AuctionInsights Exclusive

Best Match

Download the FREE eBook Best Match Made Simple

eBay’s Best Match search algorithm has sellers up in arms because, in many cases, their auctions never make it to the first page of eBay’s search results. There’s been speculation on what the eBay Best Match algorithm considers when generating the order in which search results are listed. Forbes Magazinespeculates that search results are predominately a factor of seller feedback and powerseller status (effectively squeezing out the small eBay sellers). AuctionBytes has suggested that eBay’s Best Match algorithm might be a function of many things including: powerseller status, shipping options, feedback ratings and Detailed Seller Ratings, and use of item specifics in the description and the titles. A number of sellers on the forums have speculated even more conspiratorial theories of the factors they think are influencing the ranking algorithm. I’ve always been a fan of facts vice speculation and guesses, so I hit Google looking for some answers. After a few hours of research, I managed to track down eBay’s Best Match patent application.


The bulk of the patent application describes the relationship between search terms, the text in the auction title and how relevance and desirability are determined. The Best Match patent does describe other factors that can influence search result rankings including (but not limited to):

  • pictures available
  • seller ratings
  • price range
  • geographic proximity of searcher to seller
  • shipping prices
  • time left in an auction
  • number of bids
  • Buy it Now

Update - For a less technical explanation of Best Match, view Best Match Made Simple

The Best Match patent application describes how these factors can be included and weighted in the basic algorithm to influence the search results. Indeed, eBay has already announced that sellers with low Detailed Seller Ratings for shipping and handling would no longer rank highly in their search results — so we know this a factor that influences the algorithm.

But the base algorithm involves strictly auction titles and this is was we need to have a fundamental grasp of as we decode the eBay Best Match search algorithm.

The patent application describes the eBay Best Match algorithm as “a system and method to sort search results based upon [a relevance score and] a desirability value.” It goes on to say…

This desirability value may be based upon the difference between a demand value and a supply value. Demand may be based upon user activity such as click-through, purchases, price, or location. Supply may be based upon a supply of keywords that may be the number of times a word is used in search or item title. The system and method may include receiving a search query, associating a first numerical value with a keyword that is a part of the search query, tracking user activity associated with the keyword, associating a second numerical value with the keyword based upon the user activity, finding a difference value between the first and second numerical values, associating this difference value with the keyword, sorting keywords based upon the difference values, and returning the search results of the sorting.

Huh?

Please allow me to attempt an explanation…

Determining Relevancy Score

When a visitor to eBay inputs one or more keywords into the search box, the search algorithm returns the auction listings with keywords in the title that match those included in the user’s query. The eBay Best Match algorithm now assigns a point value to each and every word in the title text string displayed to the user. Initially this point value is zero, but this will change based on the actions of the searcher.

Good Keywords Get Rewarded

If the user clicks on an auction title in order to view a particular listing, each of the words in that auction title are awarded a point. These points are used to evaluate the relevance of auctions in future searches. If the user takes further actions such as adding to watch lists, placing a bid, or actually purchasing the item up for sale, those words are awarded additional relevance points.

For example, if a visitor to eBay.com searched for “iPod nano,” the default search results would return all auctions with the words “iPod nano” in the listing title. If the user then clicked on an auction with the title “Brand New iPod nano 4 GB Black MP3 Player Sealed,” each of the words in the title would be awarded a relevancy point. These points would accumulate over time and keywords that accrue more points would be considered more relevant in future eBay Best Match searches for the query “iPod nano” and help determine the sort order when the results are presented on the results page.

Bad Keywords Get Punished

The converse is also true. The eBay Best Match algorithm identifies unique key words in auction titles that don’t get clicked and assigns a negative point value to those words. It will only assign a penalty to those words that don’t match keywords in the auctions that do get clicked.

For example, if when our eBay user searched for “iPod nano” and clicks as described above, but does not click on the auctions entitled “New iPod nano leather case” or “Used iPod nano USB charger,” the words “leather,” “case,” “used,” “USB” and “charger” would receive negative points towards their relevancy score because the searcher did not click, add to watch list, bid or make a purchase auctions with those words in their listing title.

Keyword Relevance Scores are Determined over Time by User Actions

Over time, keywords develop a relevance value by accumulating positive and negative points. The more visitors to eBay that click on auction titles, place bids, add to watch lists, and ultimately buy, the more positive points are awarded to keywords in those auction titles and these words are considered more relevant to future searchers.

When eBay visitors fail to click auction listings with certain keywords in their titles, those keywords are considered less relevant and points are deducted from their relevancy score.

Keywords can have a negative score. If, over time, users who search for the “iPod nano” just are not interested in auction titles with the word “used” them, and thus hardly ever click on titles with that keyword, “used” will accumulate a significant negative relevancy score for that search query.

Calculating an Auction Listing’s Relevance Value

Although this is a simplification of the eBay Best Match algorithm, part of an auction listing’s relevance value is the sum its title’s keywords scores for that particular search query. Let’s say for the search query “iPod nano,” the keywords below have accumulated the following relevance scores:

  • iPod (1.39429)
  • nano (1.10132)
  • MP3 (1.11403)
  • New (1.08399)
  • Sealed (0.553271)
  • Case (-0.107664)
  • Used (-0.117723)
  • Charger (-0.173231)
  • 4GB (0.530753)
  • Player (0.86456)
  • USB (0.00000)

Note: These are hypothetical scores

Hypothetical Examples

Therefore the Best Match relevance value for an auction with the title, “New iPod nano MP3 player ~ sealed” would be 5.010141(1.08399 [new] + 1.39429 [iPod] + 1.11403 [MP3] + 0.86456 [player] + 0.553271 [sealed] = 5.010141).

A relevance value for an auction with the title “Used iPod nano USB charger” would be 2.204656(-0.117723 [used] + 1.39429 [iPod] + 1.10132 [nano] + 0 [USB] – 0.173231 [charger] = 2.204656).

Interestingly, the relevance value for an auction with the title “4GB iPod nano with charger and case” would only be 1.351178(0.530753 [4GB] + 1.39429 [iPod] + 1.10132 [nano] – 0.173231 [charger] – 0.107664 [case] = 1.351178). Even though this is probably a better value than the first example (because it includes some accessories), this auction has a lower relevance score because it includes the words “charger” and “case” in the listing title. In our hypothetical example, these words have a negative relevance score for the search query “iPod nano,” because bidders who use this term have been more interested in the actual MP3 players than auctions for just accessories.

Determining Desirability

Another factor influencing eBay’s Best Match search results is something termed “desirability.” This is a relatively basic function of supply and demand. If most searchers using the query “iPod nano” are actually looking for the MP3 player vice accessories then demand for auctions selling MP3 players is high. If most of the listings with the keywords “iPod nano” in the title are actually auctions for accessories, then the supply of iPod nano MP3 player auctions is low.

Supply and Demand

Thus auctions for items in which demand is high and supply is low are considered “desirable” and score highly for desirability. Auction listings with a high relevance score and a high desirability score are going to rank higher in the Best Match search results than auctions with a lower combined score.

Auction where supply is high and demand is low are going to score poorly for desirability.

How Supply is Calculated

According to eBay’s Best Match patent application, “the total occurrence of every unique keyword in all the titles is first determined. Then the percentage occurrence of each keyword in the result set [(the auction titles)] is determined. For example, if there were 1000 items in the results [for the search query ‘iPod nano’], and the word ‘charger’ occurs in 900 items, then the supply percent of charger is 90%. If the keyword ‘player’ occurs in 50 items, then its supply percentage is 5%.”

How Demand is Calculated

“The percentage that each keyword is used in an activity data set, [‘demand’], is calculated in the following manner. Activity data is collected on a daily basis, data that includes activities such as clicked, bid, bought, or added to watch list… are associated with keywords used in item titles. For example, if for the query ‘iPod nano’ 1000 click-throughs are collected and we know the item title of each one, then the occurrence of every unique keyword in this set of titles is calculated. The percentage occurrence of every keyword in the same set is calculated. Thus, if the keyword ‘charger’ occurred in 100 of these 1000 titles, the percentage of charger in the ‘demand’ is 10%. Similarly, if ‘player’ occurs in 950 titles, its demand percentage is 95%.”

Calculating Desirability

The formula provided in the Best Match patent application for calculating “desirability” is as follows:

Desirability = log2(1.0+(dw*d))-log2(1.0+(sw*s))

Where:

  • d= demand
  • dw = demand weight
  • s = supply
  • sw = supply weight

The patent application notes that demand weight “is set to 3 by default,” while supply weight “is set to 1.” This means that demand is three times as important than supply when calculating desirability in the eBay Best Match algorithm.

Putting it All Together

The patent application is particularly vague in how the Relevance Score and Desirability Score are combined to produce a value that is used to determine the ranking in the Best Match search results. Obviously, a higher combined relevance and desirability score will rank higher in the search results than lower scores.

Additional Factors Affecting Best Match Search Result Ranking

Recall that the patent also mentions other factors that can influence the final Best Match search results placement. The factors specifically mentioned include:

  • Price range
  • Geographic proximity of seller to searcher (as determined by the seller’s address on record and the IP address of the searcher’s computer)
  • Time until auction close
  • Number of Bids
  • Seller Feedback
  • Pictures
  • Buy it Now
  • Current price compared to average sale price (is it a bargain?)

The patent specifically mentions that these factors could be weighted differently or not at all depending on corporate decisions and that additional factors could be included. We already know that sellers with low scores in their Detailed Seller Ratings for “Shipping and Handling Charges” are receiving, according to eBay “considerably reduced visibility.” This means that their auctions are buried in the search results. Therefore at least one aspect of the Detailed Seller Ratings is a factor (with substantial weight) in the eBay Best Match search algorithm.

Optimizing Your Auction Listings for eBay’s Best Match

Just as webmasters optimize their web pages for Google, eBay sellers can optimize their auction listings for Best Match search results. Your objective here is to be at the top of the first page of the search results.

The area in which the seller can exert the most influence on their Best Match ranking is the title. The title has always been important, but now it’s even more critical.

You’ll need to do a little investigating to determine what words to include in your title, and just as importantly – what words to exclude.

Search for an item on eBay, but put yourself in the buyer’s shoes when you think of the term that you are going to enter into the search box.

Next, sort by Best Match if the search results are not already sorted in that manner.

Take a note of the words used in the titles of the auction listings returned on the first page of the Best Match search results.

Now go to the middle of the Best Match search results and note what words are in the titles of the auctions listed. Does anything stand out?

Finally go to the last page of the Best Match search results and note the words included in the auction titles there? Does anything stand out?

Determine the Good Keywords

Are there any words that are more prevalent in the titles near the top of the search results that are less prevalent in the middle and the bottom of the search results? If so these are good keywords that have a high positive relevance score.

Determine the Bad Keywords

Are there any words that are more prevalent in the titles near the middle and bottom of the search results that are less common in the titles of the auctions that made it to the top of the list? These are bad keywords that have a low or negative relevance score.

Include the Good and Avoid the Bad

You want to avoid including keywords you suspect have a low or negative relevance score and include those keywords that you think have a high relevance score.

Important– Only include keywords that apply to your auction listing. Including keywords that are irrelevant or misleading just because you suspect they have a high relevance score is considered keyword SPAM. eBay is quick to cancel auctions that SPAM their auction titles and competing sellers are quick to report offenders. As eBay’s Best Match search algoritym matures, auctions with titles like “like NEW MP3 PLAYER better than IPOD NANO” are apt to be cancelled or penalized in some manner as to make them irrelevant.

Other Factors You Might Influence

As you explore sample auctions that rank well or poorly in eBay’s Best Match sort results, take note of other things that might be influencing the results of the algorithm.

  • Are there pictures? How many? Are they hosted by eBay’s picture services or some external host?
  • Is there a Gallery image? Any other features?
  • How long until the auction ends?
  • What are the shipping costs?
  • Etc.

You might just notice patterns that will tip you off to some of the different factors that are currently influencing the Best Match search algorithm. Some of them you may be able to influence and some of them you might not. Concentrate on the affecting the things that you can change.

Experiment. Build on your successes and learn from your mistakes.

Star Wars Comic Case Study

Briefly examining the Best Match search results for the query “Star Wars Comics,” a couple of things jump out at me. The keyword “lot” (ie more than one comic) is obviously a term with a high relevance score because the majority of the titles on the first page of the search results include the term. Jump back to page five of the search results, we see many of the same comics, closing during the same timeframe that have the keyword “pack” or “pak.” They are offering the same deal (a bundle of comics), but obviously the algorithm considers the keyword “lot” more relevant than “pack.”

Lesson:Sell comics in bundles and use the term “lot” in the title. Avoid the term “pack.”

Additionally, the time until the auction closes is also a factor influencing the listings search rankings. The first 4 pages of search results are pretty much limited to auctions closing in the next 48 hours. The next 4 pages include auctions that are closing in 72 hours.

Lesson: Consider the 3-day auction span to keep my auction near the top of the listings throughout.

All but two of the listings on the first page had shipping prices specified. As you went deeper into the search results, a greater percentage of the results listed the shipping costs as “not specified.”

Lesson: Specify the shipping and handling rate in the form when you are listing an auction. Simply including the shipping charge in the item description appears to be insufficient under Best Match.

The Bottom Line

Just remember, the Best Match search algorithm is about making eBay money. They want to feature the auctions that are most likely to result in a click, in a bid, and ultimately a purchase. So they want to put the auctions in front of the potential bidder that are most likely going to result in those actions.

Your objective now is to make sure that eBay puts your auctions at the top of their Best Match search results – so you, along with eBay, get the click, the bid, and the sale.

Good Luck!

Get your eBay listings to appear in the Google search results and watch your sales soar!

Download your copy of Beyond Best Match today!

Posted Jan 18, 2008

Related Posts: