View Full Version : 'ShopWindow Search Terms' - your feedback appreciated :)
Hi All,
In an attempt to improve the results from a SW search, we have a system in place that will 'map' any search term to any one category. Just to clarify, this means that the mapped search term will only return results from the chosen category. (This can be for any type of category in the tree: Parent, child etc..)
The ShopWindow team would really appreciate your input as to which simple, common words return poor results in SW. I realise that there maybe many instances of this happening but just as a starting point, if there are any which you feel need urgent amendment so that we can proceed reviewing.
Any feedback is much appreciated on this,
Kind Regards,
Adrian
SW Integration & Technical Support
adrian.botelho@digitalwindow.com
www.digitalwindow.com
For me my main bugbear is the precise nature required for model numbers on items.
eg.
All cat searches for:
KDL-40W2000U gives 2 results
KDL-40W-2000U gives 0 results
KDL40W2000U gives 5 results
40W2000U gives 2 results
40W 2000U gives 0 results
etc
Results are also in placed different categories.
Possibly not the best/worst example but hopefully that illustrates the point.
Thanks
Confuscius
04-01-08, 14:25
My view is simple - a very bad idea!
A search is a search and should run with parameters set by the user ans as far as I can see if you search within a category then you get results from that category.
More fundamentally, if you are aware of simple, common words that return 'poor' results then ask yourself what is it in the data or the search algorithm that causes it to return 'poor' results.
Even more fundamentally what you may think is poor then I may think is great as this is a matter of subjective opinion... try a search for the word 'red' then decide how you want it sorted and tell me which is the 'poor' set of results - the fact is that a search returns values according to what is specified.
Although poor is a subjective matter then some merchant feeds I would describe as 'poor' because I have a set of objective criteria against which I can assess the feed and therefore if you want to talk about poor search results then how about specifying the objective criteria against which you are judging the results. A lot of the time then the perception of poor results is often the result of a poor query - to illustrate now try a search for 'red shoes' - are those poor results?
Compare:
http://www.shopwindow.com/search.php?q=red&c=0&search.x=0&search.y=0
with:
http://www.shopwindow.com/search.php?q=red+shoes&c=0&search.x=0&search.y=0
Paul
Confuscius
04-01-08, 14:27
For me my main bugbear is the precise nature required for model numbers on items.
eg.
All cat searches for:
KDL-40W2000U gives 2 results
KDL-40W-2000U gives 0 results
KDL40W2000U gives 5 results
40W2000U gives 2 results
40W 2000U gives 0 results
etc
Results are also in placed different categories.
Possibly not the best/worst example but hopefully that illustrates the point.
Thanks
Crap in = Crap Out!
Are the above numbers in line with the data - if so then search works perfectly. Why don't DigiWin spend more time addressing the quality of the data and images collected? That would have the biggest impact on sales, IMO.
Hi Paul,
I can fully understand your point on this one but that's not entirely what i was getting at. For example, if a user searched on the term 'ipod' the first set of results you would expect to see would be the actual mp3 players themselves and not 'ipod accessories' for the first few pages. If an accessory is required, a more detailed search should be executed (E.g. 'ipod cover'). Broader, more generic terms will be left as they are.
Just to clarify, not every search query raised here will be mapped, we will review these and only use ones that we feel would positively affect user results.
Kind Regards,
Adrian
SW Integration & Technical Support
adrian.botelho@digitalwindow.com
www.digitalwindow.com
Confuscius
04-01-08, 18:55
Hi Paul,
I can fully understand your point on this one but that's not entirely what i was getting at. For example, if a user searched on the term 'ipod' the first set of results you would expect to see would be the actual mp3 players themselves and not 'ipod accessories' for the first few pages. If an accessory is required, a more detailed search should be executed (E.g. 'ipod cover'). Broader, more generic terms will be left as they are.
Just to clarify, not every search query raised here will be mapped, we will review these and only use ones that we feel would positively affect user results.
Kind Regards,
Adrian
SW Integration & Technical Support
adrian.botelho@digitalwindow.com
www.digitalwindow.com
Actually, I would expect to see all Ipod search results sorted by relevance because that is how it is currently set up.
So, the category MP3 Players conatains ipods and ipod accessories but the accessories does not contain ipod accessories - so why don't we have a MP3 Player Accessories category? Of course you can put either ipods or their accessories into either / eac / both depending on what you merchant name is.
You picked a great example!
Look at:
http://www.shopwindow.com/category.php?c=35
and compare it to:
http://www.shopwindow.com/category.php?c=40
and then ask yourself is there much point in having a category structure if you do not control the data?
If you want to create some special searches that form the basis of a list of clickable texts links then that provides away to get at 'ipod', in fact you can have some fun fetching a product where you know there are multiple prices and then ordering the results from high to low to give a price comparison.
I have to go now but I will build an ipod link later that works and one that will not work but should!
Have you thought about introducing product tagging as an alternative means and then providing a tag search? Now that I would vote for.
Paul
PS then using the first product leads I have a question for you - what is a 'Sony CPFNW001' and what categories are associted with it?
PPS Answer here: http://www.shopwindow.com/search.php?q=Sony+CPFNW001&c=0&search.x=10&search.y=12 - now I do not know what it is because ther is no description but I can get a cradle that may be a computer component - just an example of using search and people wonder why conversion is so awful - deal with the basics first, please?
Confuscius
04-01-08, 21:46
I thought it would be interesting to see what happens as a user as if I had just landed on http://www.shopwindow.com/ - hmm, get that festive feeling - must be talking about Xmas 2008?
Hmm some nice gifts - lets try this http://www.shopwindow.com/search.php?q=apple+ipod+touch&c=0&search.x=0&search.y=0 - 126 results looks promising. Now I can see 16gb and 8gb and it is sorted by Relevance, whatever that means. So lets re sort from high to low price - http://www.shopwindow.com/search.php?c=0&q=apple+ipod+touch&iListLimit=10&sListSort=hi - that's better. I can see that the 16gb is available from lots of shops - Tribal UK, Bennetts Electrical and Saverstore clearly cannot afford to pay someone to write a product description. Most of the rest of the descriptions are a bit flakey.
So what is a 16gb Ipod - hmm - 5 Merchants say it is a Mp3 player, 2 for Portable Audio, 1 for Audio Equipment, 1 for Gifts, Gadgets and Toys and, my personal favourite, 1 for an Ice Cream Maker - can't wait to Google for "how to make ice cream with your ipod"!
So now time for refinement, I want a 16gb version - cahnge my search to : http://www.shopwindow.com/search.php?q=apple+ipod+touch+16gb&c=0 - now got 23 results and I see some new Merchants and some new categories - so I have to search globally as an Ipod is also now a Computer Component according to Novatech. Also it looks like Eshopmegastore are going to give me a 8gb and a 16gb for just £199 - now that is a bit misleading or a great deal.
So now lets go for the low to high price : http://www.shopwindow.com/search.php?c=0&q=apple+ipod+touch+16gb&iListLimit=10&sListSort=lo - hmm, I can get one free! Just need to spend a few hundred quid on a Mobile Phone contract that I do not need but it is nice to know what my options are.
So it looks like £199 is the low price - try to buy it - oh no, the 16GB version is £60 more on the dropdown and I can only pre order - that Buy now button is a bit misleading. Oh well lets try again, page2.
Empire Stores - £229.99 - it's acamera category now! - page loads and defaults to 16gb version - price is £350 ish - misleading again. Next try:
Dixons - £235 - great price but..... OUT OF STOCK - quote from Dixons "Apologies, the product that you have selected is currently unavailable. You now have two options. Please select an alternative product, or revisit tomorrow as we receive frequent stock deliveries. Apple 16GB iPod Touch Please use the back button to go to the previous page" - back we go!
Play - £249.99 and free delivery and a nice bit of discount on a docking station - not bad and it is in stock. Interesting thing to note is that I would love to know how much commission each Merchant pays on this product! Shame about all those html tags in the Play descriptions - oh well, a minor irritation.
Let's continue £269 seems to be mainline price - quite a few more options if you are not motivated by price. A quick scout through to see what we find. Novatech - in stock. Woolworths - can buy but not delivered for 2 weeks! John Lewis - nice description. AdvancedMP3players will beat that Dixons price but I wonder if they will notice that it is out of stock! Marks and Spencers - OUT OF STOCK! Saverstore and Bennetts - no description and Tribal top the tree at £312.08 including VAT and delivery.
Product names - did not realise there were so many ways to name the same product - Apple Ipod Touch MP3 Player 16GB Black - Brand Product Size Colour - so how many do it like this? Guess!!!
So - just count how many things need to be corrected to return a half decent page for a search for one product - anyone want to convert each problem into anew thread for each problem? Anyone else fancy doing another example?
So it looks like the original 3 most relevant were actually the most IRRELEVANT - nice one that!
Seriously, trying to make common searches more meaningful is just about POINTLESS until you sort out the basics. Anyone for ice cream?
Paul
Paul,
Fair play - I think we understand your point of view on this now.
We do not discredit the fact that our data has to be improved, far from it infact. Only today I was on lunch with a major brand partner of ours and openly discussed the current data issues and how we are aiming to move forward with them.
You must bare in mind that this is not just something that we experience. This is an issue for any network taking data from a third party.
However, data aside, we have developed a tool that has been integrated by over 2000 affiliates, and a number of high profile brand partners; so we must be doing something right.
I personally always appreciate your input on this forum, however lets not discredit all the time and effort that has gone into improving ShopWindow over the past 6 months and making it what it is.
Regarding the quality of data, well.... ive said it quite a few times on here - im sure i dont have to share my thoughts and aims again.
Amo
Confuscius
05-01-08, 00:50
Hi Amo
My concern is and has always been how to make the best of a visitor - getting them to use the search box would be brilliant but my monitoring of my logs show that the vast bulk of visitors are 'clickers' and not 'searchers'.
Good to hear that data quality is an issue and I sincerely hope we see improvements in the future - it would be interesting to set up a couple of affiliate sites and try to split the feeds between better and poorer just to see what difference it makes to click conversion rates - mine is about 3% who hit the goto end up purchasing. What fascinates me is not why 3% are buyers but what causes 97% not to be!
Some feeds are really good, some are woeful but not providing say any descriptions does a disservice to Shop Window and I simply cannot understand why you tolerate this from merchants. I agree completely that for you to have developed a tool that has been integrated by many thousands of affiliates and high profile brand partners is a great testament to the fact that you are doing a lot more than something right.
My intention has never been to discredit all the time and effort that has gone into improving ShopWindow over the past 6 months and making it what it is. It is a great tool and the ££££'s in my affiliate account are a testament to the fact that it works. In the past month I have generated at least one sales for well over 100 different merchants which clearly demonstrates that it works in its broadest form. Based on my recent stats and the fact that Google is starting to like my sites then I would hope to generate sales turnover in excess of £1 million per annum by this time next year based on 'free' traffic, no membership and no big brand.
If you can improve the search function then that would be great but to start to presume what the user wants is a dangerous game to play.
Looking forward to more improvements and more income!
Best wishes,
Paul
Just wanted to thank Paul for those excellent pieces - nice work and very well put. You've saved me struggling to put into words those very same thoughts, and they wouldn't have been as eloquent as yours.
Amo - appreciate all the effort everyone is putting into this. Hopefully this 'great' tool will soon become a 'superb' tool.
IntroSites
05-01-08, 03:31
If anyone is interested in logging searches so you can see what terms give bad results, i have posted some code here http://www.shopwindowforum.com/showthread.php?t=461
Hi Intro,
Just wanted to say well done for the coding, looks like a great tool for capturing some very useful data.
This is excellent stuff, hope many will find it useful :D
Adrian
SW Integration & Technical Support
adrian.botelho@digitalwindow.com
www.digitalwindow.com