Over the years, we have noticed that many clients come with wrong expectations regarding matching rates. That’s why we have decided to demystify the topic – hopefully this will help many companies starting their competitor price / assortment monitoring projects.
First of all, lets define product matching rate more precisely. Let’s suppose you want to check A products on B different websites. And let’s suppose the product matching results in M matches found. So, your matching rate in that case will be MATCH_RATE = M / (A x B). So, if A = 1000, B = 10, and M = 2000, your match rate will be 2000 / (1000 x 10) = 20%
Or, for the ones who do not like math that much – here’s a more graphical illustration. Let’s suppose that each product is in it’s own row, while checked websites are in columns. In that case – in the following image X represents a combination where a match has been found. Empty cells represent no-matches.
As you can see, the above matrix is pretty sparsely populated, and actually it;s matching rate just above 10%.
Reasons for poor matching rate
When discussing possible reasons for matching rate below your expectations, we should keep in mind the following
- Maybe your expectations were too high?
- Maybe you’re dealing with own brand / white-label products which are not expected to be sold elsewhere
- Or, maybe your matching provider did a poor job?
Lets discuss each of these options in more details
Setting the right expectations
After so many years in product matching business, we have noticed that many clients come with wrong assumptions – on how many matches are to be found. This question actually translates to ‘what percentage of product assortment do my competitors share with me’. A typical online retailer would think ‘We all deal with same brands, we all get same product lists from our vendors => our assortment should be nearly identical => product matching rate should be real high’.
The reality is just the opposite: listing new products can be quite a labor-intensive work, because online retailers try to insert unique content – with each product getting it’s own human-curated product description, and sometimes with even product images customized. The reason for such labor-intensive process is the wish of each retailer to have unique pice of content – in hope that this will be appreciated by Google’s search ranking algorithm. The alternative approach (inserting products in bulk, all with descriptions / images received from the vendor, may result in SEO penalization, which all eCommerce sites are trying to avoid).
In reality, listing new products very much depends on how much workforce (which is usually coming from low-pay countries) can a retailer afford. Further, lets not forget that different retailers prioritize product listing differently (depending on which brands / categories they consider important).
Own brand / white label product matching
If you’re dealing with products which only you carry – how can you expect them to be listed by competitors? In other words, product matching rate in such cases should be 0%.
On the other hand, in such case it’s much more reasonable to go for similar product matching, with setting a threshold for accepted similarity. And, if you do go for matching similar products, you should expect real high (over 80%) product matching rate – because most of your products do have similar counterparts with competitors, don’t they?
Could it be that your product matching provider did a poor job?
There are many ways how product matching can be done. Maybe your provider went for automated this process which relied on identifiers that could not be found? Or could it be that they did it manually, but did not have enough resources to do the job properly? That’s why we always advise (and offer to perform) QA on product matching process.
Would you like PM>AI to perform this QA for you? Please get in touch, we will be happy to do this QA exercise for free, in a nicely documented way.
No-matches have a meaning of their own
Lets suppose that your product matching matrix has really few X’s (very few established matches) – like in the image above.
Please be aware that such a situation has a meaning of it’s own, which can be commercially exploited: your competitors have very few products that you carry – this means you have more liberty to raise your product’s prices!