Almost four years ago, Jack Kiefer, CEO, BabyAge.com, and his marketing team started using a Web analytics package for their eretail site. They wanted to better understand and market to their customers -- but, after a couple of months, they weren't sure about the data.
“One of the things that really kind of, right out of the gate, gave us an indication that some things weren’t right was that 27%-28% of our revenue was always listed as bookmark. In nine years, we haven’t spent a lot of time building the brand. It’s been happening more virally,” Kiefer says. “But what I can tell for certain is that 27%-28% of my revenue does not come from people typing ’BabyAge.com’ into a Web browser.”
Lisa Stein, CEO and Founder, SpinLife, found her company in a similar situation during 2005. “We could match up our sales against what [the Web analytics package was] telling us our sales were, and they just didn’t match. The cross-selling stuff never made sense ... a lot of the data just never made sense to us.”
Ultimately, both Kiefer and Stein ended up going it on their own with software built in-house. This approach may be something you want to consider or you may be happy with your current system.
If you want to check if your data is accurate, the first thing you have to do is have faith in the reports from your backend ecommerce system. You know if your backend works. If it didn’t, you would have big customer service issues daily with customers complaining about missing or wrong orders. It’s essential to trust your own reports because you’re going to need them as a benchmark to test the accuracy of other systems. Assess Your Metrics
One of the simplest ways Stein and Kiefer tested their Web analytics software was to compare the results to sales reports. If sales from your analytics software don’t match sales from your ecommerce system, something is wrong.
“When [the Web analytics software] tells me that I did $3 million dollars this month, and I really did $4 million, you know something is not right,” Kiefer says.
Then, follow these strategies with other internal reports:
-> Strategy #1. Set a few simple data points
Pick some statistics from your ecommerce system’s reports that are also provided by your Web analytics software. The easiest one is total sales. You should be able to determine your sales by source, such as sales directly from your website compared to sales from a lead-generation site. Others could be your website traffic by source and your conversion rate.
The key here is to keep the numbers simple. You are looking for data points to test your software’s accuracy at a very basic level.
-> Strategy #2. Use third-party ROI calculators
If you’re like many eretailers, you’re using other websites, or networks of websites, to generate leads. Many of these lead generators, like Google, Shopzilla and Shopping.com, offer their own free analytics packages. The reports display the leads sent, cost per lead, conversion rate and total sales revenue generated. By using these reports, you can isolate the revenue and other statistics from each lead-generation service.
“What we were doing is trying to play to the strength of the other tool,” Kiefer says. “Google analytics has a really good idea of what Google traffic they’re sending us. You’re not necessarily looking for, you know, all sorts of different cool reports, but just at a very high level. How many refers did I have from Google today? How many did I have from BizRate?”
WARNING: A down side to using a network’s reporting software is that its operators will know how much money you make from it. The tools are great for running tests, but they can leave your company’s books exposed over the long term.
-> Strategy #3. Compare the numbers
After you have run all three systems (backend ecommerce, Web analytics and third-party reporting) for a set time period, compare all the data points you decided on. If all the numbers are close, congratulations! You have nothing to worry about. But if your Web analytics software is way off, you have a problem.
That’s what happened to Kiefer and his team. “Shopping.com, their numbers, which they know how many leads they sent us and they also know the conversions ... agreed with our internal calculations,” but did not agree with the numbers from the new analytics package.
“Some months there was a 7% to 10% difference in what they said our traffic was and what they said our revenue was,” says Kiefer.6 Steps to Build Your Own Analytics System
Both Kiefer and Stein ended up creating their own in-house systems. Here’s the process Stein followed:
-> Step #1. Hire the right people
You need to have the right people on staff to get started. This includes marketers with search and pay-per-click experience and capable IT staff.
It’s not necessary to hire specialists to build the system. “It’s pretty standard stuff,” Stein says. “Any good software developer ought to be able to go in because, as you can imagine, Google publishes a lot of data about their API [application programming interface]. We’re by no means the first people doing this.”
Although it’s not necessary to designate a full-time employee to manage the system, marketers and IT staffers should engage with it every day, Stein says.
-> Step #2. Design your reports
Collectively decide what metrics you want to gather and if they’re possible to generate with the people who will build the system and use it.
“My approach to getting the right reports is to write the reports first,” she says. “We said, ‘What do we want the report to look like?’ And then we worked backward from that and asked, ‘How do we get the data to get the information into the report?’ ”
-> Step #3. Build the system
Begin with the basics. Connect the cost of a lead and the revenue it generates. Search engines will tell you which keywords that customers click on and the cost per lead they sent you. Monitoring cookies through your backend will reveal the path of clicks and websites that customers followed to find you.
“Start analyzing your major keywords first. You’re going to get 80% of your results with 10% of your keywords in most cases,” Stein says. “You don’t have to analyze 10,000 keywords to make a major impact.”
-> Step #4. Run tests and make adjustments
“We tested, pulled the data in and, of course, it never turns out the way you think it’s going to at first,” Stein says. “So, you go back to tweaking, and we’d learn from it.”
Stein’s team took three months to come up with a system that had about 75% of the features they needed. Major tweaking continued for another six months, while they’re still making minor adjustments to their algorithm today.
-> Step #5. Change your marketing
After you trust the reports, make the necessary marketing changes and monitor the results. This will give you confidence in your metrics without over-extending your risk.
“We started out doing a six-month analysis, then we moved it down to a monthly thing. Now, we’re really in there on a weekly basis looking at major keywords,” Stein says. “We simply learned and tweaked over time so our algorithm is more closely aligned with our purchase pattern.”
-> Step #6. Continually monitor the results
“Once we built our own system, we were immediately able to positively impact the percentage of search dollars that we spent relative to sale. It was critical for controlling search costs. It allowed us to tell exactly what our ROI was,” Stein says.
Her online marketing now delivers the same or better returns while costing 10% to 15% less. “Anyway you look at it, it’s a whole lot cheaper than outsourcing it,” Stein says.
Kiefer isn’t spending less on marketing; he’s spending more. “But it’s a more effective spend. When you talk about your cost of marketing going from 48% on Google, down to 20%, I mean, that’s millions of dollars that you’re putting to use more effectively.”Advantages When You Own Your Data
What’s the biggest benefit to building your own analytics system? Kiefer says he has access to all of BabyAge.com’s data and can experiment with new metrics.
For example, after a customer’s first purchase, he and his team determined how much time he has to make another sale before there is a diminishing return in order size. In BabyAge.com’s case, they have 16 days to get customers to make another purchase before their lifetime value trickles to zero.
“That sounds easy, but it’s a convoluted metric to get into. [And] it doesn’t cost you tens of thousands of dollars to get custom reports written” when you own the system, he says. “Too many Internet retailers don’t understand the value that they have in the data that they have.”
In Stein’s case, her team has automated the keyword bidding process, among others. “That allows us to reduce our marketing costs by more rapidly responding to keyword performance. If I want the algorithm to behave differently and the report to be different, I walk over to the IT office and say, ‘Hey, how about we do it this way.’ I don’t have to go back and have customized software written, I just walk across the hall.” Useful links related to this article
Shopping.com: Merchant account center:
Shopzilla: Business services: