The presentation and notes from NORDJYSKE Mediers case presented at Email Marketing Evolved - 2015EME. Click NOTES to view my comments for each slide; my verbal presentation. Some slides have been excluded...
2. NORDJYSKE Medier is…
Founded january 2nd 1767.
2 daily newspapers
27 weekly newspapers
2 radio channels
1 TV-channel
100+ websites
Turnover per Year: appr. 1.2 billion dkr.
No. of employees: appr. 2.400
First to integrate radio, TV, print, web and mobile services in
one combined content-production… in the world!
3. Some volume statistics…
NORDJYSKE Medier have:
• 50+ Newsletters
• Several flows using triggermails and Marketing Automation
• More than 1.000.000 unique and active mail-addresses
• We have more than 84.2 mio. pageviews every month on +100 websites…
• … E-mails generate appr. 8-12% of the traffic
8. Psychographics
• We have implemented Cxense on all major websites
• Cxense map users interests through a (pre-)defined taxonomy
ie. ‘Denmark beat Sweden 2-0’
- Keywords: Football, Denmark, Sweden, Zlatan, Lord Bendtner… etc.
• These keywords is rated on each user, and can be queried
What if we dont have what they want?
9. The paradox in lead-rating taxonomy
10 exposures on keyword ‘Football’ 2 exposures on keyword ‘Giraffe’
2 were read 2 were read
Hitrate = 20% Hitrate = 100%
Should we increase our volume in stories about giraffes,
instead of football?
10. Map content themes from market, not content
- Cars Lamborghini or Lada? Insurance or Racing?
- Animals Dogs or cats? LOL or animal rights?
- Fitness Running or boxing? News or equipment?
- Home Indoor or outdoor? Economy or soft news?
- Sport Handball or Football Where?
Automated taxonomies – in my experience – work best…
12. Content independency
Permission
Channel
• NORDJYSKE
• JSL
• Jubii
• Bazoom
• …
Demographics /
Psychographics
• Age
• Sex
• Location
• Interests
Content
providers
• NORDJYSKE
• Vi med Hund
• Jubii
• BilGalleri
• …
13. Example
- NORDJYSKE newsletter have a defined format and profile (constraints)
- Of ‘these available’ stories which should we deliver – Ask Cxense
- Cxense tells us User is interested in dogs…
NORDJYSKE may not have content about dogs… ‘Vi med Hund’ does.
- Deliver suggested ‘Vi med Hund’ content in NORDJYSKE newsletter
seemlessly (if the story fits in the defined format and profile - brand).
16. Permission complexity…
- Marketing laws (DK: Markedsføringsloven §6)
- Data Protection Directive (DK: Persondataloven §6)
- Cookie regulative (consent NOT implied)
A user could revoke one permission, and not the other…
What is our fallback?
This is my presentation of NORDJYSKE Medier as a case for 2015EME – Email Marketing Evolved…
NORDJYSKE Medier have a long legacy of print-media and have a large monopoly in northern Jutland. Moving into the digital age, it became quite apparent that digital news-readers is not that loyal to local media-enterprises anymore, why we in recent times have invested a lot in digital businesses that covers all of Denmark. We have also invested in businesses with a reach into Scandinavia and the rest of Europe, namely Bazoom and Industry Supply. It is also important to note, that we have an editorial team that are used to deliver content for a lot of different media and platforms – including email.
Regarding email we do a lot! Newsletters, servicemails, various triggermails and Marketing Automation setups. Our setup have become quite complex due to many different needs and business models. Email is a central part of our communication with our customers, and also have a large impact on the traffic generated to our websites. Lets’ look at our basic approach on B2C newsletters, where we disseminate our news-stories…
Jubii have 171.000 recipients on their list. These have mainly been gathered through many years on jubii.dk and the various services we have provided there. When you sign up you get a daily newsletter, that comprises all recent stories. One size fits all. Well, in fact, not all…
It always depend on what is going on ‘now’. If the newsstream does’nt suit you, the bouncerate will off course be higher. Demographics simply does’nt apply when disseminating news. It is all about interest. Some may be interested in politics, others in sports. If this is not matched in the subject-line the chance of an Open is lower.
So… we need to identify peoples interests and match our content with these. Forget demographics – we aim to deliver relevant content for all, no matter Age, Sex or Location! This can be done in a number of ways. The easiest way is to simply ask the users…
In the newsletter we provide a link where the recipienst can change their options. We give them a number of choices of topics they are interested in. This impose another problem.
1 – We rely on recipients to do this. Most don’t. Only 5% change their options, even if we motivate (!) them properly to do it.
2 – Even if they do change their options the news may still not fit. We include more, but not all due to resolution… In this case we allowed recipients to select stories about ‘Animals’ as a chosen topic. But ‘Animals’ as a term is still quite abstract. You may be interested in dogs, but not cats. Or vice versa. Keep in mind, that if we keep adding this resolution, it also becomes an obligation to deliver this content. We may not have enough content… or we would still exclude the few.
User options in this field does’nt work. What people say they want and what they actually want is NOT the same…
People don’t know what they want. They may not want what they say they want. So we should do it for them. We should analyze their behavior online and track the content they look at. NORDJYSKE Medier have 100+ websites. 9 out of 10 northeners are in contact with at least one of our media daily, and we also have a very large reach nationally. In time we could achieve some very precise profiles on what people are interested in. It may sound a bit like ‘Big Brother watching’, but in fact we only strive to deliver relevant and engaging content satisfying each individual recipient and tailoring our media experience.
We chose a system called ‘Cxense’ as our profiling-engine. I will not go into details as to why it was them and not another – there are alternatives, and if you work with B2B email marketing and/or productsales, I would probably suggest other systems to look at. But for our needs, scanning our vast amounts of content and building semantic profiles, it works. The taxonomy register keywords, and these are mirrored on the individual looking at the content. This generates a more or less a basic Lead Rating profile, but with a very large taxonomy. Next time we suggest content for the user, we can take this profile into account selecting the right content. But is build upon logic, not intelligence (like the Viterbi algorithm). This provides a few paradoxes…
One paradox is: What if we don’t have what they want? Should we deliver other current news, or dig back in time finding content of relevance although dated? We never found an answer for that – this is individual too. So this is an ongoing investigation.
Another paradox is this: Cxense don’t measure weight or volume. We dont know the number of articles exposed to the user (volume). If you get exposed to content 10 times, you may click it the tenth time, even if it does’nt really relate to you. It could be a secondary preference, it could be just the fact that we as an authority presents the content to you. We dont know. So in this example using keywords like ‘Football’ and ‘Giraffe’, they would both be measured equally. Both received two hits. It is not taken into account how many was exposed to the user. We as humans can calculate the hitrate and off course say ‘Giraffe’ weighs more. But in the world of automated digital profiling, it seems ‘Football’ is chosen. This is how we qualify and test a profiling algorithm. We dont trust them out-of-the-box. Off course we shouldnt write more content about giraffes – there is so few people interested in this field – but we should make sure, that they get it when we have it.
But when we keep creating more content about football and still present this to the same users, it imposes Bias. We could build constraints but that brings new bias. Instead we try to build logic upon the queries to Cxense and measure the weight of the keyword towards the total number of articles registered with the given keyword. It is nerdy and technical, I know, but I have to address this. If you work with autonomous digital profiling tools you have to validate its output. And there are no tools for this!
Building the taxonomy and resolution of the digital profile you also need to think about the various intents. A high abstract, like ‘Cars’ could mean a number of things. Is the user interested in Lamborghinis or Ladas – definitely not the same thing! But it could also be surrounding topics like Car Insurance or Racing News. The same with animals: You could be a dog lover or a cat person. But maybe we shouldnt present LOLcats to the people very much into animal rights and welfare. Taxonomy needs to be fused to reveal this. How does ‘Cats’ relate to ‘Welfare’ in the taxonomic model? It is not a hierarchy, I can tell you that!
But have you done it well you get a strong tool for analyzing interest that could also identify new marketing possibilities for you. In this example we made a basic demographic model dividing our recipients in gender and age. In that we can query for interest in various clothes. Women quite often buy clothes for their men. Grand parents very often look for clothes for their grand children. We can basically make a pointcloud of each and every recipient and see if there is an interest, a marketing opportunity and build a new newsletter for them. Or other models… We use their own interest profiles and give them exactly what they want, if the point cloud is big enough. This is the rationale of doing digital profiling at all; this is the real value!
Doing digital profiling have also meant, that we have disassambled our content from our mailing-lists. The traditional way have been, that the newspaper have a list, they have content and they just send their own content to their own list. Not anymore. As a company we create content in many different departments and companies. The recipient may have signed up for a list due to the demographic profile, but their psychographic may reveal they want content from another provider.
Our daily newsletter from NORDJYSKE Medier deliver current stories each morning. We ask Cxense about which story should be on top and highlighted in the subject line. If Cxense responds, that the user is interested in dogs and there is no current story about dogs from NORDJYSKE, we could opt for a story from a different content-provider ‘Vi med Hund’ (Translates in to ‘Us with Dogs’). We can take this content directly from the ‘Vi med Hund’ content database and present directly in the NORDJYSKE newsletter the user receives. The constraints we build is, that some stories may not fit the brand and profile of NORDJYSKE Medier. This is evaluated before writing it all into Apsis. This also explains a key reason as to why we chose Apsis. Speed. Agility. And a great API. Not to mention a topclass support-team. It just works and we can do it fast.
BUT… all of this is quite cool. We can build interest profiles with large detail, we can identify and extract content from a lot of different content-providers and present it seemlessly in various newsletters. Technology is no longer the problem. Everybody is talking about customer engagement, identifying relevance and need. But we are way ahead. I truly believe we’re Next to None in this field in our domain at the moment. So imagine how heart-breaking it is it to learn, that it is not legal…
An enduser would never know it. They would just get a better overall experience, better content and tailored media. But it is still illegal. They need to approve of us doing it. Not just the basic permission from marketing laws, that they agree to the newsletter. No. The setup is so complex that various platforms, businesses and databases is all working together to deliver this magnificent experience. And each of them have their own permissions.
The newsletter it self need a permission, that we can email content to the user. But we also need a permission towards the Data Protection Directive, because we want to share information between our own companies – they are different legal entities, even if they have the same owner. Personal data cant be shared by default. And Cxense is a script that stores a cookie on the users computer. This is a third permission that needs to be given. And each permission could be revoked independently of another. On each site… we have +100. This is a lot of permutations of how many versions our newsletters should be delivered. But we’re doing it – we will start getting the proper permissions and implement this in our core setup on the logical touchpoints throughout our company. We have no choice. The penalty for 1.000.000 adresses is severe if we dont! We care and we want to be the spearhead in this field.