After having dedicated two weeks to exploring the analytics tools available for the most commonly used social media platforms and think about how we can best assess whether you met your goals and objectives, we have invited Brad Fry, Director of Strategy Insights (@BradFry) at Folk (@Folkdigital), a story-telling agency based in Poole, to speak about how they measure their social media campaigns. Brad’s work includes using “web analytics and business intelligence tools to gather insights into Folk’s client’s businesses or target markets” which help provide insight that drives and influences strategy.
Brad’s powerpoint is on myBU and I highly recommend that you go revisit it. While very focused on the ROI aspect of online measurement, Brad’s presentation makes a lot of valuable remarks. I’ll highlight them below.
Brad started his talk with a re-cap of what a brand universe encompasses, that is the totality of channels used by a brand to communicate with their target audience(s) and have contact with them. There are over 100 channels at marketers’ disposal (revisit the lectures from the beginning of the semester about the social media landscape) however most companies, and in particular SMEs with which most of you currently work for your projects, use only a limited number of these and they usually include websites, blogs, Facebook, Twitter and email.
The traditional ROI model
Brad spoke about the traditional ROI (return on investment) model that puts emphasis on the conversion rates and transaction averages. His example is below:
If your website has 50,000 visits @ 2% conversion rate = 1,000 sales
1,000 sales @ £50 ATV (average transaction value) = £50,000
While this model is simple and easy to use it is only fit for transactions at the peak of the conversion cycle – that is those actions that lead a user to a purchase. However, a user’s activity online is not always so directly related to a purchase. On the contrary, a user’s interaction with a brand universe can be repeated and in various forms (via Facebook, search, a mention on Twitter), which means that the current model disregards all those intermediary steps which lead to the final conversion. This explains perhaps the reason why customer service and public relations have been struggling in the past to justify their effectiveness, impact and value of their efforts.
The multi-channel attribution model.
So if our 50,000 visits lead to @ 10% signup rate => 5,000 subscribers
10% of subscribers become customers within 12 months => 500 customers
25% of customers become loyal, and purchase 3 times a year
The multi-channel attribution model therefore considers the contributions towards the end target of each channel. It relies heavily on Google Analytics and the way in which conversions are defined by the platform (this is mainly because Google Analytics is free and is usually part of the analytics toolkit).
Google Analytics tracks and reports:
All the activity therefore happening outside Google Analytics (and a lot of it does – such as emails opened, posts on Facebook, content viewed on YouTube and more) is therefore unaccounted for. Therefore, to be able to measure these efforts one should have a brand/communications hub, a presence to which all activities on social media are channeled towards. This is why, for businesses in particular with financial goals, having a website or blog is highly recommended (as they can use Google Analytics to track these elements of conversion).
The multi-channel attribution model operates with various funnels, that is provides more weight (or importance) to one element out of all the user’s interactions with the brand within the brand universe.
Brad’s example was that of a user’s purchase of a dress (a process which takes place over two weeks):
- A shopper reads a post we wrote on Medium about party wear, she bookmarks the page.
- She signs up to our newsletter and likes us on Facebook
- She visits the site three more times from email and Facebook links
- She visits the site after searching for [cocktail dresses]
- She visits the site after clicking a bookmark (last click)
- She buys a dress
The Last Click model
The last click model, in calculating ROI, would consider as relevant (and therefore give credit to) only the final action before the completion of the purchase (the bookmark click, in our example). This means that all other interactions are unaccounted and disregarded.
The First Click Model
The first model on the other hand will credit the action that triggered the brand interactions, therefore the one that enabled the user to discover the brand. In Brad’s example this would be discovery of a post on Medium. While this could be particularly useful for instance of public relations, marketing and advertising campaigns, as Brad very well points out most platform reports back data from the point of purchase up to 90 days. This means that high engagement products (university courses could fit in here as well), this model does not assess the true first click but rather the most-recent first click (in that 90 days spectrum).
The Last Non-Direct Click Model
The Last non-direct click model attributes value to the first before last point of conversion. In Brad’s example this is search (or paid search). Like with the previous models, while this can help calculate the value of one channel, it disregards the input and value of the others.
This leads therefore to
The Linear Attribution Model
This model, to address the challenges of the funnels discussed before, proses to provide an equal share to all brand interactions leading to the purchase. This means that all actions in Brad’s model will be considered to have had an equal influence on the shopper and her purchase decision. While this solves the problem of attribution it also brings the challenge of maintaining a competitive edge between the channels.
The Time Decay Attribution Model
This model provides different wight to all channels depending on the recency they have been interacted with. Therefore, the oldest the interaction, the less the weight and influence towards the final purchase. This is by far a better model than the last click attribution one even if it still provides more importance to the last transaction.
The Position Based Attribution Model
Is a combination between all. It provides more weight to the first and last click and distributes (either equally or otherwise if your strategy requires so) the remaining weight to the intermediate channels. This is something worth considering however, this model too has its challenges, confirmation bias (changing the percentages to reflect your results) being its biggest.
Challenges with measurement
Brad’s final part of the lecture was dedicated to reflecting on the challenges of measurement in particular to those providing a single customer view. With the possibility of accessing content not only via a variety of platforms but also via a variety of devices, tracking a customer’s journey and interaction with the brand becomes very difficult if not impossible. This makes reporting, particularly of unique visitors ineffective since a unique visitor will always be considered via the entry point he/she makes to the platform. So if a user visits the website via their desktop computer and then again from their tablet, it will be reported as two unique visitors rather than one. This is the same with mobile interactions.
How does this apply to DCS?
Brad’s talk was very insightful however the conversion funnels that presented are more extensive than the kind of activities that you undertake part of the DCS group project. The attribution models however (conversion funnels) can also be used in relation to your campaign objectives but you need to consider in that case how much of your work can be captured, monitored and measured that way. Perhaps the time decay and position based models should be the ones you should consider at this point, particularly when thinking about the interplay between the three channels that you use.
As for a general learning point, Brad’s lecture emphasizes how important it is to have an understanding of measurement concept and of the advantages and disadvantages of the tools used. Finally, a clear, logical assessment of what is that you measure linked targets and objectives is essential.