What is Marketing Attribution? Definition and Types of Models

  • sree 
marketing attribution a general overview

Table of contents

What is marketing attribution?

According to Wikipedia, in marketing attribution, or marketing attribution modelling is the identification of a set of user actions towards a goal, or a conversion, which we’ll refer to as touchpoints and then the assignment of a value to each of these touchpoints. The goal of marketing attribution is to get insights into what touchpoint or combination of touchpoints influences the individual towards a goal completion or conversion.

Why attribution in marketing?

Organizations which invest money into marketing activities, be it into paid media or own media, want to know exactly what working well and what is not working well. Knowing the performance of their individual activities helps them to stop wasting time and money on what does not work well and focus their attention on what works. In the history of marketing decisions were often done based on intuition and personal experience, which leads to a lot of money plainly wasted. 

McKinsey & Co. analysts … showed a typical range of 15% to 20% of marketing budgets could be reinvested in other activities or returned to the bottom line without losing marketing ROI … $200 billion of marketing spent annually could be put to better use

Source: “Smart Analytics can tap up to 20% of lost ROI

To put it short as to why attribution modelling is simple: To stop wasting money and outsmart your competition in decision making.

Benefits of marketing attribution

Benefits of doing marketing attribution right are multiple. When multitouch attribution is being used all the marketing channels and touchpoints along the customer journeys get attributed conversion value or revenue. This means that there is one metric or KPI one can use to assess the importance and performance of every channel.

This is quite an important thing because it means marketers can stop looking at CTR’s (Click trough rates), bounce rates and engagement metrics etc. etc.

Combining the attributed conversions with how much was paid for those conversions gives the marketers a ROI. One metric they can use to assess the performance of every channel. This makes optimisiations and budget allocation so much easier.

Attribution in marketing: different models

In marketing attribution, there are four types of models we are going to cover, single touchpoint models, multi-touchpoint rule-based models, algorithmic or data-driven models and econometric models. All of them provide different insights. In general, the goal when choosing the model is to model historical data in a way which is closest to mimicking reality. This helps to increase the probability of estimating future outcomes in the most economical way.

 

  • Single touchpoint marketing attribution

    • Last-touch or last-click
      The last-touch or last-click model gives all credit for a touchpoint to the last touchpoint or click a user has before a conversion event. This is by far the most used model as it comes as default with most advertising and analytics platforms. It leaves up all activity which happens in the upper funnel by nature of model it gives performance intent-driven channels such as paid search and organic search a heavy weighting. It often leads to wrong decisions (Adidas case).
    • First-touch or first-click
      This model gives all credit to the first touchpoint on the customer journey. The nature of this model gives credit to branding or upper funnel activities which occur at the beginning of customer journeys.
  • Multi-touchpoint rule-based attribution

    • Linear-touch model
      The linear touch model gives the same credit to each touchpoint in the conversion journey. It’s best to visualise this with an example. A customer journey has 5 touchpoints: 1. paid social, 2. display, 3. email, 4. display, 5. organic search. In this case, each of the channels gets 20% of the conversion attributed. Advertisers optimising on linear touch models often also include impressions and not just clicks into the model.
    • W-shape model
      The W-shape model attributes 40% of the conversion to the first touchpoint, 40% of the conversion to the last-touch and 20% to the touchpoints in the middle.
  • Algorithmic or data-driven modelling

    • Algorithmic or data-driven models allocate credit to each touchpoint based on probability theory. They simulate the impact of the removal of a touchpoint in customer journeys. Based on this the touchpoints get conversion credits attributed. There are two commonly used models in data-driven marketing attribution modelling, the Markov model and the Shapley value model.

  • Econometric models or media-mix modelling

    • For channels which do not have touch-points available from either digital or CRM channels such as TV, radio, billboards and print advertising, the impact is usually measured using a so-called top-down model. Here often linear regressions and ad-stock models are used. This is also best explained with an example: Acme corporation is in peak season and decides to run TV and radio advertisements. The marketing team would like to know which channels and stations are the most effective. They decide to use an econometric model. From the digital and CRM side, they have the conversion journeys and from the offline advertising side, they have spot-plans which will tell them at what time of day on which channel how much money is spent. They spend $15’000 on a Wednesday night at 9.00 pm. Based on historical data they know the base-line of conversions and sessions on their website around this time, so they can attribute the uplift in sessions and the resulting conversions to TV and start understanding what time of day at what weekday is the most efficient.

What is a touchpoint in marketing attribution?

A touchpoint is an event a user or prospective customer towards the expected outcome (see Goals and conversions). Here are the most typical touch-points which are measured by marketing organisations:

  • Digital Touchpoints – measured in impressions and clicks

    • Paid search, also known as SEM or SEA
    • Organic search, also known as SEA
    • Paid social
    • Organic social
    • E-Mail marketing, also known as EDM
    • Display advertising
    • Video advertising
    • Direct traffic
  • Offline touchpoints – measured using econometric models

    • Radio
    • TV
    • Billboards
    • Magazines
  • CRM touchpoints (only where applicable)

    • Tradeshows and Events
    • Sales meetings
    • Webinars

Customer journeys, also known as the path to conversion or conversion journeys, often involve multiple touchpoints in various sequences. The higher the product complexity or price, the higher the average of touchpoints is.

Goals and conversions

Measuring the efficiency of marketing and advertising requires a definition of goals towards one works to. In a sales environment, the goal is usually clear: Revenue. On the other side, in marketing, the goals are often not so clear. In marketing organisations often work with conversions goals and measure themselves against these goals. There are different stages of readiness in measuring outcomes.

  1. Traffic

    Here the goal is to get traffic to the website. The outcome is measured in clicks and sessions. We will not cover this in great depth in this article, as most organizations already understand that this is not an ideal form of measuring performance in a marketing department.

  2. Cost per click or click-through rates (CPC/CTR)

    As the title suggests here the measurement is a cost for a click and the click-through rate from someone seeing an advertisement or content to actually clicking on it.
    The formula here is very simple:

    CPC: cost / clicks
    CTR: clicks / impressions

    As for traffic, we will not cover CPC and CTR too much here in this article as clicks are a nice thing to have but they do not actually translate into business value unless you are a publisher selling advertising on your website.

  3. Conversions

    Conversions are what many organisations nowadays – with the exception of businesses which transact directly on the website as for example e-commerce and hospitality companies are doing – are optimising on. A conversion in most cases is a form fill. The metric which organisations on this stage of readiness are looking at improving is called cost per action (CPA), in some cases, it’s called cost per lead (CPL).
    The formula here is very simple too:

    CPA: cost / actions

    Now, this is where it gets interesting: Organisations optimising their CPA know how much each action costs and can act on this information to bring the future CPA down. It, however, does not give any indication about the quality of the lead. In a very typical example:
    Acme Corporation is looking at generating leads for their sales team and of course, the number of leads is not high enough. To help the sales team, the marketing team looks at which channels bring a scalable flow of leads and optimise towards more leads. Now the number of leads increases, but the sales team is still unhappy because many of the additional leads don’t convert into opportunities and subsequently into a business outcome. Now the marketing team spent a considerable amount of money and didn’t solve the problem. One further way of optimising is looking at their website and looking at the different conversion types they have: Online chat information submission, Lead form fill, General enquiry form. Now they look at adding additional dimensions to the data anlaysis: The use historical data to understand the probability of conversion type to opportunity conversion and they also use the average deal size across all deals and apply the following formula.

    Average deal value * probability of conversion in percent

    They do this per goal type and get what is called a weighted cost per acquisition or wCPA.
    Optimising the budget on a wCPA helps organisations to increase the quality and quantity of the desired outcomes

  4. Return on investment, ROI or ROAS

    The most advanced and business outcome-focused way of measuring performance in marketing is a return on investment based.

    The formula here is very simple:

    Marketing spend / revenue

    For companies which transact directly online, this way of measuring outcomes is the most popular as it is crystal clear. If the overall revenues outweigh the expenditures, the business is profitable. Businesses which don’t transact online often have a challenge similar to what was explained in 3) Conversions and the closest they can get with their optimisations are still based on guesstimates. Now if we jump back into our example of Acme Corporation:

    To get to stage 4 their next step would be to link up their marketing data with their sales data. In order to do this they need to link up their advertising and analytics data with the CRM. The way they do so is by connecting both systems using unique identifiers to understand the touchpoints a user has online on the website to their CRM system. Now the setup is complete and marketing costs can be linked to sales revenue. Thanks to this the organisation can now measure the monetary value marketing generates and also optimise it. The executive management can now measure the cost of revenue across the organisation on a very granular level and allocate resources wherever needed.

Pitfalls and common mistakes in marketing attribution

With so much data and options at hand organisations often drink from data firehoses and fail to look at the data in the right way which ultimately leads to inefficiencies.

Data firehose in marketing attribution

Organisations are increasingly drinking from a data firehose

Some of the biggest pitfalls in the world of marketing attribution modelling are:

  • Data silos

    As mentioned in 3) Conversions, businesses with data in multiple places often fail to evaluate their marketing expenditures properly. Some examples here are:

    • E-commerce companies seem to have a great return on investment on certain campaigns but the overall business results are quite bad as these campaigns actually attract customers which return the products, which ends up costing the company advertising costs, shipping fees and a product which cannot be sold again. Not knowing this in certain countries can have a critical impact on the business.
    • Businesses optimising on CPA instead of ROI: Bringing down the CPA and increasing the number of leads might not necessarily impact the business in a positive way at all as quantity and cost don’t tell us anything about quality. Only by connecting the data between the different silos using a unique identifier an organisation can understand the true cost per revenue.
  • Ad platform marketing attribution modelling

    Most large publishers nowadays offer their own set of marketing attribution modelling software for optimisation. While they are definitely helpful for intra-channel optimisation, there are a few pitfalls to be aware of:

    • Cross-channel allocations
      Ad platforms don’t provide advice on how to reallocate budget across channels as they are only focussing on their own inventory. 
    • Double counting
      Ad platforms count conversions using their own pixel which leads to conversions being double-counted. A simple example here: A user sees an ad on paid social media and then goes onto a paid search platform. In this case, both platforms would take credit for the conversion (insert 520k wasted blog).
  • Impressions vs clicks

Historically advertisers were able to track both impressions and click journeys across all channels. With certain publishers walled garden approach, it has become impossible to track impressions. So while some channels support impression tracking and others don’t it might look that channels are actually underperforming while they are clearly not. Here the only way, for now, is to revert back to clicks for all channels when looking at journeys as otherwise, it is impossible to have an apple to apple comparison between them.

Wrapping it up

  1. Get all the data into one place before making any decision: With data silos in place, critical decisions are made on incomplete sets of data. If data is the oil of the future, you will need to look at your data pipelines now.
  2. If you are currently working with CPA metrics, start planning for the ROI age: Not only will it help you to get radical transparency, it will also save you a lot of money.
  3. Don’t rely on last-touch attribution modelling data: Marketing organisations which want to understand the impact of their work to the bottom line of the business definitely should look stay away from last-click attribution modelling.

You might also be interested in…