Make PPC advertising more profitable using returns data – Part 2


Yesterday I pointed out why returns data should be considered in PPC advertising. Only when returns information enters into tracking and bid management, is the advertising channel search able to fully exploit its potential for e-commerce profit. Because only in this way can the profit provided by an advertising initiative be correctly conveyed and optimised. The reward is highly efficient and vigorously profit-driven search campaigns. To achieve this, a few details must be taken into account. Today I’ll dive into these details, such as data quality, tracking or cancellation limits, and finish with a check list for decision makers. Have fun!

Pay attention to data quality and tracking

The basic requirement for successful bid management is detailed tracking of campaigns and user behaviour. It tracks orders back to discrete campaign elements, such as keywords (Google AdWords) or product targets (Google Shopping).

Figure 1: A typical instance of keyword conversion tracking

Professional tracking should also encompass returns. This way campaign elements such as keywords do not just receive credits when ordering in the shop, but also deductions with incurred returns. Google itself provides in-house Google Conversion Tracking. Unfortunately it cannot track returns.

However, other tracking systems are capable of doing so. They can take the returns into account in various ways, and they reflect the returns very precisely: from the lump sum of returns up to the cancellation of individual items in one order. These methods can be differentiated as follows:

  1. Shop Quote: A lump sum return rate will be deducted from all orders. This is the least accurate of all the methods, since it is heavily based on averages, but it is still better than nothing.
  2. Segment rate: Different return rates are created for different segments (e.g. product types or brands).
  3. Shopping cart level: All orders can be cancelled in tracking. To this end, the tracking system needs to know the actual orders, which can then be cancelled via the order number.
  4. Item level: Individual items of an order can be cancelled. This type allows the shop operator to represent a situation that is not entirely unknown to him: The customer orders several variations and calculates their return deliberately in doing so. To this end the tracking presented in Figure 1 has be carried over for each differentiated item in the shopping cart.

For the latter two mechanisms it is recommended to automatically factor in the occurrence of returns. In this case the shop typically automatically assigns a cancellation feed, which is effected through a technical interface in the tracking system. Using aggregations, for example, the most profitable PPC channel brands in the shop portfolio can be conveyed before and after cancellation. It is not uncommon in this case to encounter surprises and new insights.

Brands before and after returns
Figure 2: The top 5 brands of an outdoor shop change after returns

Cancellation limits for bid management

Good tracking prepares the right conditions in order to provide returns data for PPC. However, with bid management there is still a problem, since it works with the most recent historical data. If the cancellations are factored into tracking via order or basket level, you cannot know all the returns in a 30-day return policy. Because this will only occur in the future. Therefore it is helpful to define a cancellation limit. All sales that occurred prior to this cancellation limit are considered secured (see Figure 3).

Period considered by Bid Management
Figure 3: A cancellation limit facilitates bid management in dealing with returns

Bid management now has to work with data having complete and only partially known returns. It can proceed in several ways. First, bid management can exclusively work with the cancelled conversion values. In this respect it is accepted in the purchase that as yet unknown cancellations are not calculated into the new bid. On the other hand, it is possible that bid management will always work only with non-cancelled conversion values, ​​and that it will assign to these a cancellation estimation. The cancellation limit plays practically no role in this case. In the third option, bid management segments the data amount to the cancellation limit. The basis for real cancellations is provided by data prior to the cancellation limit; for the current data, a cancellation estimation with extrapolation.

Exercise caution in estimates

If bid management works with non-cancelled conversion values ​​or if the amount of data is segmented to the cancellation limit, cancellation estimates are necessary in both cases. Caution: If the offered product segments have very different return rates, the estimates should be based as little as possible on average values. A differentiated approach is necessary, because particularly good keywords for product segments with disproportionately low returns will be penalized, because the standard cancellation rate will still apply to them. And due to a bid that is too low, unfortunately they will be discouraged rather than encouraged. In short, the revenue potential will not be fully exploited because of too little traffic. However, particularly bad keywords for product areas with above-average return rates, will not be penalized sufficiently. Because of bids that are too high, there will be too much traffic, which in the end not only drives up the advertising costs for unprofitable clicks, but also the service costs due to increased processing of returns.

It is worthwhile to look more closely at the quality of tracking data. Whoever does not factor returns into the tracking and subsequent reporting, will only be able to evaluate the efficiency of PPC activities with difficulty. It holds in each case that: If the profitable modulation of the PPC channel plays an important role, bid management should necessarily include the occurrence of returns in the calculation of bids.

Check list for decision makers

If returns data in pay-per-click advertising is to be taken into account, internal information must be obtained in advance. Toward this end the respective departments or experts can be consulted.

Questions regarding business intelligence and purchasing

  • Determine which of your brands and product types particularly deviate from the average return rate for your shop.

Questions regarding tracking

  • Find out how the returns dealt with in bid management can be factored into the tracking system: Not at all, as a (lump sum) rate, or even precisely per order?
  • Will return data (affected orders, cancelled items) be automatically factored into the tracking system – in order to also cancel the recorded revenues and margins of returned orders at the keyword level?

Questions regarding bid management

  • Is the consideration of returns for the optimization objectives of bid management relevant?
  • How does bid management take into account the occurrence of cancellations?
  • How does bid management estimate incomplete cancellation rates for the current month?


Alexander Paluch

Alexander has several years of experience as an IT consultant and Product Owner in the e-commerce industry. Today he works as a Senior Product Manager for crealytics.

    Find more about me on:
  • googleplus
  • linkedin