Bid Management Requirement #2: Tight, Deep, Smart, Flexible Tracking
- Deep Tracking: Item Level tracking provides granular insights on what products sell best after general searches. This information can help inform page design and merchandising decisions. If coupons and handling charges are treated as individual items analyses can reveal what keywords tend to attract discount/coupon shoppers versus full-price shoppers.
- Smart Tracking: What to count and how to count it matters.
- Cookie windows should be calculated based on observed data and offline marketing efforts, and choices should not be limited to 7, 14, or 30 days. For some of our clients a 2 day cookie makes sense, for others it's 10 days. The limits of the system shouldn't dictate this important component of the metrics.
- The tracking system should understand the difference between brand and non-brand traffic. While the fraction of customers clicking on both a non-brand and a brand ad is relatively small (2 - 6% of brand purchases for most retailers) we think credit should be given to the non-brand ad regardless of the order in which they were clicked.
- Some day, retailers will have a really smart credit allocation mechanism that will allow them to construct infinitely complex rules for assigning credit for orders. There are a number of companies heading down this path and we wish them Godspeed, though we may try to beat them to the punch with the first really good system. To date the tools are pretty rudimentary and inflexible, though they're getting better. When this day arrives the Bid Management system should be able to absorb partial credits seamlessly.
- Flexible Tracking: Goals and internal visibility shouldn't be limited by your search provider.
- The system should support, to any level of depth, other internal analytics packages both by passing tracking to that system and by accepting source codes from that system for use in credit/ no credit calculations.
- The system should be able to handle Sales or Margin or Leads or Catalog requests or hybrids of these, such that keywords that generate few sales but lots of quality catalog requests are bid according to a truer picture of their value.
- The system should be able to incorporate additional data after the time of purchase that may reflect on the value of that transaction, whether that's frauds and cancels, or account funding information. We'll talk more about "backfeeds" in the next post.
What you track and how well you track significantly impacts the quality of the work your analysts and your bidding algorithms can do.
Part 1 of Bidding Technology Requirements