Traffic cannibalism. Is it necessary to conduct contextual advertising for branded queries?

Problems

Branded search traffic converts into actions better than any other type of traffic. I want to increase the volume of branded traffic to get more inexpensive actions.
If the project does not have positions in organic search for branded queries, but there is information demand, then there is no doubt that context for branded queries needs to be maintained.
But if the site is in first place in search results for all branded queries, the answer to the question – “is it worth running context for branded queries?” is not so obvious, there is a risk of traffic cannibalization.

Cannibalism of website traffic sources is a partial or complete flow of traffic from one source to another (one source eating up another source). That is, traffic from one source grows at the expense of another source and the total increase in traffic is small or absent altogether. For example, in a period there were 100 visits from search, in the next period they bought 100 visits from context, but the number of visits from search in this period decreased to 20, and the sum of visits from both sources was 120 visits, and not the expected 200. This is an example of partial eating of search traffic by contextual advertising.

Method

To answer the question “is it necessary to conduct contextual advertising for branded queries?”, it is necessary to conduct an experiment:
Launch an advertising campaign in context, measure the results (number of visits) with advertising and the results without advertising and calculate ITR (Incremental Traffic Ratio) – additional traffic ratio.

ITR=(nON – nOFF)*100/K,

where nON is the total traffic from organic and contextual traffic during the period of running the advertising campaign,

nOFF – organic traffic during the period of no advertising campaign,

K- branded traffic from RK.

ITR is calculated as the difference between the total number of visits from organic and advertising traffic during the period of running an advertising campaign and the number of visits from organic traffic during periods when there was no advertising campaign, divided by the number of visits from branded advertising campaigns.
That is, we calculate how much the total traffic has increased or decreased in relation to the attracted advertising traffic.

How are possible ITR values ​​interpreted?

In an ideal world, 100% cannibalism would flourish.

The ideal situation is as follows: a site in the SERP (Search Engine Results Page) is in first place, and above it there is one contextual advertising ad leading to the same site.
The ad text and position snippet are the same, and both links are visible to users on all devices/resolutions.
The user clicks on the link that is the shortest to reach with the mouse. That is, on the ad in context.

In this case, ITR = 0%. There is no traffic growth, i.e. organic traffic completely switches to context. Zero-sum game. Ideal traffic cannibalism.

But the real world is not perfect.
The picture that the user sees in the search results is more diverse, it has more links than the two from the ideal picture. The links have different anchors. In special placement there is not one ad, but several, in organic search the user sees several positions, they have different snippets. For some queries, the search engine gives several links to the site in the SERP.

All this makes the possible outcome of including advertising campaigns for branded queries not as clear-cut as in an ideal world.

If ITR is greater than 0% and less than 100%, then the context brings in more traffic than it takes away.
That is, there is additional traffic.
This means that some users would not have reached the site without advertising, but would have followed other links on the search results page, therefore, the context for branded queries is justified.
How justified it is is determined by the cost of additionally attracted traffic and conversions on it. But it is clear that 80% growth is better than 10%

If ITR is 100%this means that context does not reduce search traffic at all.
This is the ideal option for the complete absence of traffic cannibalism. The CTR on the search did not change, and all users who are accustomed to clicking on the first links in the context were additionally attracted.

If ITR more than 100%this means that contextual advertising magically increased the CTR in organic search results. That is, links to the site began to be clicked more often than without advertising. This situation is possible if the ad advertises the subsequent organic link.

If ITR takes negative valuesthis means that the context had a negative impact on the total traffic. That is, some users decided not to click on either the contextual ad or the organic link at all. Theoretically, this is possible, but it is difficult to imagine how to achieve such a result.

Experiment

To correctly compare periods with and without advertising, we used a “checkerboard” with
with a discreteness of one hour. That is, the advertisement was switched on according to the schedule every hour, and the next day the hour of switching on was shifted, then repeated. And so on for two weeks.

Why such complexity? To minimize the influence of external factors on the experiment that appear if you alternate a week with advertising and a week without advertising sequentially.

In our case, two weeks were enough to obtain results with an acceptable margin of error.
Moreover, the advertising campaign used all variants of brand queries for which the site has the first positions – 72 queries in total.

After the end of the campaigns in Yandex Metrica, we received data on visits from organic traffic and visits from our text advertising campaign by hours.

In our case, search traffic is quite large compared to the brand segment, so to improve the accuracy of ITR calculation, organic traffic was filtered by the same list of queries that the advertising campaign was conducted on.

Interestingly, 12% of clicks in the context in the site statistics fell during the hours with disabled advertising. But, nevertheless, we also counted them.

We collected data on types of traffic in a table, calculated it and got ITR = 81%

ITR=(nON–nOFF)*100/K

81%=(439–272)*100/207

Efficiency

ITR = 81% is it a lot or a little, good or bad? And how much is needed?

Ideally, you need to calculate ROMI for additionally attracted traffic
ROMI Inc = (Profit from additional traffic requests – Costs for branded context) / Costs for branded context).
And if ROMI Inc is greater than zero or meets target indicators, then the context for brand queries is justified.

Profit from requests for additional traffic is calculated as the average profit from requests from the site, multiplied by the number of requests for additional traffic. The number of requests for additional traffic is calculated as additional traffic, multiplied by the average conversion of contextual branded traffic.

If it is impossible to calculate ROMI, and in our case it is – in B2B such a situation often occurs, you need to compare the cost of action for additional traffic (CPA Inc) with the target value of the cost of action – CPA targ.

CPA Inc=(CPC/ITR/CR)

Where CPC is the average cost per click in the advertising campaign for branded queries

ITR – additional traffic ratio

CR – conversion of branded contextual traffic

In our case, CPA Inc = 1729 rubles = (46 rubles / 81% / 3.3%)

The target CPA targ value in the project is 7000 rubles. With an ITR of 81%, CPA Inc = 1729 rubles, which is very good, since the cost of action is significantly less than the target value.

What is the maximum ITR value for a project?

This is the value at which CPA Inc will be equal to CPA targ. Therefore, the minimum ITR value for the project (ITR min) can be calculated using the formula.

ITR min= CPC/CR/CPA targ*100

CPC – average cost per click in the advertising campaign for branded queries

CR – conversion of branded contextual traffic

CPA targ – target value of cost per action for the project\.,

ITR min = 17.4%=46 rubles/81%/7000 rubles*100

The minimum acceptable value of the additional traffic coefficient for the project is 17.4%.

Conclusion

In general, the method turned out to be quite easy to implement.
It is quite possible to apply it not only to branded queries, but also to any segments of the project's semantic core, for which there is a suspicion of the possibility of traffic cannibalism.

To avoid having to reinvent the wheel, we are sharing a template of a working file for conducting an experiment. The template is available by link.

Author: Rublevsky Roman, Director of OOO “Summa Technologies”