LikeCube works with all kinds of location related data, like ratings, reviews, bookings, checkins, and metadata about places. And we work out the best way of tailoring this data for your users. So if you feel you need an expert to get you to market now, contact us.

This session is part of the Social Media track on Day 2 (at 14:40). Come and join us on Wednesday 6th October. 

And if you couldn't attend in person or want to know more, please feel free to contact us for a chat or view the session's presentation here.

True, one-to-one marketing can yield benefits that will enable you to increase your brand’s loyalty and reduce your reliance on SEO and SEM.

But with so many ways to do it, it’s hard to know where to get started and most importantly what is right for your site.

Here are a few tips you might want to consider:

1. From segmentation to machine learning.

The standard approach to personalization, which is CRM based, is through organising, analysing and segmenting the database and then crafting targeted marketing messages delivered at specific groups.

This approach yields benefits given the right amount of investment and skill sets, but may show limitations in addressing the fact that people’s desires aren’t static. The most innovative travel companies are now embracing machine learning to offer a real-time personalized discovery at the individual level. A bit like Amazon books recommendations, but for hotels, trips and restaurants.

Machine learning has the power to surface reviews from liked-minded people and answer questions such as “I liked this hotel in New York, where should I go in Paris”. Machine learning solutions leverage data, lots of data, so while you may not plan for it today, it is definitely worth start investing in it today, in terms of data collection.

2. Collect data, all you can.

The data you own is a potential goldmine. So collect it, keep it and make sure it is clean and usable. We typically see two types of data.

The one linked to places, which is any metadata that can describe a place, like place information and description as well as tags, attributes and editorial content. And the one linked to users.

User data can be explicitly given (user generated content) like ratings, reviews, checkins, wishlists. It can also be inferred via search terms, clickstreams and bookings either from your website or mobile application. All this data is worth something in terms of personalization.

So if you don’t collect it today, consider starting now. If you already do, think about what future data you should be gathering (like mobile activity or geolocation) to increase your data personalization capital.

3. Personalization is a journey.

Machine learning techniques are slowly filtering to mainstream, but still have a considerable way to go, because of the challenges in terms of understanding, implementation, performance and scalability.

However, you can start with low-tech personalisation steps, based around tailoring search results and navigation (e.g. using profile and history information), while allowing users to add content (tags, ratings, votes, reviews, etc).

Once you have collected large amounts of user data, you can use cloud-based solutions to convert that into deep personalisation and business intelligence. The nature of your data should drive your solution, rather than the converse.

4. Social graph or taste graph?

The two are complementary. Exploiting the social graph is important for virality, and, with Facebook’s likes, a good source of user data. But only few friends have similar taste to ours and, most importantly, we do not have enough friends to cover the places and things we need opinions/recommendations for.

With the taste graph, we can make use of “collective intelligence”, not to build trivial average ratings, but to create sophisticated mathematical models of users and their taste that can be used to make personalised predictions.

5. Value to the user first.

Personalization is first and foremost about engaging users in a more relevant discussion. One where they feel they will want to hear more, interact more, while getting to what they are looking for faster.

For example, if a user has booked a hotel, in your next newsletter to that user, you might want to consider their last booking as a reference point to future recommendations. In general, it is always worth asking: “Would I trust this information, would I find it relevant to me?”.

Often, the information is too generic and miss an opportunity to engage. In our experience, the more narrow-casted the message, the better in terms of generating trust and loyalty.

This may mean you may not reach 100% of your audience with personalization, but the one you will reach should have a far greater value than the booking they will do. These will become loyal to your brand and by-pass search engines to find you.

NB: this post was originally published Tnooz



Source: Five ways to make a travel site more personal (http://www.tnooz.com/2010/09/29/how-to/five-ways-to-make-a-travel-website-more-personal/)

The keynote was followed by many questions, a confirmation of the interest of Online Travel Agents in personalization.

Quoted by HEDNA, Ford told attendees: “We are experiencing exponential growth in data that can be used for purposes that we have never thought of before”.
"Qype is a service that is putting user data to new uses. Its aim is to match a place you like, say a coffee shop, with other coffee shop reviewed by people that are very much like you. Qype epitomizes the trend of narrowcast rather than broadcast. Narrowcast focuses on explicit data geared toward the users’ preference".
"Hyper-targeting in advertising is an emerging trend, allowing for customized delivery of messages to customers", noted Ford. "And we’ve only begun to tap into the many ways we can utilize mobile to deliver relative consumer information".

So if you are an OTA (online travel agent) or TPI (travel partner intermediary), remember that the era of machine learning in travel and leisure has started - and LikeCube is there to get you through the journey.

For more, check out the keynote slides here.

This week at TWTRCONF, the subject was debated by: Foursquare Dennis Crowley, Gowalla Josh Williams and SimpleGeo Matt Galligan. Yet no clear outcome as "Galligan admitted that the industry was going through a  'Wild West' period, and that it would be some time before the sector’s conceptual foundations and technical know-how shake down into an equilibrium" (kudos to Jesse Stanchak at SmartBlogs for capturing this).

We're not there yet, but part of the solution lies in the ability to improve the signal over noise ratio. Push marketing has been mentioned a lot as a monetization channel since the availability of geolocation became mainstream with smart phone. But it can't just be turned on because users would be bombarded by ads from every single shop they walk nearby.  And asking users to manually follow specific brands as a way to filter push marketing simply isn't a scalable solution. When you walk in a high street or city centre with 100 shops, you might be likely to shop at any given time in only 10% of them. And on a specific day and specific time, maybe 1 or 2 of them.

Behavioural understanding of places people might like is a way to deal with improving the signal over noise ratio. And information like checkins is one possible source of data for this.

Having the ability to predict places one might like is a marketer's dream. Geolocation alone isn't enough, but geolocation + personalization might well be the answer, because your chances of conversion are of a different order all together. One might say that if done really well (predicting the right places), you might even start considering push mktg as a feasible option (as signal/noise ratio will be good enough).

Cloud based recommendation engine for locations is what we do at www.LikeCube.com.

So if you are a web/mobile app capturing user location and want to experience these benefits, give use a call.



Tony Fish introduced the panel, raising the point that in companies quest for our personal data, we, as individual, have to think about where we are on the public-private data sharing continuum.

He brought the idea that there are 2 forms of public data:

• the paper/radiowave style public (tv, paper press) which has a very short lived footprint
• the internet style public (social media, youtube, twitter, everything internet based etc) which stays for years to come, with implications that we are still trying to understand and often learning at significant cost (at least in terms of reputation, brand etc).

Tony also raised another point which is that the way we perceive this continuum not only depends on our mood (on those bad days, we don't share as much) but also on the value we feel we get from the service we share data with (or actually give data to is probably a more accurate description in many cases).

The backlash recently experienced by Facebook and Google wasn't surprising. While many argued that too much data considered private was made public without the user's consent, from the end user point of view, there was probably a lack of perceived gained value.

One suggested we should actually own our data and trade it with websites for a cost.

If services would deliver more value, such thoughts wouldn't exist, but there is clearly an issue of lack of value to date.

So if you are a website owner, when you ask your users to give something (eg, I like this), it isn't abnormal that they expect something back in exchange right away, not 12 months down the road. The idea of instant reward is built-in many of the most successful sites, as a way to lower the barrier of engagement.

Qype has enabled such feature using the LikeCube service, with their quick rating game. Any visitor (anonymous new user or registered qyper) can play this game which allow you to rate on the fly places you might like and get instantly rewarded for sharing this info with personalized recommendations. A neat way to engage users to share their data with clear benefits.

So how are you going to entice your users to share their data?

There were interesting discussions about why mobile booking isn't really being adopted for hotel bookings in the UK. There seem to be a consensus that since bookings are done about 30 days in advance on average, people typically use their PC. But for more last minute activity, the panel agreed mobile might be more used more. Someone also suggested that the real mobile opportunity for travel is once the hotel/flight booking is done, where mobiles can be a support tool and new channel for ancillary revenues (eg once I am at the destination). In short, the mobile opportunity isn't about displacing existing revenue streams, but rather about exploiting new channels.

That said, our lack of time and mobile usability were also mentioned as key limiting factors on why booking isn't picking up on mobile. One person from the panel touched on that point when saying that search needs to be improved and that the mobile opportunity for travel companies is for smart people to come up with creative results.

While I agree much innovation is needed in the travel space, at LikeCube (http://www.likecube.com) we believe that mobiles aren't great tools for searching because of their form factor limitations. This directly impacts the experience of booking on a mobile. The problem with booking on a mobile is not booking, it's finding the place in the first place. On a PC, we got a plethora of information and screen estate to research and make that decision, but on a mobile, it simply isn't an enjoyable experience.

When it comes to music (last.fm or Apple Genius) or movies (Lovefilm, Netflix), we now assume and accept the services understand our taste and help us find what we'll like, bypassing the need for search alltogether. The same concepts are certainly applicable to travel, and even more on mobile

After all, wouldn't it be better use of our precious time if information would be more relevant to who we are, what we do or want and where we are?

One way to limit this content overload and give it some context or ‘trustability’ is through the social graph – what do your actual friends think about it? What would they recommend? Where have they been ? What have they liked?

You can see the value as it replicates real life, but there are clearly some problems and limitations with this approach too.

How many friends have you got? How many friends have you got in the places you want to go to? How many of those friends can you actually trust to give you a good restaurant or hotel review? But finally, and crucially, how many of those friends can you actually access when you need to, to give you those reviews?

A good way to think about this is comparing the power of Search to the power of Bookmarks. Let’s imagine for a minute that Google wasn’t there, and that for any information that you were looking for, you would start browsing from one trusted source, until manually crawling the web you find what you are after. How much more powerful it is to have the wealth of universal data indexed and then ordered by relevance to your query. There is no comparison.

In the same way in a leisure query we need to be able to access and filter vast amounts of data, but in an even more personal way that involves user taste. So what we are talking about is the need for personalised filtered search in the leisure context. Not sharing but filtering. Leveraging all data on the web, powered by all user activity – but made personal by centering it around your own personal activity. That means literally learning from the available data to give the best available answer.

At LikeCube we believe that this is when you really start having something useful.