Four Stages of Attribution Modelling in 2022

challenges of the Attribution Modelling implementation

At Windsor.ai, our team managed to talk with many leaders in Digital Marketing, whose companies are at different stages of data-driven decision making and implementation of Attribution Modelling. I would like to share our findings and the key challenges of four categories that have been observed:

  1. We don’t use attribution but we should get started ASAP

This group accounts to 45% of the marketers we had discussions in with in 2017

Marketers in this category often work in Financial Services (Credit Cards, Insurance), Automotive, or Consumer Electronics, especially in companies which have been around for a long time and traditionally have had healthy margins. Often the media-buy here is distributed across multiple agencies

The key challenges of this group of marketers are:

1) Internal alignment and alignment with agency is perceived to be a showstopper (elaborate on how this affects end goal, likewise for the other challenges)

2) Technical marketing analytics-skill set in-house

3) Ownership of data across different teams (Acquisition, CRM, …)

4) Legacy systems in place

Interestingly >90% of the larger organizations have a proper tagging and tracking system in place, meaning they sit on untapped gold mines. Analytics have been integrated by most of these companies and have collected data for a long time. This makes it very easy to move forward much faster than expected when getting started with attribution.

  1. We use Last-Click Attribution

This group accounts to 45% of the marketers we had discussions in with in 2017

Marketers in this category can often be found in Hospitality, eCommerce, Travel and some FMCG companies with a mid-range margin. Often, the companies start moving a part of the media-buy in-house, have full transparency over the numbers and work with agencies to do the programmatic media buys.

The key challenges of this group of marketers are:

1) Understanding how social and upper funnel activities affect conversions

2) Measuring the impact of impressions/views from mid-funnel activities

3) Uncertainty of media-spend allocation especially as it is heavily biased towards the SEM/SEA side

4) Integration of programmatic data into the conversion journey

The fact that last-click modelling can be applied means that the tagging and tracking is properly set-up. The next logical step of multi-touch attribution (MTA) is easily achievable. With common CRM and marketing automation systems like Salesforce and Marketo in place it would even be simple to do online to offline attribution modelling using the actual sales revenues.

  1. We use an independent attribution software

This group accounts to 5% of the marketers we had discussions in with in 2017

Marketers in this category are mostly found in Europe and North America and most of them are either agency professionals which have clients with a significant online ad-spend or companies which hired a team of ex-agency professionals to move the media-buy totally in-house. Marketers in this category are mostly running “always-on campaigns” with a significant spend and usually the margins in their industries are low and the volume is very high.

The key challenges of this group of marketers are:

1) The investment it takes internally and externally with adding tracking and tagging on all these pages repeatedly is very high

2) Global head-office purchased a nice multi-touch attribution solution, but no-one actually knows how to use it in the region and the support desk is somewhere far away.

3) The model used is a black-box and we don’t exactly know based on what criteria it analyzes the performance of channels.

Nowadays tracking and tagging technologies offered by existing solutions currently available are installed in all of the cases helping to remove complexities around additional tracking when moving to this stage. Challenge 2 and 3 are again dependent on choice of vendor and location of the marketing analytics team.

  1. We have an agency doing custom modelling for us or have a data-science team doing our own models

This group accounts to 5% of the marketers we had discussions in with in 2017

Marketers in this category are either the top ecommerce spenders having a large in-house marketing analytics team or advertisers which run large multi-million dollar campaigns within a certain time window.

The biggest challenges we see here are:

1) The data is modelled done on a project basis, additional costs are required to re-run and tweak the model

2) The data-scientist which built the models to run the attribution leaves the company and the replacement wants to build a new model or is not familiar with the code

3) The management tells the data-scientist to change the model frequently, making it hard to measure impact over time

Marketers in this category were at one point confronted with a make or buy decision. Reasons for choosing make might be that the technology stack is built in-house which makes it hard to integrate an attribution software relying on standard API’s or just simply that at the time when the decision was made, there was no attribution solution out there which would be able to do modelling to an extent which was expected. 

Attribution Modelling – Our outlook

One thing was consistent across all the marketers we talked to: The desire to become more digital and to squeeze more revenue out of the marketing investment.

Of course, the bigger it is, the more likely the C-Suite is looking at marketing ROI. We see a huge potential for marketers from category 1 and category 2 to move into category 3 and 4 in the near future.

Being able to significantly influence and steer the movement of sales in a predictable way by adjusting and redistributing the marketing budget across channels and campaigns is the dream of every marketer.

 

You might also Like to read:

Shopify CLV and net revenue attribution

Post-View and Post-Click in Attribution: An Overview

Data driven attribution models