Max Obermuller, Optimisation Analyst at Infectious Media, explains why using CTR's to lead your click based RTB campaigns is a bad idea
Last week we got asked by an advertiser to run a click based RTB campaign with click-through rate (CTR) as the lead KPI metric. It is not immediately obvious, but there are strong reasons why this is a bad idea.
In display advertising, there are a range of KPIs that advertisers can use to judge the performance of their campaigns. The various KPI’s all have their strengths and weaknesses and will give advertisers different insight into how the campaign is performing. KPIs, which are usually based on user clicks (CTR and CPC) or user conversions (CPA, CPUL and ROI), are chosen carefully by successful advertisers based on the end goal of the campaign.
With CTR, this has long been used for testing performance in display advertising, and has proven to be a valuable metric when purchasing inventory via a fixed CPM when buying direct. But CTR is not a valuable way to measure performance when running a dynamic CPM RTB campaign, and here’s why.
As we know CTR measures the rate at which users click on the ads that advertisers have served, so a high CTR can indicate to advertisers that:
- 1. The content of the site they are advertising on is relevant to the product being advertised.
- 2. The site attracts users belonging to an advertiser’s target audience.
- 3. The ads are displayed in a good position on the site, allowing users to easily view and click on them.
- 4. There are not too many ads cluttering the site.
When buying directly from a publisher, an advertiser has only a limited amount of information about each impression that has been served, and that information is fairly static. The advertiser knows that they are paying exactly the same CPM for each impression and could also have some idea of the domain or domains on which their ads will serve (depending on the transparency of the buy). They may also have some information about the publisher’s audience, including how many users they are likely to reach and the demographic of a large proportion of those users.
With this information the advertiser can make certain inferences as to where and to whom the ads have been served and in this case, CTR can provide some useful insight into how the ads have been received by the expected audience. With the CPM remaining the same for all bought impressions, an eCPC can be calculated using the CTR and with all other aforementioned variables remaining relatively static, any or all of the four indicators listed above can be inferred from a high CTR.
A high CTR figure in a direct buy therefore, can inform the advertiser that they are achieving a good eCPC and that the campaign is generating a high level of interest.
When buying advertising through RTB, it is sometimes assumed that CTR is just as valuable; however this is not the case. The reason for this lies in the fact that RTB purchased impressions will vary in almost every way from one impression to the next. An advertiser could spend £100 on RTB advertising and buy ads on thousands of different domains, in a variety of different placements, serving to a number of different demographics, for anywhere between £0.10 and £10 CPM. Given that each impression is so different, we have a huge number of data points available to us each time an ad is served and can no longer make the same generalised inferences into audience pools, price and ad location as we can with direct buys.
The below example demonstrates how a CTR figure becomes obsolete if even the CPM alone is variable. Given that all of the previously mentioned metrics are also variable in RTB, the effect demonstrated in the example is therefore magnified when buying through RTB. Looking at the CTR figure for the whole of the campaign therefore, does little to tell you about that campaign’s performance.
Example
All that CTR measures, is the rate at which users click the ads that they have been exposed to and does not take into account the actual cost of each click. The examples below demonstrate how a variation in inventory price can bring about far different figures for CPC and CTR.
Case 1
We spend £1 on Inventory A (Average CPM of £0.33) and £1 on Inventory B (Average CPM of £1)
Inventory A - £1 buys 3000 impressions (CPM £0.33)
These 3000 impressions generate 2 clicks
CTR = 0.067%, CPC £0.50
Inventory B - £1 buys 1000 impressions (CPM £1)
These 1000 impressions generate 1 click
CTR = 0.1%, CPC = £1
If we were to optimise towards CTR and had to choose between the two inventory sources above, we would optimise by excluding Inventory A and continuing to serve on Inventory B, as A has a far worse CTR than B. We would not have taken into account however, the fact that we are actually paying half as much money for a click when serving on Inventory A. By selecting inventory A, the CTR is worse but we would achieve twice as many clicks for our money.
Case 2
We spend £1 on Inventory A (Average CPM of £1) and £5 on Inventory B (Average CPM of £5)
Inventory A - £1 buys 1000 impressions (CPM £1).
These 1000 impressions generate 1 click
CTR = 0.1%, CPC £1
Inventory B - £5 buys 1000 impressions (CPM £5)
These 1000 impressions generate 1 click
CTR = 0.1%, CPC = £5
In this example, if we used only CTR to optimise we would believe that Inventory A and B are equally valuable inventory sources. In reality, although the rate of clicks is the same, the cost per click of Inventory B is 5 times that of inventory A and so we should recognise that Inventory A will generate far more clicks for the advertiser’s money.
In dynamic pricing RTB, these scenarios are constantly occurring and we can never know the true value of Inventory through the CTR alone. Using CPC as the performance metric will give us a far better overall picture of how much value we are actually getting for our money.
What should you be using as a performance measure?
We believe that when running a click-driving campaign through RTB, CTR can be used as a guideline for optimisation but an advertiser should aim towards an overall CPC goal rather than a CTR goal. Where an overall CTR goal will do nothing but restrict an advertiser to buying inventory with a given CTR, a CPC goal will allow an advertiser to make use of the abundance of inventory that is available to be purchased in real time.
A CPC goal will allow an advertiser to locate those inventories which demonstrate a favourable CPM-CTR combination and will allow us to achieve the highest number of clicks for an advertiser’s money. What this can do is give an advertiser a clear picture of exactly what has been achieved from their spend, whilst also giving them the opportunity to really take advantage of what is on offer to them in RTB. A CTR goal on the other hand, will restrict an advertiser to only a sample of available inventories and given the absence of static metrics, will provide them with no real idea of how their RTB campaign is performing.
