust your grand parents recommendations to go clubbing?.
* We need a few advices to make an opinion. How likely is it that we know someone with taste similar to ours who's been where we want to go for holiday or more simply diner? The answer is not very likely.

An interesting complement to the Nielsen analysis would be to know the relative usage of these various forms of advertising. The top category is probably very small because of the reasons I've listed above.

And so we have to fall back on editorial content and consumers opinions, which are hard to trust, because we don't know the people who wrote them. While we may develop some level of trust with editorial content (as we get used to read the same person's article and develop what I would call a taste filter), we can't really do this with consumer opinions.

That's of course an area where we are very involved at LikeCube, We're changing the way people perceive User Generated Content (UGC) and we greatly enhance its value, quite a deal for web sites that have invested so much in UGC but are still struggling to see real value out of it.

In addition to providing really accurate personalized recommendations (some users have even qualified them as "quite spooky" in accuracy), our solution helps filter consumers opinions and extract the ones that are from people closest to us in terms of taste.

As a result, our solution can personalize each product page (think of a hotel or a restaurant review page) with reviews from our taste neighbors, reviews we can trust.

This solves a big issue for consumers when they read opinions ("Should I trust this person?") and I wouldn't be surprised if over time the 70% from the Nielsen report for "Opinion" moves toward the 90% from people we know, if not beyond.

Tuesday, 24 February 2009

Hotel industry in need of data or innovation?

In a recent interview from Sandeep Govil and Natasa Christodoulidou, the author was talking about the need to better understand your customer interaction with the hotel booking sites. Essentially, "hotels have a long way to go in talking to guests about the product, its attributes, locale, etc. in a way that each individual guest wishes to receive that information"

According to Natasa, "the key to working on CRM data gathering approaches in a manner that engages the customer as little as possible while obtaining the information that makes the biggest impact on the goals as a service provider is simply personalisation".

It is clear that there is a lot of value from capturing more data, as long as data is available. However, there is a significant challenge in the hotel industry. One that I usually describe with the number 4.

If users have used a site less than 4 times, then they are unlikely to use it by default for their next booking. They simply don't have much loyalty to the site yet.

Above 4, the users become more loyal to the site and use it as a natural starting point to their hotel research.

Now this is clearly over-simplifying a complex problem for marketers and CRM experts. But I think it's worth the analysis.

So in terms of numbers what does it means. Well the below 4 represent 95% of the bookings. 95%. To give you another representation of the challenge, the average number of bookings per user is below 2... Building user statistical patterns with such a low amount of data is unreliable. Hence the challenge of marketers and CRM specialists.

Now there is an interesting parallel that can be done here. 4 is also the number at which personalization solutions based on collaborative filtering (people who like the same hotels I liked also liked...) start showing interesting levels of accuracy and related impact on conversion rates. Before 4, consider yourself in a cold start situation.

We've addressed both sides of the challenge at LikeCube.

Leveraging our expertise in semantics, ontology and search, we provide very innovative solutions to address the cold start issue. Rather than asking data and trying to profile users as suggested by Natasha, our SaaS solution enables hotel sites to offer a powerful way for users to express their preferences and filter their results accordingly. Conversion rate improvements are expected to be of the order of 10% to 20%.

And for users that have crossed the '4' gap, our state of the art collaborative filtering solution could up conversion rate significantly more. The other key benefit here is that users get the noise filtered out automatically, including reviews and other form of user generated content. This of course results in drastically improved user experience. They will see first the content from their taste neighbors, the people most like them in terms of preferences, as opposed to hundred of reviews, most of which are non relevant to them.

So how are you going to differentiate from your competition to secure the 95% of users that are in cold start? And what are you doing to leverage your investment in UGC? Is data or innovation the answer?

Feel free to contact us at info@likecube.com for your free Site-Matcher evaluation.

Reblog this post [with Zemanta]

Monday, 13 October 2008

Agile programming, god or bad

I was having an interesting discussion with my colleague, Ismael, today about Agile programming. I was arguing the benefits of Agile, and the use of stand ups in the morning, allowing to bring a team together and quickly review the project progress.

Ismael agreed in principle but felt this rigidity of standups was against the whole principle of Agile... and he cited one interesting blog from Steve Yegge

Steve's definitely loving his job a Google!!! I love it I agree there is a lot to learn from Google's way of doing things, just like there is a lot to learn from Academia or the way startups work.

In my previous company, I remember when engineering started to introduce agile programming very much like google's way. The first thing they said to the business is: "forget being rigid about a delivery date. We will launch whenever we are ready to launch".

OMG, be assure a lot of adjustments had to take place. This wasn't well received because it felt that engineering wasn't tuned to the need of the company.

One specific area that was highly dependant on launch date was the launch plan and all the activities around it (training, marketing activities, enablement of partners, trade shows, press releases, analyst briefings, etc).

These activities take up to 2 - 3 months to execute. So the business needs to know in advance when the release date is, so that they can best invest their precious marketing $ to maximize the buzz in the market and attract the most new potential customers.

Clearly Google doesn't have that pb, nor does academia or pre-customer startups.

As Keith puts it in his comment to Steve's blog:

"Yep, and why is that? It's because most of us in our industry are writing sotware for paying customers (who might happen to share an employer with us) who have a "soft real time" idea of the value of the features we build: ie, the value of the features is at a maximum at some time t and declines, perhaps rapidly, after that. These three kinds of development are all quite unlike this.

Google's products and services seem mostly to exist for the purpose of making online advertising through Google more attractive, and that's a continuous process.

My client's projects generally exist to make some business process 1) possible, 2) faster, 3) higher quality before the competition do that. The business sponsors of these projects want to know when the benefit is going to start acruing. So they need schedules. Rightly or wrongly, that's what they need. And that's why (in contrast to the named methods that camed before) the Agile methods tend to put an emphasis on early delivery of business value, and hence the focus on schedule. In XP, though, note that we would rather reduce scope that slip a delivery. "

This said, at my previous company, the business has learned to be more flexible about release date and engineering more accountable about their progress toward a set goal and more transparent about the planned released date. All in all, best of 2 worlds.

I am now looking forward to leverage even more this approach with LikeCube.

Sunday, 12 October 2008

The good news about the IPO market...

Todd Kimmel's post, Advanced Technology Ventures, warns that the IPO market will basically remains idle for the next 18 months.

"The IPO window is closed and likely won’t re-open until 2010."

However this downturn will profit a few players in the markets, and these are the very small companies. The reason is that early-stage companies have very small burn rates and can more easily afford focussing on catching up or developping new technologies, while the medium size startups have to downsize and struggle to support an existing customer base.

As Kimmel's puts it: "This is not good news for anyone, except early-stage companies that plan to have their heads down for the next year or more working on proofs of concept and reducing technology risk. Those companies will be in a better position than those in need of capital to scale."

So are we going to see a few Davids out-smart Goliaths in the next 18 months?

Wednesday, 1 October 2008


I was working on getting up to speed with my new venture, LikeCube, which I am joining this month. LikeCube is a fantastic team combining over 16 years of semantic academic expertise. I have too say I got a lot of catch up to do here. Though my job will be less the technology and more driving the company through first customers and investment

This said, I needed a good refresh and this blog put Semantics in simple words... And I found it quite amazing how much LikeCube aligns with most of the points in this blog.

I am a firm believer that their is a huge opportunity in the market for companies seating between the content providers and the consumers. This is our sweet spot.

Saturday, 6 September 2008

The slow rise of social search

There been quite a bit of talks recently around social search growing fast, led by Myspace and Facebook.

As the Wall Street Journal pointed out, online marketing is dominating advertising with a healthy 20% growth, despite the economic turmoil.

Interestingly, simple advertising remains the preferred way, the low hanging fruit as opposed to complexly designed ads.

It also seems that Facebook is under exploiting their search monetization. Even though Myspace has tried to crack the space with their Google deal, it is believed to be under performing as per Google.

So when will the large social networks start leveraging their social graph to deliver better search? This remains a highly complex field to be able to calculate taste neighboors and apply them to do better search result or recommendation. Who will be the first?

Friday, 11 July 2008

Web 3.0. It's all about you.

So back to this Web 3.0 story.

The first generation of the web was all around e-commerce and making your inventories available for sale on the web.. winners: ebay, amazon. All this worked really well because no one except for the companies themselves could modify the content of their sites... so company centric, controlled content were at the core of of Web 1.0

The second generation of the web (The Community Strikes Back) saw the rise of the uncontrolled content generation. Blogs allowed the masses to instantly comment, share, communicate, exchange, express, basically participate in the web content, no longer under the corporate control of the few businesses in charge. Everyone could become a critic as well! Imagine those companies having spent millions on fine tuning their product messaging, and here comes the average student from his dormitory slashing through the product in no more time than it would take to mix your gin&tonic... ahhh... So the Web 2.0 is all about the community and the uncontrolled generation of content

Well here comes the web 3.0, leveraging the 1.0 marketplace, the 2.0 community and focusing on the last bit that wasn't addressed properly so far, you, me, the individual itself.

Today's challenge resides in making use of the amazing amount of unstructured content. And the best way to make use of it is to make it relevant to every single person... And one person trust as much his close friends (in the real world) as well as his digital friends (the communities) as well as the professional critics (cf web 1.0 companies)

So Web 3.0 is all about you, your tastes and what you like.

Who do you think will control the future, the companies with large amount of content or the ones able to recommend the right content to the right person, even if the content they own isn't as large? I bet on the last one.

Hello World!

The easiness of starting a new blog (this one took me 21 sec!) is one of the single most important factor in the pace at which the personalized recommendation industry has been growing. Indeed today everyone can be a critic. Well that's nothing new from the past, you could be a critic to your friends, recommending or not to go see a movie. But the main difference today is that you can broadcast your experience to a far wider audience.

This has started to generate quite a bit of problems with the professional critic industry. Like the music industry, they started rebuffing the underground movement, claiming that because they didn't have any proper credentials, Mr Lambda critics weren't as good as theirs. Well why would they? after all, the professionals have been witting them for years and years, know their market sector inside out, know the trends and fashions, places to be and the one to abandons. They even know what we, the un-professionals, are supposed to like. Hence their professional critics are tuned to a very specific ecosystem of readers, market sector, geolocation etc.

But unfortunatly for the professional, I personally care as much for the professional critic as for my friend's recommendation, if not more for the latter. The trust factor! Why would I trust more someone I don't really know as opposed to someone that I know, that I probably had many smashing parties with, seen fantastic movies and tried the best restaurants in town... according to my taste at least.

Well all this to introduce a topic that I've been interested about quite a bit recently, which is the personalized recommendation market, something fascinating, which is the essence of the web 3.0...

I'll come back to it in my following post.