Product matching

Automatch – continuous automated product matching

In case your industry uses highly standardized product identifiers like EANs, MPNs, UPCs, or ASINs, and if you need more than 1000 products matched this way, Automatch is the way to go.

What is Automatch?

Automatch is a continual automated product matching process that can be performed at specified intervals (daily, weekly, monthly), depending on your needs.

Depending on the match approval procedure (which is configurable to be fully automated, manual, or hybrid) products that get matched end up in your Price2Spy account, where they get regularly monitored for price changes.

How does Automatch work?

Automatch works in the following way: all existing products in the client’s account that still don’t have matching URLs on particular competitor sites are attempted to be matched – automatically.

In other words (for our clients who are already familiar with the Price Matrix report): Automatch will automatically try to fill in blank cells in the Price Matrix.

The matching is performed according to a set of matching rules, which we define together with our clients. The rules take the following into account:

• • product data searchable on a particular competitor site (for example; some sites may provide EAN/UPC search, others may not);
  • variations in product naming (example: ‘Galaxy S5 Silver’ vs ‘Galaxy S5 Gray’;
  • matching accuracy (determined by several factors, for example, linguistic proximity);
  • the need for a manual check on matching accuracy: the matches can be considered safe enough to be automatically approved, or they would need manual approval.

Why Automatch?

• Executed continually
• • With manual product matching, we have learned from our clients’ experience that they quite often face a situation that particular products are not found on a competitor site at the time the initial product matching was performed. However, after some time, a competitor does include it in its offer, but Price2Spy does not know about it – and hence it does not monitor its price;
• • On the other hand, Automatch will keep trying until it finds the product on the competitor site.

• Matching efficiency/costs
• • The costs of manual product matching tend to get too high once we speak of 10 000 products or more;
  • Furthermore, the time needed to perform the matching is often too long (weeks, if not months);
  • This is where Automatch kicks in, within a few days you’ll have up-to-date matching data. Additionally, cost-effectiveness gets better and better as the number of products to be matched goes up.

Please check more details on our Automatch page.

Manual product matching

Over the years we have learned that time is the most valuable resource in the world of eCommerce. Hence we have noticed that many of our clients are short on time when it comes to matching and entering competitors’ products / URLs into Price2Spy.

Our manual product matching team can help in such situations. What we’d need from you is a list of products you would like added to Price2Spy, followed by a list of your competitors.

Our task will be setting up all these products in your Price2Spy account, so you get notified of any price changes detected from that moment on.

Each of your products will be checked on your competitors’ sites, and manually matched with yours. The result is a human-made, nonsense-free, high quality list of matches, whose reliability is much higher than what you get from price comparison engines (like Google Shopping and others). When performing product matching, our team will check not only product names, but also technical specs and product images – in order to ensure that a match is indeed a proper one.

When setting your account up, we will perform first  500 matching combinations for free. As for the rest of products, please contact us for a quote. (Since our product matching team is located in Eastern Europe, we are able to offer very competitive prices for this service).

One important thing to keep in mind: although our data entry team is located in Serbia, which enables us to offer lower manual labour costs, we insist on the quality of product matching we do for our clients. This part of Price2Spy team has been steadily growing over time, thus accumulating significant matching experience across various industries, across many different markets. Further, quality Assurance (QA) is an integral part of this process, performed by dedicated QA team members.

Our product matching team has the ability to perform matching in any language – so far we had tasks not only in English, but languages like German, Spanish, Arabic, Swedish, Japanese, Portuguese, Russian etc etc

As for the industries – over 9 years we have performed task for more than 25 different industries – the most recent one being rental properties (where matching is especially difficult because it involved matching of apartments / rental properties – which is by no means an easy task).

Hybrid Automatch via Machine Learning

If your products are of such nature that they cannot be effectively matched by unique identifiers (no EANs, UPCs, MPNs, etc), and there is a need for manual matching as a continual process – you should consider Hybrid Automatch via Machine Learning – part of Price2Spy Enterprise package.

Hybrid means that machines and humans are working together. Put it simply: Machine Learning (ML) algorithm will go through products from both your and competitor sites and make sure that only meaningful matching candidates get suggested. However, it will be humans who will be doing the final check – whether that candidate is really a good match or not.

The result is a process where matching accuracy is very close to 100% – ensured by the human effort. However, the amount of such human work is much much less than if product matching was done with no ML help. Further, as more and more matches get identified, ML algorithms will get smarter and smarter (and less and less human effort will be required). That’s what learning is all about.

Product matching via ML was a project which took us 18 months of development in cutting-edge technologies. The problem is very complex, no other service (and there are many of them operating worldwide) offers such a feature. That is why we have prepared a series of blog posts explaining in detail why we went for such a project, and how we pulled it off.

• Product matching via Machine Learning – Abstract
• (Part #1) Product m via Machine Learning – Introduction to the project
• (Part #2) Product matching via Machine Learning – Important decisions to be made
• (Part #3) For ML experts – why is product matching so difficult
• (Part #4) Preparing the ML training set
• (Part #5) ML training Implementation
• (Part #6) Evaluating ML training results
• (Part #7) Product matching via ML: Post-processing
• (Part #8) Product matching via ML: Testing on various industries/languages
• (Part #9) ML does work, but it’s not magic
• (Part #10) Product matching via ML: The results

New competitor / retail website discovery

In Price2Spy we believe that you (as our client) know very well which websites we should monitor for you. However, from time to time we encounter cases where that’s simply not possible, because new websites keep popping up.

In that case, we would happy to offer you a website discovery service, no matter if we’re talking about competitor websites, or new retail websites selling the products manufactured by your brand.

The way Website discovery works is the following:

• We’d periodically check if there are new websites selling your products. The period depends on you and can be anything from 1 week to several months. You can also specify if we should check all sites worldwide, or we should focus only on specific markets
• We won’t automatically start monitoring product prices on these websites. Instead, we’d provide you with the list of new websites, and wait for your approval – which websites you consider relevant
• Only after your approval we’d start matching process on selected newly discovered websites.
• This process is applicable both for Automatch, hybrid Automatch or manual product matching

One special case of website discovery would be the discovery of new sellers on marketplaces like Amazon. It works the same way as website discovery, but this time we won’t be returning a list of newly found websites, but newly found sellers instead.