Facebook järjesti Tanskassa seminaarin Markkinointimixin mallintamisesta ja attribuutiosta tammikuussa. Seminaarissa käsiteltiin markkinointimixin mallinnuksen kahta päähaastetta, jotka ovat historiadata ja metriikat. Facebook on kehittynyt viime vuosien aikana hurjaa vauhtia, jonka vuoksi historiadatan keruu mallinnuksiin on haasteellista. Toinen haaste on oikean metriikan valinta – mikä on vertailukelpoinen muihin medioihin ja mikä on paras ennustusten rakentamiseen. Facebookin lanseeraaman Partnership ohjelman kautta saamme parempaa dataa ja entistä parempia koulutuksia sekä Facebookin että Instagramin suhteen.
At the end of 2016, Facebook launched the MMM Partner Program, aimed at improving modelling results via co-operation with agencies and advertiser customers. This January, Annalect was invited to participate in Facebook’s Marketing Mix Modelling (MMM) and Attribution seminar in Copenhagen, aimed at informing and inspiring econometric professionals and modelers who are building bridges between the online and the offline.
We sent out a team to find out more – here is a summary of the key points that struck us.
Get in touch with us below to hear more and ensure that your data tools and analytical insights are keeping up with the landscape.
Do you know your reach?
Currently 15.8 million people access Facebook every month in the Nordics and 82% of them return to Facebook every day. Another attention-getting number was presented – 92% access Facebook daily on mobile devices (September 2016). Marketers need to focus on mobile more than ever before. When creating marketing content, a one size fits all approach will not work. The average attention span of a mobile user is 1.7 seconds. Recycling your 30 second TV spot is a waste of money.
In the past, there used to be fewer media channels but they could all achieve high reach levels individually. Now, the number of media channels has grown resulting in the lowering of the reach of any given channel. However, with right combination of media, the total reach can still be kept very high. And this leads us to the main message of this event: measuring cross-media impact is crucial.
Measurement needs to keep up with media
The most important move that an organization can make is to shift from opinions and gut-feeling to fact-based insights grounded in the real world.
Media has been going through revolutionary changes in the past decade, with the pace of change accelerating in the past few years. The impact of traditional print media has begun to wane and social media’s importance grows exponentially. Social media itself keeps evolving, with Facebook itself having undergone numerous changes in the past few years, for example with the introduction of Instagram and Facebook Live.
Measurement techniques need to keep up with the change to remain accurate and relevant.
Key challenges in analyzing Facebook performance: history and metrics
Modeling the performance of Facebook as a part of the media mix suffers from two key challenges. First off, 3-5 years of historical data are needed. It’s hard to even remember how Facebook was used five years ago or what formats were being displayed at that time. There were no carousel ads, no looping videos, and Instagram had only just been launched. In addition, the number of users has exploded since then. How can this type of historical data be properly interpreted? What is the best way to use it in modelling so that useful predictions can be made? There are many ways to tackle this issue, the easiest of which is to split data into groups based on time (on yearly level at least), campaigns, or ad format.
The second issue involves metrics. Almost everything can be and is measured nowadays. But, which are the most relevant predictors for future conversions and comparable with other media? According to a Nielsen Brand Effect analysis, 90% of the people who buy your product never click on your ads. This makes the question between should the modelling metric be clicks or impressions quite irrelevant. It’s only a certain type of people who tend to click a lot whereas (almost) all people in FB get exposed to ads. The effectiveness of paid vs. earned media was showcased by another Nielsen study: 75% of buyers aren’t fans of your product, 24 % are not on FB at all – only 1 % are fans.
From the modelling part, it was great to realize that the discoveries we have made in our modelling cases here at Annalect are also recognized in Facebook as well as in other modelling companies. And on the other hand, the challenges we face are similar for all modelers. Via the Facebook MMM Partner Program we will get better data and more transparency and inside training on Facebook and Instagram to better tackle the data issues.
New solutions for Multi-Touch Attribution
“Attribution is only as good as the data and models being used”.
There are still some problems with multi-touch attribution, e.g. cookies don’t capture cross-device actions. According to an eMarketer study (US 2015), 50% of people use three or more devices and based on a TNS Consumer Barometer study (2015), 41% of people finish an activity with a different device than they used to start it. Also, the choice of model used makes a big difference in results – this defines how credits are allocated for parts of the customer journey (last click, last touch, time decay etc.).
Facebook now offers a few solutions for better attributions. Some of them are still at the beta stage but one current solution that is already in production is FB Pixel, which can be used to pass conversion data from your website to Facebook. The Facebook audience can be split into test and control groups and with the conversion, data lift can be calculated by comparing test and control groups.