We’ve had a few queries recently about how search works and in particular how Folksy is indexed by Google and how Folksy on-site search works and what improvements are planned. Whilst both are about items being found by potential buyers, they are two separate questions which I’ll tackle in turn and suggest ways you can help to improve your chances of being found in search results both externally and through Folksy.

(image by moneyblognewz)

External Search

How does Google index items for sale on Folksy?

Overview

Google constantly crawls the web, revisiting pages already in its index, and crawling new pages if finds along the way.

In terms of a Folksy item page, that means Google will usually first find the existence of the item through a link on another page. Some of those links may be from other Folksy pages, for example:

• Category pages
• Shops pages
• The homepage
• Other item pages
• Blog posts
• etc.

Improving your item’s chances of doing well in Google search

Bearing all of this in mind, how can you maximise your chances of your items being seen in search?

We’ve had a couple of questions about discrepancies and changes recently too, so we thought we’d try to clear those up.

Question 1: Are view counts artificially enabled by Google spiders?

The discrepancy some people have noted between the view counts we used to provide on the admin screens and the metrics Google provide through Google Analytics which many people have integrated, is probably explained by the fact that the count we used to provide was probably influenced by ‘spiders’ or robots used by the search engines when indexing the web. Each time a spider / robot indexes a page it is generally counted as a page impression.

Whereas Google Analytics only measures views in a browser where javascript is enabled so is a more accurate measure of actual people looking at a web page.

Question 2: Has the re-design affected results from Google Image Search?

On researching this we found that it is explained by Google changing the way Google image search worked. They did this in mid-July last year in the UK, and this meant almost all image search traffic was re-categorised from being ‘referral’ traffic to being treated as search traffic. Here’s the graph showing where the ‘referral’ traffic appears to drop:

The search traffic wasn’t lost, it seems most of it was re-classified as organic search traffic.

Changes to the naming convention of images on Folksy (we append image file names with the type of image they are e.g. main etc) have had no bearing on how they are indexed or returned in results. Interestingly, people coming to Folksy via image search convert into sales far less than ‘normal’ web search (which is explained by most people looking for an image to use rather than looking to ‘shop’).

There’s more info on the Google Images change over on Search Engine Roundtable

Internal Search

How does internal search work on Folksy?

For items, the current search technology we use creates a searchable string by joining together a number of different fields such as title, item description, category name. We then have an algorithm that gives these fields different “weights” which equates to a higher score when the search string is queried. If someone searches using multiple words then only those items that match all of those words are returned, so if you search for “pink floral skirt“, only items matching all of those words will be returned (33 as of now). These words could appear separately anywhere in the listing.

If you use speech marks to define a term, for example “pink cushion” you will return all items that have that exact term (in this instance 8). If you search for “pink” and “cushion” without the speech marks it returns many more (799 items as of now) as search is returning all mentions of cushion and pink. Ideally we would like the results to return all those results with the term “pink cushion” first, this is called n-grams for approximate matching and is something we will be working on in the next phase of development on search (see below).

Shop search works in a similar way to item search, using the fields shop title and also shop description and username where we append an appropriate weight to the different fields in the string we create.

The advice on how to maximise the likelihood of your items being seen in search is to describe your item really well, both in the title and description and inspiration fields but also using facets such as colour. The fuller and more accurate the description, the better.

Next stage developments

Search is quite a complex area and one which is constantly being looked at by any service with a large amount of content.  There are three areas of search we are enhancing:

1. Relevancy – creating a more advanced algorithm to return more relevant results (including n-gramming – basically term matching – and stemming – a process for reducing words to their root form e.g. plural to singular etc.)
2. Filtering – providing functionality to allow people to filter search results by price and also by category
3. UI – improving the search field to incorporate auto-suggest based on popular search terms, best suggestions in the search results as well as grid view and list view

This work falls into our workstack after Guest Login and Checkout improvements have been delivered. The estimated start date for the search work is March 1st.