Which TV commercial is worth watching again?

Anyone who advertises on television should also be able to measure the success of the campaign. That means: assigning the actual value of the commercials to additional purchases and sales. This is where cross-channel TV attribution comes into play.

Challenges in optimizing TV campaigns

The times are long gone when high advertising recall (Brand recall) or recognition (Brand recognition) represent the most important key figures for evaluating the effectiveness and efficiency of TV campaigns. Because more and more marketing managers are relying on data-driven analyzes, alternative KPIs and evaluation methods are needed here.

With classic TV advertising, the challenge is to identify which users actually saw a TV commercial. There is still no deterministic method of matching a user who has seen a TV commercial with his or her interactions on the advertiser's website.

From a technical point of view, this means that there is no unique identifier (e.g. cookie ID) in the TV area that can be assigned to a user who has visited an app or website. That is why statistical methods are used here.

It's easier with programmatic and addressable TV advertising over OTT or streaming services. Because more information can be tracked through the TV ad viewer, e.g. B. the IP. The challenge here, too, remains that many people see an advertisement on TV - but order the advertised product using a different device, e.g. B. via your laptop or smartphone. The advantage is that cross-device tracking (from addressable TV advertisements to the device on which the purchase was made) is much easier in this case. This is inter alia. because there is a higher probability that both devices are in the same IP network.

You might also be interested in: Case Study: How Spring was able to increase its ROAS for TV by 256% with the help of our Unified Marketing Measurement solution.

How does cross-channel TV attribution work?

In general, there are two different approaches to assessing the revenue and performance impact of TV advertising. These are as follows:

The micro perspective

The micro perspective includes the evaluation at the user level as well as the tracking of all touchpoints and interactions. The broadcast times of the TV spots in connection with the website or app visits as well as other activities are watched. This approach enables the assessment of how watching the TV commercial affects a user's direct reaction within the minute following the commercial. It is assumed that the TV ad encourages the user to visit the advertiser's website within the next few minutes after seeing the TV ad. However, this perspective does not take into account the longer-lasting brand effects that a TV commercial could have on a user, which prompts them to use the advertiser's website e.g. B. to visit until or again two weeks after the TV spot.

The macro perspective

The macro perspective takes into account the aggregated visits / leads / sales data, e.g. on a daily, weekly or monthly aggregation level. It is not necessary to assign TV spots to users directly. This analysis enables the overall impact (direct response and branding) of TV advertising to be assessed. Methods such as Marketing Mix Modeling (MMM) and the extended time series forecast can be used here. The disadvantage of this approach is that it is far less accurate and not applicable in all cases. Depending on the product and type of TV advertising, branding or direct response-oriented, one of the two methods is used. There are also cases in which a combination of both approaches can be useful. For example, to understand the additional impact television has on branding.

Micro perspective - user-based TV attribution

In general, this approach relates the airing of a TV commercial to a statistically significant increase in the number of visitors per minute on the advertiser's website. All visits and visitors from SEO, SEA Brand and Direct are taken into account. Traffic from other sources, such as B. from retargeting campaigns or emails, is usually not caused directly by TV advertising.

The biggest challenge is to get a baseline to understand what the number of website visits would look like after a TV commercial. In concrete terms, this means: If the initial value is calculated at 1000 visits per minute and the actual number of visits per minute after a TV commercial is 3000, then one can conclude that 2000 visits were made by this TV commercial.

Adtriba dynamically calculates the baseline on a minute-by-minute basis using the most recent data from times that are not influenced by TV commercials. By comparing the initial value with the actual number of visitors, Adtriba's TV allocation algorithm calculates whether there was a statistically significant increase in the number of visitors for every minute.

In particular, care is taken to ensure that outliers and minutes in which a significant increase was detected do not distort the further calculation of the baseline value. This rolling recording of statistically significant peaks (eng .: lifts) enables a dynamic setting of the number of minutes to be taken into account after the TV commercial has been broadcast. As a rule, TV analytics providers always consider a static time frame, e.g. B. 4 or 8 minutes, which can lead to an over- or underestimation of the performance of the TV advertising.

For example, the time window is set to 4 minutes. However, the actual and measurable impact of the TV commercial lasted 7 minutes. Then all visits in the 5th to 7th minute and later conversions from these visits cannot be assigned to this TV commercial.

Number of individual visitors per minute, baseline value and peaks (signal)

In the diagram above, a significant peak is labeled "Signal" every minute. The green line shows the dynamically calculated output value. A TV spot was broadcast at 9:01 p.m., at 21:02 the starting value was 3 unique visits per minute, but the actual number of visits was 111. This leads to an increase of 36 (= (111-3 ) / 3 and a 97.3% (= 108/111) probability for every visitor to the advertiser's website in minute 21:02 due to this TV commercial. If one of these 111 visitors orders in the future, 97.3% will be able to For example, if there are 20 orders from these 111 visitors in the future, 19 can be assigned to this TV spot.

Only the inclusion of minutes with statistically significant increases in TV attribution is crucial. Otherwise, TV spots on smaller (in terms of range and GRP) TV channels with a higher frequency are very likely to be overrated. Users with high conversion and purchase probability tend to have more visits to the advertiser's website per se, e.g. B. because they research details about the product. A TV ad that is broadcast 10 times an hour on a small TV station is more likely to be shown in the minutes immediately before a buying user visits than a TV ad that is broadcast once per hour on a major TV station will be shown. Failure to take into account the significance of the peak can result in too many conversions being credited to a small TV broadcaster.

It is also important not to set a static baseline; z. B. by calculating the median. Or, even more seriously, the mean value of the number of visitors in the period or month before the TV broadcasts begin. This could lead to the success of TV advertising being significantly overrated. TV advertising not only has a direct impact on the number of advertiser's website visitors after the TV commercial, but it also increases the total number of visitors. Failure to take this into account would overestimate the efficiency of a single TV commercial.

An example: A TV campaign started in February and Spot A will be broadcast on March 10th at 3pm; there is a significant increase resulting in 1200 visits at 3:01 pm. The average number of visits before the TV campaign started in February was 500 visits per minute. The dynamic baseline is 800 visits at 2:59 p.m. The specified median method would measure 700 increasing visits, while the dynamic approach would assign 400 visits to this point in time.

How does attribution work for overlapping TV commercials?

In the case of TV spots that overlap, the so-called peak or uplift pattern of each individual TV spot is taken into account in the attribution. So if this appears without overlapping with another TV spot, this is used for a weighted distribution of the increase between the TV spots. An example: Spot A, which is broadcast at 3 p.m., then leads to a significant increase in the number of visitors over the next few minutes. Spot B, which aired at 3:03 p.m., also resulted in a significant increase for the next few minutes. The peak and uplift patterns for Spot A and Spot B, if these are broadcast without overlap, are shown in the following table:

Table 1 - Average peak or uplift per minute after the TV commercial was broadcast

Table 1 Example: In minute 4, the increase in visitors after TV spot A was broadcast is on average 9.If the starting value is 20 visits per minute, the average over all previous broadcasts of spot A is 200 (= (uplift + 1) * Baseline) visits per minute.

Table 2 - Attribution of visitor and attendance probability for overlapping TV spots

An example based on Table 2:

  • At 3:04 p.m., minute 5 after TV spot A was broadcast (average uplift of 8, see Table 1)
  • and minute 2 after TV spot B was broadcast (average uplift of 9),
  • there are 155-20 = 135 additional visits,
  • of which 8 / (8 + 9) = 47% (64), which can be attributed to TV commercial A.
  • and 9 / (8 + 9) = 53% (71) on TV commercial B.
  • This leads to a 41% (= 64/135) probability that every user visits the website at 3:04 p.m. due to TV spot A (or 46% for spot B).

This results in an estimate of how likely it is that a visitor has seen the TV ad and visited the website as a result. We can use this probability in two ways, which are explained in more detail in the following section.

The isolated TV attribution

The first approach involves estimating the impact of the TV spots in isolation, without taking into account other marketing touchpoints. The majority of TV analytics providers, TV broadcasters and agencies also follow this approach. The problem here is that this method harbors the risk of overestimating the advertising effectiveness of the TV spot, as further marketing touchpoints are not taken into account.

In order to estimate the number of conversions that should be assigned to a specific TV spot in this way, all users who visit the website after this TV spot in the minutes with significant uplifts are flagged as belonging to this TV spot cohort . If a user from this cohort orders a product, this order is assigned to this TV commercial, weighted with the probability that this user actually came to the website as a result of this TV commercial.

An example based on the data from Table 2: Assume that out of all 155 users who visited the website at 3:05 p.m., 10 users place an order a little later. Then 2.5 (25%) of these orders would be allocated to TV spot A and 6.3 to TV spot B. This also applies to all minutes with significant uplifts after TV spot A has been broadcast in order to calculate the attribution of the overall conversion for TV spot A. If a user visited the website several times after the TV commercial - first on a Monday, then again on Tuesday - the value assignment would be divided accordingly. Based on the uplift weights of each individual TV commercial when broadcast without overlap with other TV commercials.

Integrated and cross-channel TV attribution

The second approach is to integrate the potential TV advertising touchpoints at the user level into the analysis of the user journey and into the attribution modeling. This approach enables the success of TV advertising to be assessed in a holistic, cross-channel context. Through the calculations described above, we know for every user who visits the advertiser's website how high the probability is that this can be attributed to the TV ad being watched.

Referring to the above example from Table 2, let's assume that a user visited the website at 3:03 pm. There is a 45% probability that the visit can be traced back to spot A and a 40% probability to spot B. These two TV advertisements are integrated into the user journey of this user. If that user had another paid search click, added something to the shopping cart, and then bought the product right after clicking on a retargeting campaign, their specific user journey would look like this:









User journey with two potential and weighted TV ad views

The integration of TV spot touchpoints in the user journey analysis provides insights into which other marketing touchpoints are part of the customer journey that can be influenced by TV campaigns. For example, it is possible to understand in which areas the expenditure for digital marketing campaigns has to be increased so that the traffic of a certain television station results in conversions. Adtriba customers make it possible to analyze the interaction of TV advertising and digital marketing channels on the user journey and detailed level.

The combination of User Level TV Attribution with CLV Attribution is particularly interesting for advertisers. In this way, not only the influence of TV spots on first orders, but also on all subsequent ones, i.e. the customer lifetime value, can be assessed. An example: Let us assume that a television station has so far successfully attracted first-time buyers, but not users who buy repeatedly. With the help of CLV Attribution, however, its competitive position improves compared to a TV broadcaster that attracts fewer new customers, but generates higher CLVs through better acquisition of repeat buyers. Only CLV-based TV attribution enables holistic performance optimization of TV ads.

The following diagrams from the Adtriba dashboard are examples of the type of analysis that this integrated approach makes possible:

Adtriba Performance Report including TV as a channel in comparison with other marketing channels

Adtriba Customer Journey Analysis: User Journeys with a TV Ad View as the first interaction

Adtriba Funnel Analysis: In which part of the customer journey TV ad views take place

Adtriba offers two options for analyzing the performance of a TV advertising campaign; the isolated and integrated TV attribution. In principle, the integrated approach should be preferred. In particular for comparison with existing tools or with what TV agencies or broadcasters enable, we also offer the isolated perspective.

Would you like to learn more about how Adtriba integrates TV advertising into user journey modeling? Then we look forward to hearing from you at: [email protected] You are also welcome to book a demo in which we will answer your questions: