One of the most repeated stats about online marketing is that 44% of the purchases that happen online from advertising are for products unrelated to the ad(s) clicked on. If an online marketer were able to see the ad(s) that a consumer clicked en route to purchase, and discovered that the product they purchased was unrelated to the ad(s), how would they attribute sales credit across the ad(s)?

During the Attribution Management Forum 2.0 (AMF) on Jan 29^th, 2009, we posed that question to an audience of hundreds of senior online marketers and asked them to vote on the attribution model they would use in the following scenario:

In this scenario, a consumer did a search looking for “running shoes” and clicked on an ad for the Finish Line, but not did buy. They did a second search which was more refined and looked for “woman’s Nike Shox” and still did not buy. They then did a branded search for “Finish Line” and made a purchase. However, the product they purchase was unrelated to the first two ads.

We provided the audience with 3 attribution models they could choose from. They were as follows:

If you voted for:

A) Attribution Rule A, there were 3 ads involved before the sale. Regardless of which product was purchased, all 3 ads contributed and deserve equal credit for the sale.

B) Attribution Rule B, the product sold does not matter. However, when a branded term is used at the end of a path, it is being used to navigate back to the Finish Line. The customer could’ve found the Finish Line through the address bar or bookmark, however, they chose to use search again. In this instance, the consumer has already decided to buy and no sales credit goes back to the branded term at the end of a Purchase Path. Credit for the sale is split evenly between the first and second search.

C) Attribution Rule C, the purchase has nothing to do with what they searched for. Therefore, credit is excluded from the terms that do not relate to the product bought. All of the credit of the sale is attributed to the final ad for “Finish Line”.

The following charts highlight how the AMF participants felt attribution should be given, broken out by the indicated Attribution Management Experience Level:

What is interesting from these results?

• We presented the same scenario in the first Attribution Management Forum, but did not disclose the type of product sold. All the audience knew was that a sale occurred after clicking on the same 3 search ads. Over 76% of the group with the most AM experience voted to exclude giving credit to the branded term at the end of the path and favored dividing the credit across the first 2 search ads. In the scenario above, where the product sold was not related to the ads, only 17.95% chose to vote the same way.
• Almost 36% of the experienced group voted to give credit solely to the branded term at the end of the path. This was not the case in AMF 1.0 where the experienced group thought a branded term at the end of the path was used for navigation and was not deserving of sales credit.
• It would be interesting to know how the audience would have voted if the 3^rd search term was not a branded search, but instead was another running shoe related term.
• Why do people search and click on ads for one type of product, but then buy something very different? One obvious reason could be in the case of gift shopping. Perhaps you really wanted to buy someone shoes and when you got to the site another product caught your eye. Or perhaps your selection of shoes was not good enough or your prices were too high, but your selection and prices on other unrelated products of interest were in line.
• If a retailer could isolate ads that sell unrelated products, they could try to understand why they have difficulty selling those items. It probably has to do with the 4 P’s of marketing: Product, Price, Promotion, or Place.

**Disclaimer- in order to do our analysis we broke the audience down into 2 buckets: Bucket 1 is based on skill level and use of AM today. The 2^nd bucket is based on the type of business they work for: B2B, B2C, or agencies. In the observations, analyses, and questions you see below, we have taken the liberty to share what we think are the most interesting data points.

Poll 3:

In this poll, the majority in each group believed that distributing the profitability evenly between the Comparison Shopping Engine (CSE) and branded term was the correct rule. However here at ClearSaleing, we would be in the minority because we would exclude giving credit to the branded term.

CSE with Branded Search

One of the most common Purchase Paths we see when CSEs are involved is that a consumer starts with a CSE, discovers a product/place where they want to buy from, but then they go to a search engine and search for that company by name. Perhaps the reason a consumer does this, is that they are uncertain how a CSE makes money. Therefore, they want to check the site directly to make sure the CSE is not charging a premium on price.

We would like to hear from the audience why there is such a drop off from excluding branded terms when a search was involved as opposed to when a CSE was involved.

**Disclaimer- in order to do our analysis we broke the audience down into 2 buckets: Bucket 1 is based on skill level and use of AM today. The 2^nd bucket is based on the type of business they work for: B2B, B2C, or agencies. In the observations, analyses, and questions you see below, we have taken the liberty to share what we think are the most interesting data points.

POLL 2:

The only group where the majority believed you should avoid crediting a branded term at the end of a Purchase Path are those who indicated that they are doing AM and that it is working well. Although, there were at least 37% in every other group that felt that same way.

At ClearSaleing, we are firm believers that branded terms at the end of a path don

**Disclaimer- in order to do our analysis we broke the audience down into 2 buckets: Bucket 1 is based on skill level and use of AM today. The 2^nd bucket is based on the type of business they work for: B2B, B2C, or agencies. In the observations, analyses, and questions you see below, we have taken the liberty to share what we think are the most interesting data points.

Poll 1:

No matter the skill level, across all the groups that were polled, the consensus is that Even Attribution is a more accurate rule. The second most popular rule is Last Attribution. However there were a few in each group that believed 1^st attribution was the best rule.

What we would like to know, is from the groups that indicated they do AM and it works well, is there some data that supports this position?

• For example- what is your method for proving one attribution model works better than another?

• What changes in campaign performance have you seen after switching from a last attribution model to even?
• Finally, for everyone, are there any other models, aside from first, even, last, that you believe are even better?