At RKG, we've long argued that much of what you hear about the "search funnel" is overblown. Over six years ago, we presented a study of paid search click streams supporting this viewpoint, and it's a topic we've come back to a few times since. With multi-channel attribution data, including in our Q4 report, we've found that, more often than not, only a single digital marketing channel precedes an online purchase, but also that different attribution schemes can still shift credit significantly for certain channels. Combining those two ideas, we recently analyzed click path data for orders that were preceded either by a paid search touch or an organic search touch within 30 days of the purchase to see what the numbers might tell us about how searchers are utilizing the two channels. This data covers orders placed in early February and it comes from our proprietary Attribution Management Platform. For simplicity, we're ignoring direct traffic and touches from other marketing channels and focusing solely on search dynamics.
All Orders with a Search Touch
Our 2006 paid search click stream study came to a few high level conclusions, including: 1) many click streams are short, 2) many click streams are redundant, and 3) your brand name matters a lot. These findings apply equally well when we include organic search touches. We'll address all three, but let's start with the first two points. On average, for 63% of orders with a search touch, there was only a single search touch. For another 15% of orders, there was more than one search, but all search touches were on the same phrase and all were from the same channel, either paid or organic. So, very quickly we reduce the number of search orders requiring any meaningful thought on credit allocation to about 1/5 of all orders: The picture gets murkier when it comes to assessing two other segments: 1) orders where all search touches are from a single channel, but there are multiple phrases (about 13% of search orders) and 2) orders where there are search touches from both paid and organic (about 10% of search orders). Divvying credit across keywords appropriately in the first case is not trivial, nor is it a settled issue among marketers. But, it is a topic we have addressed previously in the links above, so again, check those out for more detail on our take. Generally, we would say that giving 100% credit to the last keyword touch before an order is an extremely close approximation of the credit each keyword would receive if credit were based on each keyword's true impact to all click streams. That is assuming we make a distinction between keyword phrases containing our own brand, which are primarily navigational in nature, and non-brand phrases.
Orders with Both Paid and Organic Search Touches
Taking that point, and focusing on our other remaining segment, the 10% of search orders with both a paid and organic search touch, we can further break down user behavior. Among this group, we find that a little over half are orders where all of the search touches were on brand terms. In these cases, one could argue that neither channel necessarily deserves much credit as the navigational brand searches were likely driven by other factors such as offline marketing. We should still care about attributing these orders properly, but since we should be steering primarily by our non-brand performance, they are not the ones that are going to keep us up at night. What remains are two groups that comprise about 5% of our original total of search orders: 1) cases where there are both paid and organic searches and all are on non-brand terms and 2) cases with multiple search channels and a mix of brand and non-brand terms. Getting the attribution of these orders "wrong" could result in over or undervaluing the non-brand contribution of either search channel and change the amount of resources we are willing to devote to them. In the first case, when all searches are on non-brand terms, we find that moving from a 100% first to 100% last touch attribution scheme shifts credit across channel nearly 80% of the time. That may sound high, but again, most click streams are very short -- remember about 2/3 of all search orders had just a single search touch. For a multi-channel order, most search click streams are also the shortest possible length -- in this case, two touches -- and in every single one of those cases, credit will shift from one channel to another. It's important to note that we can't just assume it all eventually evens out as paid search is on the losing end about 2/3 of the time when all phrases are non-brand and credit shifts channels. In other words, giving 100% credit to the last touch is not as good an approximation for multi-channel search orders as we found it to be for single channel search orders. When we have a mix of brand and non-brand touches across multiple channels, we are not only concerned with how credit should be split by channel, but also whether we are giving undue credit to navigational brand searches. Again moving from first to last touch attribution, we find that credit shifts from non-brand to brand, or vice versa, in about 3/4 of these cases. Credit is over two times more likely to shift from non-brand to brand than the reverse and paid search is more likely to lose a non-brand to brand cross-channel shift than organic.
Moving Beyond Search Touches
With orders where credit shifts from non-brand to brand and across channel, we are now talking of a subset of a subset that equates to about 1.2% of all orders with a search touch. This is a small segment, but it's really just our worst case scenario, and it doesn't represent all ambiguous orders. But, if we can determine how best to divide credit in these cases, we can apply that knowledge to the larger pool of search orders and to orders with other channels in the mix as well. In the grand scheme of attribution, the overlap between paid and organic search and the resulting potential for shifting credit due to various allocation schemes is actually relatively minor compared to those scenarios involving certain other channels. Not surprisingly, discount distribution channels like affiliates generally hold much greater potential for skewing credit allocation. As marketers we need to make sure that our attribution models are tuned for these factors with our insights, but without our unduly putting a finger on the scale to get a specific outcome. One last takeaway, with just 10% of search orders having both paid and organic touches and 63% having only a single touch from either, our results speak to the need to do both SEO and PPC well, but that doesn't mean that those efforts can't benefit from one another. While some customers may only click organic links and we may have just one opportunity to sell a search customer on our products or services, we should position ourselves to best take advantage of that opportunity using the knowledge we've gained across all channels.