When a consumer searches for your brand name, clicks on your search ad and makes a purchase, who should get the credit for that? Did that purchase happen because of a specific marketing tactic on television? Radio? What about display or social media, or even the search ad itself? These questions cut to the essence of attribution, which is leading to major turf wars within marketing organizations. If, for instance, a new attribution model assigns a greater value to television, that team can expect to receive a greater share of budget going forward. But this decision may not sit well with the search team or other marketing groups vying for budget. It doesn’t help that attribution models often involve imperfect data and myriad assumptions. The truth is there is no perfect way to know the value of each marketing tactic. What I’ve come to believe is that all tactics are interdependent, and are important to stoking demand at the top of the funnel and closing the sale at the bottom of the funnel. In that spirit, I recommend that all search marketers resolve to try the tests outlined below. Executing them can prove the reliance each tactic has on the others, promote more accurate attribution models, and might even change turf wars into pizza parties. 1. Turn off search – Yep, I wrote that. If you want to prove how vital search itself is to lead-flow, turn it off. A great way to execute this test is to “go dark” in three tofour smaller markets that represent different parts of your demographic. Throughout the process, you’ll want to measure many specific things in those markets, such as net sales, social media chatter, store visits, phone calls and loyalty metrics. As part of the test, find objective ways to measure how often customers complain about how difficult it is to do business with you when search is “off.” An underappreciated feature of search is how easy it actually makes purchasing for the consumer. Run the test for one tot three months, and when the test has completed, assess how long it takes at previous spending levels to return to your pre-test metrics. This is incredibly valuable at quantifying the impact of “pausing search” at times when budgets get tight. 2. Turn off television. – Execute this test in a handful of diverse, smaller markets. In addition to insights derived by the metrics stated above, you’ll have the added insight of search impressions and search clicks. Run the test for one to three months, and you may notice a “half-life” for television ads. Many branding ads induce “real-time” searches for consumers that are ready to buy, as well as searches over the coming weeks or months for those who are still considering solutions. If possible, run similar tests across different markets for radio, display and other campaigns to understand those tactics’ effectiveness. As with the search exercise, assess how long it takes at previous spending levels to return to your pre-test metrics. 3. Test how non-brand ads affect brand searches – Bidding on your category in search is much more expensive than bidding on your brand, and will drive up your cost per acquisition (CPA) metrics. Search marketers are often loath to reduce brand spending in favor of category ads, since they are higher risk and subject to unexpected fluctuations. Category ads, though, are an opportunity to have attribution models work in favor of the search marketer’s budget, which tends to be scrutinized purely by CPA. Category ads induce purchases and also improve awareness, meaning that they deserve a share of phone calls, purchases or in-store visits that happen later. One way to test the impact of category ads is to increase category ad spend in several markets and reduce it in others. Measure the changes in brand search impressions, brand volume and the “unexpected” changes in net sales from that region. While there are certainly more sophisticated tests that can be done, this data should start to give you a healthy appreciation for the contributions of each marketing tactic, and begin to help you quantify the impact of major budgetary changes. And who can argue with that?