E-Commerce Case Study: DollarDays Cashes In After Move to Solr

• Successfully transitioned from database search to commercial search to Solr
• Scaled search features quickly and flexibly with the growth of the business
• Quick, convenient index updating without heavy hardware footprint significantly reduces TCO

Customer Overview

DollarDays is an online wholesaler and closeout company that helps small businesses to compete against big box stores by offering more than 50,000 products at prices that meet or beat those available to large chains. Small retailers, non-profits and religious groups can order their goods – ranging from clothing to office supplies, toys, and greeting cards — at competitive wholesale prices, and take advantage of tools that instantly create seasonal departments and marketing materials.

In Search of Search at DollarDays

Initially DollarDays implemented search through direct SQL queries using light matching.  With only 5,000 products to search through, query times were manageable. However as DollarDays grew and added more customers and products, it became difficult to keep up with queries, so a hosted site search solution called Atomz was chosen. User queries would be submitted at DollarDays and then run at Atomz. This became an issue as DollarDays wanted keep visitors on their own website and not refer them externally.

Endeca was next chosen as the replacement search application and was used for a few years until a variety of issues came up:

• Cost – support costs were high
• Performance – scheduled updates had to be run hourly, and it was difficult to deal with the constant switching to a secondary Endeca engine – updating the first instance, switching back, and updating the second. This happened often as changes were made to the product lineup daily, as products became available or went out of stock.
• Large indexes were chewing up large amounts of disk storage

In 2007 DollarDays decided to look for a replacement for Endeca. In looking for an alternative Endeca, DollarDays first investigated Lucene, but found the need to write a server from Lucene to be more than they’d bargained for – which led them to Solr. They downloaded Solr and found it easy to install and get running. After testing, Solr proved to be reliable, and met all requirements – and the price was right.

“Solr, hands down, beat Endeca’s capabilities.” Ryan Grange, IT Manager, DollarDays

The fact that Solr was open source software was significant – DollarDays uses open source wherever possible. They want to avoid being dependent on a single vendor, and they like to can see source code and make modifications as necessary.  Solr scored an advantage on this front, and was thought to be easier to tweak to get desired results when compared to proprietary search software.

While open source was a key factor, management at DollarDays felt more comfortable about Solr once they learned that Lucid Imagination was there to provide support and consulting as required.

Solr Results

Once DollarDays implemented Solr, product records could be added or deleted on a single instance, with no need to stop and then restart, or re-index. Solr’s ability to add, remove, and edit items without going offline was key. Solr could easily keep up with rapid changes. Now the search application could be optimized at night when the system is least busy.

Complaints about difficulties in finding products have been reduced since implementing Solr, as search results have improved.

After switching from Endeca, DollarDays saw much greater efficiency in storage requirements – “the  Endeca indexes were enormous – we were getting panicky about the hard drive space needed, but with Solr, it just goes and goes.  We’ve got massive space to spare now”, said Ryan Grange, IT Manager, DollarDays.

Solr Solution

DollarDays saw a number of key benefits in Solr 1.4 and had an aggressive schedule to adopt it in production. They wanted to quickly take advantage of improved search suggestion capabilities, and performance improvements. Once implemented, they found that Solr 1.4 was significantly faster running updates and re-indexing documents, in fact updates were accepted two to five times faster, and re-index of the entire product line took about half the time. Another key feature was the ability to promote or elevate key products in customer queries.

Finally, the nature of the business at DollarDays – rapid turnover of a high variety of items across hundreds of categories – lends itself to Solr’s flexible categorization, especially through the use of faceting and boosting. The site is organized around dozens of store categories, from arts and crafts to lawn-and-garden, and nested categories within them. Solr’s ability to boost certain results to ensure that certain keywords always match a defined set of results helped them align search with promotional objectives.

Says Grange, “If the customer searches for this term, we would like these documents to appear at the top,” which when you’re running an ecommerce site, it’s very important to be able to promote particular items and that was something that we weren’t getting before.”