Archive for the 'Nestoria Rank' Category

Drop in pageviews

Published by Ed on February 10, 2010 in Nestoria Rank and New Features. Comments

Fellow Nestorinos!

One thing we occasionally get asked about here at Nestoria is how many pageviews users generate. My typical response, generally to the horror of the question asker, is “too many”. Why? Well, because it’s easy to create more pageviews – just don’t give people the best answer. Instead force them to click “next”, “next”, “next” to find what they want. A good search engine should be presenting relevant content to users as quickly as possible and sending them of to that content. More pageviews would indicate we’re doing a bad rather than a good job.

Why do I bring this up now? Well, a few weeks ago we began intensively testing a fairly significant change to the Nestoria search results page, and we’ve now rolled it out to all users. The goal: reduce pageviews dramatically by helping users get to their dream home as quckly and simply as possible. Not to bore you with the technical jargon, but we’ve now gone to a full AJAX implementation – basically relevant content should just “automagically” appear as you move the sliders, the map, etc.

Screen shot 2010-02-10 at 19.15.40

So far feedback and measured behaviour has been very positive. As with any change, it’s unlikely we got everything perfect on the first try (actually this isn’t the first try – we’ve tested many implementations of this), so please let us know your feedback. And big thanks to all of you who have already sent your thoughts.

Whether you like this change or not, of one thing you can be certain, even as you read this note, a significant percentage of Nestoria users are exposed to a new test. The path to perfection is endless. We grow by walking it.

One final technical note – big thanks to everyone working on jQuery. We made the switch, and are loving it.

Same Same is Bad Bad

Published by Ed on April 18, 2009 in Nestoria Rank and New Features. Comments

Fellow Nestorlinians,

As is our habit, from time to time we like to update you on our efforts to improve the quality of the property search experience on Nestoria. We put a great deal of effort into addressing all aspect of a high quality search experience, not least being relevance. I’m thus please to reveal that this week we went live with a significant improvement to the algorithms we use for detecting duplicate listings. Because we aggregate from many sources, it often happens that we have the same property from mutiple partners. This is clearly not an optimal user experience.

Techncially detecting duplicates isn’t always trivial as the information available to us from different sources can be, well, different.

Duplicates are evil!

We’re pleased that we’ve been able to make good progress on this important part of relevancy. As with all aspects of improving quality, the work is never “done”. No doubt there are still the occasional dupes lurking out there. Please let us know when you spot them.

On a related note, a few months ago I dashed off a post about quality (it’s not the same as quantity) and why we believe it’s better to show our users fewer high quality listings than many bad listings. Our finetuning of deduplication is another step in that direction.

Finally, for our international fans, rest assured the same improvements have been rolled out across Nestoria in all the markets we’re active in.

Enjoy the weekend!

Quantity and Quality

Published by Ed on September 11, 2008 in Nestoria Rank and search. Comments

Fellow Nestortonians,

in your internet wanderings you may have come across long-time property industry expert Bob North’s recent post over at Estate Agency News discussing the comprehensiveness of various property websites (portals and search engines). Bob’s monthly analysis is always interesting, if you’re not a subscriber I recommend you become one. It’s a must read here at Nestoria HQ.

His attention this month focuses on a seemingly simple question: “How many property listings do the various players have?”. We get asked about this all the time. As we’ve always tried to stress here on the blog in our occasional discussions of Nestoria Rank, comprehensiveness is an important metric, but only one of several, the others being usability, freshness, and of course relevancy. Nevertheless, probably because it seems so easy to measure, comprehensiveness seems to be the metric people fixate on. We think it is given far too much importance, despite having some major weaknesses.

First of all, which listings get counted? We believe users are interested only in relevant results. So, should we count listings that are sold or “sold subject to contract”. Given that we focus on helping individuals find a place to live, should we include commercial properties? We think not, and do our best to exclude such listings from our index. As another example, we often see things like parking spaces listed. Of course people need parking spaces for their cars, we just don’t think they belong in a residential property search engine.

Secondly there’s the question of duplication. Often we have the same listing from multiple sources. Is that one listing or multiple? Another good example is when we see very similar units all in the same building or development. Sure different flats may have slightly different dimensions or prices depending on the view and such, but is a user really seeing many different listings? Or are they seeing one option in their choice of where to live?

Next, there’s the issue of sampling bias. When comparing only one or to areas it’s easy to (knowingly or unknowingly) skew the results.

Finally, of course, comparing search engines and portals is a bit like comparing apples and oranges. The two provide very different user experiences with only marginal overlap.

So in conclusion, yes, as Lenin once said “Quantity has a quality of its own”, and of course having a comprehensive offering is critical. But, as with almost every web metric it’s important to understand the meaning behind the number. Comprehensiveness is easy to misinterpret and even then it’s only one ingredient in the property search recipe. Rest assured dear reader we’ll keep our attention on all aspects of creating a compelling service.

Please let us know what you think of our efforts and how we could do better.

Nestoria Rank update – December 2007

Published by Ed on December 28, 2007 in Nestoria Rank, New Features, search and technical. Comments

Fellow Nestorwegians,

The time between Christmas and the New Year is a natural chance to look back on the last year and what we’ve accomplished. The next few days we’ll do a few more reflective posts summarising some of the accomplishment of the past year.

First up we thought we’d revisit Nestoria Rank, the algorithm we use to try to present relevant results to property searchers. Over the last few months we’ve continually experimented with changes designed to get people to the right house or flat more quickly.

Some of these changes have been very visible, others are more related to the algorithm itself. Let’s go through some of the more visible improvements first:

We now allow you to search by property characteristics.

screenshot of searching for property characteristics

This means searches like: an unfurnished two bed flat to rent near Little Venice with a balcony are now easily possible (right now we have 111 of them in the database)

Likewise, we’ve added the ability to filter your search only for “new builds”

We know newness to the market, or ‘freshness’ as we call it is a critical part of how people search for their next home so we now show you how long a property has been on the market and allow you to sort properties by age.

We’ve also integrated more and more local content to provide context to your property search. Most recently we’ve updated the London tube lines to add the London Overground, but we also added sports facilities and parking spaces in the last few months.

But those are just some of the new features you as a user can see. Rest assured we’ve been equally hard at work optimising the backend systems that you don’t see. Over the last few months we’ve been experimenting with systems that learn from individual user behaviour and adapt in real time. This is still in test mode, but we’re almost there. So don’t be surprised if the guy sitting next to you in the office starts getting Nestoria different search results than you.

Our thanks go out to everyone for the feedback we get on new features, but especially we’d like to thank those folks who volunteered to be betatesters.

So, what does 2008 hold? Well in the immediate future you can expect better integration of the metadata we recently made available via our API, and we’re always playing with data visualisation.

BTW – If you don’t like the way we’re displaying search results you have two choices: please let us know what we could do better and we’ll try to run a test with your suggested modification, or you can try your hand at our API and build your own interface. In the last month there have been many interesting uses of the API.

When a picture isn’t worth 1000 words

Published by Ahmed on July 17, 2007 in Nestoria Rank and about us. Comments

Here at Nestoria, we like to provide our users with quality information, in order to enable them to search for their dream house quickly and efficiently.

This means providing snippets of key information, and of course, lovely thumbnail (small) images of the properties in question. However, some sites provide us with images that sadly turn out to be far from lovely, and have little relevance for someone searching for a home to buy or rent.

Take for example these gems:


But we don’t mean to single these few out, there are actually quite a few bad apples. Here’s a little quilt of ‘image not available’ images. And these are just a tiny, tiny percentage of the ‘bad images’ we come across.

bad images

We’ve developed algorithms to weed out these less useful images, and promote those listings that do have relevant photographs.

This is just one way in which we aim to make the life of the property searcher easier. If you ever come across an unsavory image in our listings, please get in touch and we’ll make sure it doesn’t show up again.

How do we detect these sub-par images? One word: magic.

Enjoy your picture-filled property searching.

Economist coverage of vertical search

Published by Ed on July 15, 2007 in Nestoria Rank. Comments

For those that haven’t seen it, this week’s edition of The Economist has an article about vertical search. Though the article focuses mainly on the health care vertical in the US, it does a good job of summing up the advantages (and challenges) of vertical search for the layman. One quote in particular stands out:

In short, relevance is king, says Monique Levy of Jupiter which suggests that a vertical search-engine that successfully pairs a broad target market with a complicated topic can do well

We couldn’t agree more, which is exactly why we spend so much time working on relevancy. We’ve developed our own algorithm, called Nestoria Rank, which we use to show the best results for a user’s search for a home to buy or rent. We occasionally document our work on Nestoria Rank here on the blog. We’ll post another update about some of the new tweaks we’re currently working on in a few weeks.

It’s very encouraging to see someone else recognizing the importance of relevancy.

On a related note, long time Nestoria Blog readers may recall our first ever post in June of last year: “Vertical Search, What’s That?“

Nestoria Rank update – June 2007

Published by Ed on June 8, 2007 in Nestoria Rank. Comments

Nestoriaks!

Today we once again take a bit of time to review recent changes we’ve made to the core algorithms that make Nestoria the simplest way to find houses and flats in the UK. As long time readers will recall from our past updates we collectively term these algorithms “Nestoria Rank”.

First and foremost, we recently launched a big improvement to the technology we use to detect “duplicate” properties – when we get the same house from multiple sources. In the same way that we work hard to eliminate spam and expired listings, it is critical to correctly detect duplicates. Property searchers aren’t interested in seeing the same home multiple times. Unfortunately, it can be surprisingly hard for a computer to detect that different property descriptions from different sources are actually the same house. As an example, here’s a screenshot of a time in the past where we just weren’t getting it right:

duplication

While the prices are the same, the listings have different descriptions, different photos, come from different estate agents, and most critically, are geocoded slightly differently. Nestoria users will be pleased to know we’ve been able to solve all these issues, and are now much better at detecting duplicates.

Additionally, in the last few months we added the ability for property searchers to immediately find similar homes to ones you like. And we’ve done a lot of work on sorting and categorising properties so you find only highly relevant results.

Of course along with these subtle algorithmic changes we remain committed to improvement on all other aspects of what makes a great search engine:

  • In terms of usability we’ve had very positive feedback on our sliders. More importantly we’ve seen from our internal metrics that the average user is now finding relevant homes in fewer pages. We of course continue to test new usability improvements all the time.
  • We’ve increased the comprehensiveness of our database of property by adding new developments.
  • We’ve greatly improved our API so that anyone can play with our database of over 650,000 listings.

So, a few improvements. And of course we did all for both the UK and Spanish property search. Nevertheless, we still have a long way to go though, and you’ll be happy to know we’ve only scratched the surface of the ideas we have. We’ve got a few tricks up our sleeve for the coming weeks.

While our personal metric of success is how many people we help find a home as quickly and simply as possible, it was a great validation of our work when Nestoria was recently chosen by Google to serve as the Google Code case study for successful use of the Google maps API.

As always we look forward to your feedback – please let us know how you think we’re doing.

Nestoria Rank update – February 2007

Published by Ed on February 26, 2007 in Nestoria Rank, about us and search. Comments

Hello Nestoria fans

As promised, here’s a long overdue update on some of the work we’ve done over the last few months to improve the property search experience on Nestoria.

Strictly speaking, Nestoria Rank is the name we give to our relevancy algorithm, but there are many aspect of creating a compelling search experience, and we’ve been fiddling with several of them:

  • Comprehensiveness: our database now has over 450,000 high quality listings of properties to buy and rent all across the UK. What do we mean by high quality? Well, we send a lot of time trying to weed out listings that are already sold or “sold subject to contract” or listings that are outright spam. In total we actually have almost 600,000 listings, so you can see we’re very serious about weeding out dubious listings.

    Of course we’re always on the search for more listings. If you have access to property listings and want to get them in front of a highly relevant audience, we’d love to hear from you

  • Usability: We’ve spent some time tweaking our algorithms, and the site is now 30-40% faster on average. How do we know? Well, we plant little javascript beacons on some pages that allow us to measure how long it took for our pages to render.

    We haven’t played much with our interface the past few weeks, but we now have a large enough user base to begin testing minor tweaks on a small subset of users. So don’t be surprised if you start seeing some little changes (hopefully improvements) in the coming weeks and months.

  • Analysis systems: Our pure property search engine approach to the market is meeting a need for more and more internet users. According to comscore our number of unique UK users has grown by over 50% the last three months.

    Now, you might read that and say “Ed, what do I care how many people use Nestoria, I just want to find property fast.” and you’re right.

    Nevertheless, the more people we have using Nestoria the more chances we have to learn. So we’ve invested effort in building analysis tools to help us see when we’re not quite doing as well as we could be. This is also why we spend time on our API and our co-branded service. With the tools we have in place, the more users we have, the more we learn, the better we can make the property search experience.

  • and last, but not least, our dear friend Relevancy: we’ve continued to tweak the old Nestoria algorithm over the last few months. We now have over 10 different elements in the equation we use to determine which flats and houses are shown first. It’s still early days (we have many more ideas), but we’re already seeing via metrics like repeat usage and percentage of users clicking on first results that our emphasis on relevancy is being well received.

So, does all this mean Nestoria is now perfect? No, of course not, we still have a long way to go, we only launched to cover all of the UK five months ago.

But we wanted to let you know what we’ve been up to, and most of all let you know that the ground work has now been laid for some deeper and continual improvements.

But that’s just what we think, let us know how you think we’re doing.

Nestoria Rank update

Published by Ed on November 29, 2006 in Nestoria Rank, about us and technical. Comments

Friends, we know many of you have been anxiously awaiting our update on Nestoria Rank (our proprietary algorythm for determining relevance – first discussed here) and today your patience is rewarded.

The Nestoria team has been hard at work since our last update. We've added new ways to filter the property listings, new ways to use the map to narrow your property search and loads of new types of local information to qualify an area (photos, government data, more transport data, and council tax data).

All of that is great, but it doesn't really address the issue of relevance. As anyone who works in the search industry will atest, even measuring relevancy is a non-trivial challenge. When results are relevant, you don't even notice, you just have a great search experience. So we thought we'd let you know about some of the challenges we've been grappling with.

  • Comprehensiveness: We've significantly grown the size and geographic coverage of our database. Whether you're looking to buy a bungalow in Belfast, or a two bedrooms in the North Laines of Brighton, you can find it on Nestoria.

  • Natural Language Processing (NLP): Unfortunately much of the data we're dealing with is unformatted text manually created by humans. We love humans generally, but unfortunately humans make spelling errors, typos, are sloppy with cut and paste, etc. So we've developed some algorithms to parse human entered text and extract the true meaning that's relevant to the home searcher. For example, does a property have a garden or a balcony, or how many bathrooms the house has. You can expect a lot more progress on this front shortly.

  • Disambiguation of place names: Relevancy in property search has two main pieces: determining the correct location for a query, and then ordering the results in a relevant way. Disambiguation is a key part of the first part of relevancy. In the UK there are many place names that exist multiple times. For example there are several towns named "Newcastle". Building logic to know when to default to a certain location (in this case Newcastle upon Tyne and not Newcastle, Staffordshire) and when not to default (for example is someone searches for "Bangor") isn't simple.

  • Ranking based on quality of a listing: We've begun experimenting with a very involved quality score for each listing based on many factors like how well we were able to geocode the listing, age of the listing, quality of the description (as determined by NLP), and a variety of other factors. This is another area we'll be focusing on heavily in the coming weeks.

Of course we have a few other little secrets as well. We can't reveal everything here, but we can reveal that we remain hard at work on relevancy. As always, it's an ongoing challenge, and one we're excited about diving deeper and deeper into.

Have an opinion about how we're doing? We'd love to hear from you.

What is Nestoria Rank?

Published by Ed on September 25, 2006 in Nestoria Rank, New Features and about us. Comments

Loyal readers, as promised, I’d like to explain a bit more about our default property search algorithm: Nestoria Rank.

We know that the providing a quality search experience depends on several elements: relevancy, freshness, comprehensiveness, data quality, and usability.

Relevancy is a very tricky thing – even consistently measuring it is tough. But it’s one of the things we spend most of our time working on.

When we first launched Nestoria in June our search was fairly basic. Since then we’ve tested several different algorithms that depend on many different parameters. Obvious elements of the equation are things like distance from the center point of a region and quality of the listing (does it have a photo, how comprehensive the data is, etc), but there are many others. We’ve also incorporated several ideas and comments from our users (many thanks).

We measure things quite closely, and the tests have been very revealing. Over the weeks we tweaked the parameters, and we now feel proud enough to actually give our algorithm a name. And thus Nestoria Rank was born.

Like any newborn, Nestoria Rank still has a lot of growth potential. We think right now we’re at about version 0.1 . Improving relevancy is an ongoing process, and we have a long list of ideas we’ll be testing to improve the property search experience. We’ll keep refining in the coming weeks and months, please keep the feedback coming!

A picture of the team discussing the next step in Nestoria Rank development:

The team at work discussing Nestoria Rank