Paid Search on the Margins

My monthly paid search column at SEL in case you missed it. In paid search the more accurately we can gauge and react to the "true value" of the traffic for each ad the better the performance of the program. Conversion rates are a pretty poor proxy for value. A $200 order is probably worth more than a $20 order, but they're both "conversions." In lead gen a "mortgage refinancing" lead from Beverly Hills is probably worth more than a lead from that same keyword in Middle-of-Nowhere, Mississippi. Teasing out these important differences is a good first step towards improved performance. But just as sales per click is a better proxy for value than conversions per click, margin per click is better than sales per click. Any retailer knows that all $100 orders are not equal because the cost-of-goods can vary dramatically. Tying the paid bidding to the actual margin driven is tremendously important for any business with significant differences in margin structures. However, getting at that margin data can be tricky. Many paid search managers "hack" at this by determining the average margin for each product category or manufacturer brand and set cost to sales efficiency targets accordingly. This is a good and helpful first cut, but falls short of ideal by a good measure for a number of reasons.
  • The buyer's actual purchase doesn't always match their search. This takes the form of buying products other than what they sought, and buying products in addition to what they sought. A quick study of two clients in very different categories revealed that this not only applies to people searching by product category, but even to people searching by manufacturer brand. In the case of an apparel retailer, 16% of the sales revenue came from brands other than the ones referenced in the search phrase. For a consumer electronics retailer the average was 12%. Particularly interesting though was the range of variance: for some manufacturer brands the "dis-loyalty/ add-on" effect was as much as 32% for the apparel retailer, and 63% for the CE retailer! Not surprisingly, the lower-ticket, accessory brand searches are more likely to generate a large fraction of sales for other stuff than the higher ticket brands.
  • Discounts and promotions depress margin percentages. During promotions, the assumed margin percentages are guaranteed to be wrong and can result in overly-aggressive advertising.
  • Return rates and cancel rates can vary significantly between categories and brands. Cancels and returns can materially impact performance and variations in these rates can render assumed averages dangerous.
  • Spillover rates to phones and stores vary as well. Higher ticket, higher consideration purchases are more likely to prompt phone calls and trips to the brick and mortar store. Failure to account for these effects can leave opportunity on the table.
More accurate measurement of the true value of traffic requires back feed loops in order to:
  1. transmit the accurate margin value on each order;
  2. update leads with actual valuation information after the fact;
  3. knock out orders that did not ship; and
  4. capture the orders driven by paid search that close over the phone.
Even with all this in place, it may be inadvisable to set a single margin-based ROI target across the portfolio to achieve an advertiser's goals. The lifetime value of customers varies depending on their initial purchase. Bidding more aggressively for those customers than for the bargain-hunting seasonal shoppers can be profitable in the long run. Some keywords and categories are more likely to attract first-time buyers to your site, and many smart marketers happily pay a premium for those terms. Those advertisers interested in their bottom-lines are advised to refine their tracking and targeting to get the clearest sense possible of their true ROI. As the old saying goes, you can't hunt what you can't see.
Join the Discussion