Prediggo VS Solr

If there is one common question asked by all of our customers who look at our eCommerce Search engine, then it is this one: "What is the difference between Prediggo Search Engine and Solr?" <p >This post will not try to kill Solr, because it's a good search engine. However, we want to show that it's not the best option when it comes to eCommerce. Why?

First of all, let’s look back and see where Solr comes from. Solr was created at CNET Networks as an in house project for their internal website. Its core objective was to index and search web pages, and thus it was designed as such. But searching the web is not like searching products in a database. The best example why this is different is the search for a general keyword like “notebook”. As Solr is a textual search engine, it finds all the products with the most “notebook” keywords, which happen to be the accessories rather than the notebook themselves. As a result, you will have tons of accessories on the first pages, while the actual notebooks will be far behind. The reason is because Solr is not able to understand the meaning of the keyword “notebook”. Prediggo is different and was designed to understand the context behind keywords and thus show better search results.

Nowadays, Ecommerce vendors are usually not IT geeks, where most vendors have a marketing background. Unfortunately, Solr was designed for IT engineers and is still is today. In short, you cannot double click on an icon and have it run in 2 hours. It requires complex installation procedures and settings in Tomact (the web server that hosts Solr) to have it run properly. Then, working with it is not any easier. One of the most common things an eShop will need to do is add or remove synonyms. Well in Solr, you will have to manually edit a text file (usually in /conf/synonym.txt). Then, you must update the XML schema to tell Solr how it should interpret the synonyms (ie: by writing something like :

&alt;filter class="solr.SynonymFilterFactory" synonyms="synonym.txt" ignorecase="true" expand="false"> &alt;/filter>

). To make things more complicated, these files are usally on an Linux Server somewhere at your hoster, while the eCommerce vendors will have a Windows PC at the office (i.e: welcome to the world where line breaks are different in Linux, and rights on directory usually not 0777). This is another point where Prediggo is different as all of the settings are done from the Prediggo Back Office which is a web interface that can be accessed and easily maintained as long as you have a web browser and internet (and the access code obviously…). Now, Solr has an admin back office, but so far it only contains basic search and index statistics.

Finally, but not least, the customers support. Ecommerce search engine is one of the most critical tools in an eShop, where over 70% of visitors use it at one point or another. What happen if it goes down? Who do you call? It’s an open source project with a community. So it does have a bug tracking system, but it can take weeks before it gets solved. Can you afford to stay weeks without a search engine?

Posted on March 25, 2012

Recommendation Engine or Recommendation Roulette?

What makes a recommendation engine?

Any technology that powers on-site personalization strategies is technically a recommendation engine. However, not all recommendation engines are created equal. A truly effective recommendation engine has three key components;

• Understands the context of each user’s shopping behavior (real-time user profiling)
• Analyzes the product catalog using multiple parameters (attributes, popularity, price, inventory and more)
• Supports custom, multi-level strategies for each recommendation displayed (cross selling, up selling, merchandising, customer also bought and more)

So when a site only uses a transactional based recommendation strategy like “Customers Also Bought”, then they really don’t have a true recommendation engine. They are relying only on analysis of their catalog based on one parameter, purchase interaction. <p >As you can imagine, using only one product relationship parameter dramatically limits the level of personalization and often ignoring the context of that user’s shopping behavior.<p >So when you are evaluating a recommendation engine, remember to consider how it approaches the three key elements;

• Real Time Unique User Profiling
• Product Catalog Modeling
• Custom Business Rule Support

Having these three elements means that your recommendation engine can adapt to the preferences of your users and handle the complex personalization requirements that really drive incremental sales and customer satisfaction.

Good Selling!

Posted on March 21, 2012

Convert The Traffic You Have

The profitability equation for webshops is fairly simple;

Driving Traffic + Converting Traffic = Profitability

At Prediggo, we see an overwhelming emphasis on the first part of the equation, driving traffic. It seems that webshops expect their traffic building efforts to also perform the conversion role, even as they see their conversion rates drop during a heavy traffic push.

This is a dangerous approach because traffic building strategies generally involve efforts that significantly reduce margins.

• Aggressive Pricing for Comparison Sites
• Deep Price Discount Periods
• Google Keywords and Banner Campaigns
• Affiliate Commissions and more..

And at the same time you are cutting into profit margins, you are reinforcing a negative behavior, Buying vs. Shopping. In fact, focusing on traffic building alone can reduce profitability to the point that the additional traffic becomes a negative ROI proposition.

This is why Prediggo urges webshops to prioritize converting more of their existing traffic through strategic personalization. Just adding highly targeted item recommendations to product detail pages can result in a 10% increase in conversion. A comprehensive product targeting strategy throughout a webshop can increase conversion by as much as 50%!<p >And unlike traffic building, strategic personalization positively impacts your business performance in more ways than one.

• Up to 50 % More Conversion – More shoppers find something they want!
• Up to 40% Purchaser Interaction – Personalization engages shoppers!
• 5 Times More Page Views – Highly relevant suggestions inspire more shopping!
• Doubled Avg. Order Value – Engaged shoppers tend to buy more!

<p >So the next time you are thinking about ways to build profitability in the short term, remember to include on-site personalization in your planning. You will see a significant increase in the ROI than with traffic building alone.

Posted on March 24, 2012