Build-A-Bear Workshop has built its reputation on creating a unique shopping experience. If the chain’s network goes down, however, the entire customer experience is in jeopardy. By adding a new high-speed network, the specialty chain is ensuring that downtime is a non-issue.
St. Louis-based Build-A-Bear Workshop is a $474.36 million company that helps its visitors bring an array of teddy bears, animals and cartoon characters to life. Like most retailers, Build-A-Bear Workshop grows more reliant on the data it collects during every customer visit, and it needs to ensure that it can transmit and store data as efficiently as possible.
“We have a significant amount of data moving between stores and our corporate office on a daily basis, including credit card, check and gift-card authorizations, as well as internal data needed to update PC and kiosks,” said David East, director of IT network services, Build-A-Bear Workshop.
However, an aging DSL network was taking its toll on data transmission. Five years ago, the chain established a virtual private network that ran on top of a 128K DSL network. “At the time, we did not have the same data needs as we do today. As we’ve grown, our network began running out of bandwidth and response times began slowing,” he explained.
The high number of voice-based telephone lines used at store-level also taxed the network. Each location had five lines. Three were reserved for incoming calls and communications between headquarters and other stores. The other two were backup connections for the DSL network.
The company’s first remedy was to expand bandwidth to 192K, and even 384K where it was available for a reasonable price. But that strategy still fell short.
“We worked with vendors that reported more than a 90% rate of success when connected to 192K,” East explained.
“Because of our positioning within shopping malls and the telecommunications companies’ central offices, we experienced rates between 50% and 60%,” he said.
This prompted Build-A-Bear Workshop to transition onto a network based on T1 lines. As a result, the chain wanted a MPLS [multi-protocol label switching] network that functioned on a T1 circuitry. (MPLS gives each data packet a label that designates its network path, thus speeding up network traffic and better managing flow.)
After putting out a bid to major providers, it chose to work with AT&T, San Antonio.
With an MPLS network in place, Build-A-Bear Workshop was able to remove one of the two back-up telephone lines from each store. “Using a T1 line makes the second back-up line unnecessary,” East explained.
The network also helped the chain transition all internal communications between stores and the corporate office to a VoIP (voice over Internet protocol) network, which allows voice data to be sent over the Internet.
The chain also tapped Cisco, San Jose, Calif., and added T1 interfaces and upgraded its routers at each store.
Next, the chain created an arrangement between its corporate office, its backup facility located in Columbus, Ohio, and its stores. Stores are connected to both facilities.
“If anything happens at corporate, critical information is also filtered through the Columbus facility,” East explained.
The same goes for data traveling between corporate and stores. As transactions are processed at POS, the network pushes up the transaction through corporate to the Internet processor. The same flow handles data being pushed to stores, including intellectual data, graphics, or frequent shopper information.
To ensure privacy, Build-A-Bear Workshop uses a private connection when filtering credit-card data to its processor. The same action is taken for other sensitive data, including customer information.
Build-A-Bear Workshop implemented the solution within four months and began testing the solution across 90 stores in October 2007. Since completing the first phase of the project in December, the chain has increased its uptime from 99.50% to 99.99%.
“This may not sound like much, but it makes an impact when there is downtime involved,” East said. “We have had extreme cases where DSL would go down and it could take days to get it back online. Now our latency has dropped from 150 milliseconds to between 20 and 40 milliseconds.”
“Now we have a plan in place to separate voice traffic from the rest of our data communications, and circuits will let us increase bandwidth as needed, at an easier pace,” he said.