RSSGreek real-estate catches up with Web 2.0
Put together on October 17, 2008 4:29 pm by Dimitris
10 Comments
Following up to my previous post, I think it’d be brilliant if there was a meta-site to collate all the information in the major real-estate sites in Greece and tap on their collective broad user-base and its potential. Let me explain.
To begin with, such a service would include the few basic newspaper classified sites and the major real-estate agents’ ones. As it grows, further sites could also be added although that’s not necessary. The main concept lies in the meta-service scraping on a regular basis the major existing websites and processing that information. Processing means a number of things. Firstly, it should be able to re-display all ads in a homogeneous way regardless of their originating website. That way the visitor would do a single ‘master’ search and get back what there is to be found from all major sources. Secondly, duplicates could be removed from this summarising display – but more on that later. And thirdly and most importantly useful derivative information can be made available using data mining techniques.
But before I go into the data mining, let me expand a bit on the scraping. Now, by definition, classifieds are very concise pieces of text and as such can be analysed into some basic components like neighborhood, price, square meters of land, as well as type or storey if we’re talking about an apartment, extra features (like parking availability, number of rooms, type of heating etc) and even the street (if it’s stated). All this is possible because the vocabulary is particularly limited (perhaps 100-200 different words) or definable (by looking it up in a street directory). It’s important to note that despite the random structure and display in the various sites, the vocabulary remains the same across them and although some customisation will be required to process each site differently the essence of the information will be handled identically. (So if a classified has identical fields and comes from different sites it can be flagged as a duplicate.) Moreover, due to this peculiarity with the vocabulary one can also experiment with natural language searches. A proof of concept that this approach works can probably be seen in Rento which seems able to handle search queries in natural language. (Disclaimer: I do not know how well Rento does it – just that it does it.)
Furthermore, a lot of these individual websites rely to some extent to ads and marketing campaigns to supplement their revenues so depriving them of visitors (who would be able to view the ads in the meta-service instead) would be a major faux-pas. That would eventually lead to the ban the scraping script – if not to legal measures. So, it is important to ensure that once a visitor has found a preferred classified they are transferred back to the original website and that particular piece of text – to view the complete information about it, namely the contact details (or just the form to enter the code they will retrieve via SMS). This visitor transfer could also form the basis for a revenue exchange deal between the meta-service and source-sites. This is similar to the deal between YouTube and copyright holders of videos in their site: instead of flatly taking down videos with ‘their’ content they give them the option to share the revenue.
Back to the data mining features, there is a large amount of value that can be derived from these data. To begin with, simply the number of items for sale or rent is a useful figure – especially if known per location. Another type of result could come about by combining the location (which will vary in detail from neighborhood to exact street – depending on what’s available) and the price. For example, the most expensive and cheapest locations and their average prices can be determined. Conversely, alerts can be issued if any outliers from the average (i.e. possible bargains) have surfaced. Furthermore, accurate price ranges can also be calculated per square meter for each location. In addition to this, the meta-service can accumulate data over time (it should be able to remember when an ad has first appeared) and it can compare how different (or neighboring) location prices and other features evolve.
This meta-service can this way provide real-time statistics and become an as accurate as possible platform for the whole industry in a relatively easy way. This processed information can form the basis of a premium package that can be used to monetise the website. Such packages could be offered in conjunction with a free functionality – that would e.g. have limited features or would not go as back in time as the premium. An ad campaign can also be included in the website as well as an API offering access to the information – possibly for a charge if used for commercial purposes.
Now, scope-wise, an online real-estate business may be one of the few examples where going local instead of global is probably a better idea. Interest for a piece of land or site either either with the aim of living in it or housing a business usually comes from within the country – instead of from an individual or a company who has little knowledge of the culture and the geography. For example, large US sites like trulia and zillow, which admittedly, have considerable markets on their laps already, operate by focusing on the internal market. So, one could say the scope of such a meta-service should be mainly Greece.
However, there are some exceptions to the rule obviously and the first that comes to mind is tourism – a field where Greece is strong. The Greek islands but also the mainland have long been an attraction for foreigners (esp. Western Europeans) and a small fraction of them have settled in Greece. Indeed, lately a growing trend has been to transform rural Greece according to the model to other Mediterranean countries (e.g. Spain and Italy). This means turning large areas full of hotel complexes and villas, that are either put up for rent for the summer or the whole year or bought by foreigners (for instance, Russians, Germans, French etc). In the longterm therefore a niche global – or at least European – market may also develop that this meta-service could tap on.
What do you think overall – is this a good idea for a start-up?
(photo by oldyankee)
tags: idea |
Digg | |  |
Del.icio.us | |  |
Ma.gnolia | |  |
Newsvine | |  |
Technorati |
Comments
10 Responses to “Greek real-estate catches up with Web 2.0”
Leave a Reply - in English Please
kati san to http://www.landguru.gr/ ?
Yes, something like that. I did not know of landguru.gr’s functionality until I wrote this post – johnyzar of greekestate.gr also sent me an email informing me about it. Landguru.gr is a good approximation of what I had in mind – but could be further improved – esp. design- and interface-wise.
pliroforiaka tha sou po oti i agora gia real estate “online” mechri tora pou milame mporei na perioristei sta “rentals” (enoikia) ante kai ligo sta “leasing” (an kai poli spania) kai auto giati einai psiloperiergi i agora mesitias stin ellada… kai pali den to echo psaksei endelechos alla gnorizo oti iparchoun themata apo mia paneuropaiki eteria antistichi -sto “meta”- pou den mporouse na vgalei akri edo pera
to thema polisi akinitou einai psiloperiergo… eidika otan emplekete mesitis/kataskeuastis/endiamesos me oti an sinepagete auto (gia tin diki mou “meria” enoo… ).
oti theleis to email mou to echeis…
Thanks for the tips. Yes, exactly because real-estate is a large industry in Greece it’s only natural that it’s complicated and subject to all sorts of ‘irregularities’: there’s too much at stake. If you could dig up the European company which had to pull back from a similar Greek venture as what I’m describing due to excess market ‘immaturity’, that’d be great.
As for the actual closing of the deal (whether it’s a sale or a rental) it is hard to please all intermediaries (agents, construction, seller, buyer) but what I’m suggesting is about understanding the part of the market that makes it to the ads. The rest of the industry and the parties involved are outside its scope.
Thanks again for the comments:)
apla epikinonise na ta poume kai na se enimeroso schetika… apla den einai thema kratous…alla sintechnion pou kratane to offline “kleisto” kai ara kai to “online” echei tin idia tichi…
Hi Dimitris,
I am one of the developers behind landguru.gr
What do you find lacking in the current interface and design? Maybe we can improve it.
By the way the statistics we offer publicly are based on a small fragment of the data we collect and analyze internally. We haven’t yet decided how to utilize these data.
Regards,
Dimitris
http://daremon.gr
@daremon, thanks for following the blog closely (your tips on ForumView are still on our TO DO list).
Regarding the statistics you offer, I’ve made some suggestions in my post of what I would do if I had access to your data. Do you have or are you preparing processed data packages to be offered to individuals or real-estate businesses perhaps? Such a project might require considerable man-hours in connecting with other business, for example, but it might be worth it – esp. if you can provide them with information they need and an advantage over the competition.
Some random points regarding the interface. I was kind of surprised that you can’t select whether you’re renting or buying – apart from by the price range. I’d also like to see better integration with a map interface – e.g. whenever the address or at least the street is provided I’d like to see that somehow. For example, the list view in resulting ads is a bit ineffective – esp. when it can’t be expanded to more than 10 per page – a rento.gr-like embedded map might be more useful (at least for those where an address is available)
Other minor things include some explanations on how popular areas and ads are selected – as well as what the News section covers. I realise that the site was setup to run on ‘auto-pilot’ – at least to begin with – but it probably could do with a bit more attention – e.g. a proper blog.
Overall, though, your website is really good and contains real value – esp. in terms of the large number of websites you’re drawing data from.
Hi and thanks for your comments!
Reports+statistics: indeed there are a lot of ways to use the data and as I’ve said we aren’t clear on how to properly utilize them. We have some more advanced reports ready which we’ll probably make public soon. Furthermore we have done some analysis on a proprietary reporting tool and some in Weka and you *can* get interesting results from the data.
Renting and buying: right now we focus on property selling. You can’t search for rental ads.
Map interface: we really want to do it but it’s not easy. Having a random page’s HTML and trying to decipher the street (and street no) is hard! Also it’s hard to obtain raw data (e.g. all street names of Greece). However we are experimenting with it.
Popular areas and ads: what most people search and click respectively.
News: Articles from selected Greek sites (mostly informational ones) containing certain keywords.
Your most insightful comment is the “…was setup to run on ‘auto-pilot’ … but it probably could do with a bit more attention”. You are right and we’ll try to change that in the coming weeks.
Thanks!
Just a couple of things for your reply:
Statistics: what seems like interesting results to us might not be that interesting to a real-estate agent or someone who is looking for a house. I think to properly address that, one would like to focus on real-estate agents (as they are easier to understand compared to the crowd of buyers/sellers-you can ask them!): what info they already have and how they get it, what’s missing from what they have now, what they would really like but can’t get – and perhaps landguru.gr can provide.
Renting: If I look for a central neighborhood in Athens and put an upper price limit, I get some ads – presumably of apartments for rent.
General difficulty: The ideal (and very hard) solution would perhaps be to pick and choose the search engines landguru.gr looks at when a user makes a query – or a real-estate agent searches for processed information. This is perhaps too much of a hassle – but the next best thing would be to discard the many sources and keep the best ones. Best means most updated, most detailed and those that most people use. And that will make your work much easier/feasible. You don’t need to scrap all your work but you can give users the option between ‘lots’ and ‘good’ data.
Rental ads are in the index by mistake, usually because people mis-categorize when inserting their ad.
Keeping only the best sites would make our work somewhat easier although some problems still have to be solved in a completely generic way. Also it doesn’t really provide a solution to e.g. the map problem as most sites provide free text for the address (e.g. xe.gr)