Blinding You Without Science -- SEM Lights the Way

Labor Day has passed and kids are waiting on corners once more for yellow buses and wearing backpacks bulging with books. For me, September always brings back memories of those early mornings heading off to learn History and Science laden down as if embarking on a hike in the Himalayas. In fairness, I think the backpack load looks worse now for school kids, although I do find consolation knowing they no longer have to carry it all on foot, up-hill both ways.

In talking to marketers, I believe there is an element of our old school curriculum that is applicable to our online campaigns today that we would do well to recap now and again. Somewhere in the early chapters of those textbooks we lugged for years, we all were introduced to the scientific method, which in the hands of innovators like Franklin, Salk and Edison conquered diseases and even the dark. Of course, many of us destined after graduation for offices instead of laboratories promptly filed the steps of the scientific method away for easy recall somewhere between our mental copy of the Magna Carta and the cliff notes to Jane Eyre.

Yet, conducting well-structured tests with sound planning and good execution is now a requisite skill for any marketer tasked with growing business online.  Getting into the rhythm of proper testing takes practice, but once you get rolling with an ongoing cycle of posing and testing new questions and then iteratively testing the questions raised by those previous tests, you’ll find yourself continually optimizing your marketing spend and accomplishing more for less.

All marketing channels have opportunities for testing, but one of the easiest and most impactful channels to get started with is your paid search marketing program, which can be tested and tweaked by a multitude of variables such as:  ad text messaging, ad position, landing page, keyword bids, day of week or time of day ad serving and match type.  In fact, there are so many data points and averaged metrics in paid search, that if you are not consistently testing or at least analyzing the data by looking at it from multiple vantage points, you can often be blinded to which aspects of your program are performing well or poorly.

So, in the spirit of the back-to-school season, I’ve broken out the old textbook scientific method steps and populated them with examples familiar to a search marketer. And for those of you averse to wearing white after Labor Day, you won’t even need a lab coat.

1.)    Ask a Question

a.       I compete with other advertisers selling digital cameras. What is the most compelling promotion I can offer customers searching for related terms?

b.      I offer discount cruise packages, and I don’t have enough budget to serve ads on related keywords all the time.  What are the best times to display to maximize sales?

2.)    Do Background Research

a.       My competitors sometimes offer free shipping and discounts and our sales department says that as long as promotions remain under 10% cost of sale we are profitable.

b.      Looking at my site analytics, I see conversion rates from paid search are highest on weekends, but so are costs. I also have delayed conversions from customers researching during the week and purchasing on the weekend, so return on ad spend by day is murky.

3.)    Construct a Hypothesis

a.       Messaging “Free Shipping” in my search ad for digital camera keywords will lift sales without negating profitability.

b.      Ads served on discount cruise related keywords during the weekend have a better return on ad spend than those serving during the week.

4.)    Test Your Hypothesis by Doing an Experiment

a.       For two weeks I am rotating three ads evenly for digital camera keywords.  They are exactly the same, except one ends with a free shipping offer, one ends with a 10% discount offer and the third has no special offer.

b.      I have paused my existing campaign for cruises and made two active copies of it, and I am tracking clicks through to conversion.  One campaign serves only on weekends and the other serves only on weekdays. 

5.)    Analyze Your Data and Draw a Conclusion

a.       Hmmm. When I measure the ads through to conversion, the free shipping offer ads have a 6% conversion rate, versus 5% for the discount offer and the control ad. But the conversion rates were highest on my low margin cameras, which diluted profitability.  Interestingly, the click thru rate on the discount ads was highest overall, and so it yielded the greatest volume overall at the best return.

b.      Wow. I ran the two campaigns for a month.  The weekday campaigns actually had double the return on ad spend that the weekend campaigns had.  Evidently, customers make their latent conversions on the weekend after researching during the week.  There is more search volume on the weekend and those customers do convert right away, but CPCs are much higher.  It looks like I’ll get more bang for my budget if I limit search ad serving to Monday through Friday.

6.)    Publish Your Results: Ok, so publishing results in journals might just be for the scientists.  But, be sure to document the good and the bad test results internally so other departments and marketing channels like print, television and email can benefit from the insights.  And, of course, whoever signs your checks should definitely hear about the great work, too.