Well thought-through mul­tichan­nel marketing is often the basis for success in online business. Online shops must find the ideal mix in order to have a chance against the com­pet­i­tion in e-commerce. Each channel has its own power. During the customer journey, there are often many different contact points where ad­vert­ise­ments are displayed so the customer can become more familiar with the product or brand. A marketing strategy that is based on a tailor-made at­tri­bu­tion model should therefore be par­tic­u­larly promising.

Defin­i­tion

At­tri­bu­tion models are an online marketing in­stru­ment that makes it possible to derive both forecasts and re­com­mend­a­tions for future marketing measures from previous results. The principle of at­tri­bu­tion follows the approach of assigning con­ver­sions or sales to certain channels of the entire con­ver­sion/sales process in order to determine their concrete value for marketing.

The right at­tri­bu­tion model should help to allocate marketing budgets as best as possible as well as increase the Return on In­vest­ment (ROI). How this works exactly and how the different models differ from each other is explained in this article.

What is an at­tri­bu­tion model?

Most companies rely on a colourful marketing mix of email marketing, SEA (search engine ad­vert­ising), SEO (search engine op­tim­isa­tion), affiliate links, and social media. The focus is on the target group and the budget is divided ac­cord­ingly. It is clear that, as a marketer, you want to get as much out of every channel as possible - at­tri­bu­tion modelling aims to make this easier. In addition to the customer journey as an overall picture, at­tri­bu­tion models can also be used to record and analyse the various contact points (“touch­points”) where the customer meets the brand. In the best-case scenario, the at­tri­bu­tion models can provide in­form­a­tion about which touch­points were decisive for the purchase decision and what share of the turnover the in­di­vidu­al measures had.

Note

The touch­points of a customer journey differ from brand to brand, and from campaign to campaign. The following example shows a customer journey with social media marketing, re­tar­get­ing display ad­vert­ising and SEO/AdWords (SEA) contact points:

A user ...

- is made aware of a product by a sponsored post on a social network,

- then gets shown an ad due to re­tar­get­ing,

- then searches Google for in­form­a­tion on the man­u­fac­turer and product,

- is then taken to the shop after clicking on an organic hit or an ad­vert­ise­ment, where they finally buy a product and become a customer. (Con­ver­sion).

What types of at­tri­bu­tion models are there?

Different values are assigned to different contact points depending on the at­tri­bu­tion model. The models help to determine which channels are the most important and therefore which need to be pri­or­it­ised.

A monetary value is assigned to each channel depending on its share in results. The weighting then de­term­ines how the marketing budget should be dis­trib­uted. Below are some at­tri­bu­tion models from common marketing practice:

First-click at­tri­bu­tion

With this classic at­tri­bu­tion model, 100 percent of the con­ver­sion value is assigned to the first click or channel that the customer has in­ter­ac­ted with.

Pro: Best suited for campaigns that aim to increase the pop­ular­ity of a brand or product

Contra: Offers little scope for op­tim­isa­tion and little insight into the impact on con­ver­sions and revenue; dis­reg­ards all other channels and ad­vert­ising means

Last-click at­tri­bu­tion

The last click or channel is the most important contact point and is assigned 100 percent of the total con­ver­sion value.

  • Pro: Useful tool for marketers who are only in­ter­ested in con­ver­sions and therefore do not need to take non-con­vert­ing actions into account
  • Contra: Doesn’t consider whether other channels and ad­vert­ising means have played a role in the customer journey before the con­ver­sion

Linear at­tri­bu­tion

With the linear at­tri­bu­tion model, the pro­por­tion of con­ver­sion and turnover is assigned equally to all channels of the customer journey.

  • Pro: The easiest way to analyse marketing campaigns with a mul­tichan­nel approach; ideal for op­tim­ising the entire customer journey
  • Contra: All touch­points are assigned the same value, even if their share in the con­ver­sion varies greatly. There is hardly any op­por­tun­ity to optimise specific areas in a targeted manner

Time decay at­tri­bu­tion

In the time decay model, which is based on the concept of ex­po­nen­tial growth, the time de­term­ines the value of the channel. The closer a channel or a contact point is to the time of con­ver­sion, the higher its monetary share is. The half-life typically measures seven days - a touch­point seven days before the con­ver­sion therefore receives half the value that a touch­point receives on the day of the con­ver­sion itself.

  • Pro: Captures the im­port­ance of each in­ter­ac­tion on the customer journey, but gives greater weight to the touch­points that actually con­trib­uted to the con­ver­sion
  • Contra: Too strongly tailored to con­ver­sion op­tim­isa­tion; in­flu­en­tial touch­points at the beginning of the customer journey receive very little attention if the sales cycle includes a long decision phase

Position-based at­tri­bu­tion

This at­tri­bu­tion model is a mixture of first and last-click at­tri­bu­tion and is often referred to as the bathtub model or u-shaped model. Here, the first and last contact points are rated higher than the others. Both the beginning and the end of the purchase process are at­trib­uted fixed monetary shares (40 percent each by default). The stages in between share the rest of the con­ver­sion value.

  • Pro: Ensures that all touch­points are taken into account, but focuses on op­tim­ising the first and last in­ter­ac­tion; weighting can be in­di­vidu­ally adjusted
  • Contra: There is a risk that two very low-value touch points could be rated too highly

Data-driven at­tri­bu­tion

Data-driven at­tri­bu­tion models should map the customers’ behaviour as ac­cur­ately and in as much detail as possible. In­tel­li­gent al­gorithms process the collected customer data in real-time in order to determine the value of the in­di­vidu­al touch­points and to adjust them regularly.

  • Pro: Very dynamic model that in­cor­por­ates all de­vel­op­ments in the customer journey and regularly adjusts the weight­ings of in­di­vidu­al in­ter­ac­tions
  • Contra: requires a very large database; no real control over the hierarchy of touch­points
Note

It’s also possible to define your own model for your company, which is specially adapted to your company or campaign ob­ject­ives and is not con­trolled by an algorithm. With this custom at­tri­bu­tion, you define the weighting of the in­di­vidu­al channels in­di­vidu­ally - according to various factors such as position, time de­pend­ency, or existing traffic. You can also define different at­tri­bu­tion models for different customer groups (new customers vs. existing customers) or product groups.

At­tri­bu­tion modelling: Example of the models in use

The preceding sections have explained what at­tri­bu­tion modelling is and what types of models exist. But how are they used in practice to optimise marketing and campaigns? To answer this question, the at­tri­bu­tion models presented above are applied to the following example scenario. The search terms entered by the user represent the recorded touch­points:

For the in­di­vidu­al at­tri­bu­tion models, the re­spect­ive weighting of the four keywords is as follows:

  • First-click at­tri­bu­tion: The first in­ter­ac­tion is the decisive criterion, which is why the con­ver­sion is at­trib­uted 100 percent to the first keyword “overnight stay sheffield”.
  • Last-click at­tri­bu­tion: The last in­ter­ac­tion is the decisive criterion, which is why the last keyword “4 star hotel sheffield linda” gets 100 percent of the con­ver­sion value.
  • Linear at­tri­bu­tion: All four keywords have exactly the same share - i.e. 25 percent - of the con­ver­sion.
  • Time decay at­tri­bu­tion: The weighting of the four keywords increases in order of their input. While “overnight stay sheffield” has hardly any part in the con­ver­sion, the time decay model at­trib­utes the main part to the last search phrase “4 star hotel sheffield linda”.
  • Position-based at­tri­bu­tion: If you select the standard rate for the position-based model, 40 percent of the con­ver­sion goes back to “overnight stay sheffield” and “4 star hotel sheffield linda” re­spect­ively. “hotel sheffield” and “4 star hotel sheffield”, on the other hand, each have a 10 percent share in the con­ver­sion.
  • Data-driven at­tri­bu­tion: Since the dis­tri­bu­tions in this at­tri­bu­tion model are in­di­vidu­ally created and can change con­tinu­ously, no concrete or relevant values can be named here.
Note

At­tri­bu­tion modelling based on keywords (like the example here shows) is offered by Google Ads by default: So if you advertise with Google’s ad service, you don't need any ad­di­tion­al tools to apply these models to your campaigns. The exception is in­ter­ac­tions with display network ads for which at­tri­bu­tion models can’t be used.

What ad­vant­ages do at­tri­bu­tion models offer?

Many marketers con­cen­trate on the last click or customer in­ter­ac­tion when analysing their own campaigns. The channels used before this point and the actions carried out during the customer journey and their influence on the con­ver­sions are therefore often not suf­fi­ciently ac­know­ledged during the analysis. At­tri­bu­tion modelling offers a smart way out of this one-di­men­sion­al­ity and has the following ad­vant­ages, among others:

  • At­tri­bu­tion modelling provides deeper insights into the in­di­vidu­al check­points of the customer journey than many tra­di­tion­al forms of analysis are able to.
  • At­tri­bu­tion models enable the op­tim­isa­tion of the entire customer journey from the first to the last in­ter­ac­tion.
  • At­tri­bu­tion models cover the entire range of pos­sib­il­it­ies between manually specified and highly automated campaign mon­it­or­ing and op­tim­isa­tion.
  • At­tri­bu­tion helps in the long run to dis­trib­ute the available budget optimally or to invest primarily in the most powerful marketing channels.

The models enable a higher data accuracy and, thanks to mul­tichan­nel tracking, in­ter­fer­ing factors are minimised such as cookie stuffing (when a user receives a third-party cookie from a website unrelated to that that they just visited).

Go to Main Menu