Twitter’s algorithm changes likely mean fewer consumers will see a brand’s tweets.
A new generation of web analytics enables retailers to dig deeper.
In the past three minutes, ShopNBC.com has sold two watches and a sapphire bracelet. Loran Gutt, director of customer marketing, is watching the orders flow in real time. Not only does he know what shoppers have ordered, but he can easily figure out where they came from (most often one of the major search engines), what search term they used and what route they took to their purchase once they arrived at the site. Using analytics from WebTrends Inc., he automatically calibrates bids on hundreds of thousands of paid search terms to get maximum impact for minimum investment.
ShopNBC.com carries 10,000 items-everything from bracelets to elliptical trainers to mattresses. As an arm of the cable channel ShopNBC, the web site can show extremely erratic sales patterns for individual items depending on how recently they’ve been on television, Gutt says. “We’ll sell three of something on Monday, four on Tuesday, and then 5,000 on Wednesday because it was on TV,” he says. “Then on Thursday it’s 2,000 and on Friday we’re back to four.”
A guy and a spreadsheet
As a result, paid search terms don’t always work equally well, and their performance can change in a matter of hours. If an item has been on TV, the searcher may actually be looking for ShopNBC’s version of it; if not, the site doesn’t necessarily have any more appeal than others that have won the same keyword. “One of the key business questions is how much to pay for one of those ads,” Gutt says. “You can end up overpaying if you’re not getting qualified traffic.” Until 18 months ago, ShopNBC could track performance on only about 300 paid search terms. And even that took weeks. “We had a guy and a spreadsheet,” Gutt says.
These days, ShopNBC is the beneficiary of a trend toward analytical tools that take a site’s information and not only analyze it but act on it in real time, measure the result of the action and refine it for next time. Online retailers are using analytics to learn about their web sites in ways they wouldn’t have known about before.
And the information is coming in easier-to-digest formats. Brian Induni, executive director of the Web Analytics Association, worked for WebTrends for several years.
“We used to tout that we offered 500 reports right out of the box, “ he says. “It wasn’t until we asked our customers that we realized there was a problem with analysis paralysis. Big companies can afford to hire experienced analysts to develop summary reports. Smaller companies don’t have that luxury, and they struggle. Either way, an executive team isn’t interested in pages and pages of information-they want a dashboard with two or three red or green lights to indicate whether things are good or bad.”
Earning its keep
Even better is an analytics package that can be programmed to take action automatically. ShopNBC was the first client to use a new WebTrends product called Dynamic Search. Using a product and sales data feed from ShopNBC’s internal systems, the application computes how much a given term is worth: sales of a particular item, or category of items, divided by the number of buyers who arrived at the site after using a particular search term. That ratio determines the optimal bid price for the term. Instead of waiting weeks to decide whether he should change his bids, Gutt can have them changed automatically as often as he chooses. “If we see that 600-threadcount sheets are hot today, we can bump up our bids on all four engines simultaneously,” he says. The service can identify and test new search terms as well.
ShopNBC is paying for the Dynamic Search service on a month-to-month basis, and will continue as long as it’s earning its keep, Gutt says. He declines to reveal how much he pays for it but says it pays for itself by ensuring ShopNBC.com wins keyword bids for the words and phrases that are most profitable, and by finding effective new phrases. It also saves meeting time.
“We used to spend three to five hours a week meeting” to strategize paid search, Gutt says. “Now we spend zero hours. We input our business rules into the program and it places ads consistent with the rules. We meet occasionally to discuss changes in the rules-like whether we should do something different for Christmas.”
Analyzing internal search
While ShopNBC uses analytics to refine its Internet search engine marketing programs, Jamestown Distributors uses automated analytics to hone its internal search function. The seller of boatbuilding and marine supplies began as a hardware store in 1977, started a direct-mail operation in the mid-1980s and launched JamestownDistributors.com in 1999.
It conducts about a third of its business online and booked $3 million in web sales in 2006. The company recently began using multivariate analysis-looking at different competing scenarios to determine which is most effective-to make better use of its web site’s screen real estate and tweak placement of new products to maximize interest and sales.
Jamestown’s site carries 15,000 products, a small fraction of the items in its thousand-page print catalog. Using basic analytics from Fireclick Inc. and a search optimization program from Mercado Software Inc., the company’s site computes three weighted metrics when responding to an internal search query: page views, product age and gross sales, says vice president Mike Mills. An eventual refinement will be to include gross profits vs. sales, he says.
If the site used only sales and page views, then new products, which score low on both metrics, would automatically drop to the bottom of any list of results. By adding product age as a variable, “we can dump new products into their categories and trust that the cream will rise to the top,” Mills says. “We can launch many more new products because of the analytics. If they start selling, they come up even higher on the page.”