Your LIVErtising 3 stones of knowledge

July 5, 2018

To start you preparing the LIVE exam actively, here is the necessary information about the format and contents of the exam.

  • <May 5 update: add your questions and check answers in the comments at the bottom of this page – a preceding exam post also displays many questions and answers, which you can search with CTRL+F>

YOUR EXAMThree stones of knowledge FORMAT

The contents correspond to the information shared during the lectures, on the basis of the slides – I mean, the oral contents shared during the lectures, not just the slides. The evaluation itself will be based on three concepts from the course. You’ll draw at random three cards from a triple list including:
(1) core concepts (2) supporting concepts (3) illustrating concepts.


The above list is complete for the 2018 exam.


You can still get hold of the student’s notes on our Onedrive.

I understand you may want to ask me questions about some of the above concepts. I invite you to submit all your questions as a comment at the bottom of this post. This will enable me to answer publicly and help you all. In earlier posts about the exam you’ll find my answers to such questions.

Like this? Share it!
Tags: exam,

68 Comments. Leave new

Hello ! Is this correct ?
> why = mission = objectives
> how = vision = strategy = “big” goals
> what = tactics = actions = micro goals
Thank you !


Seems correct, bearing in mind that a word such as “how” is polysemic and can be understood in different ways. OK?


Hello Sir,
According to Simon Sinek, the WHY corresponds to the vision and the HOW corresponds to the mission.
Therefore I’m wondering which one is correct?
Thank you


Isn’t it :
> Why? What’s the purpose = vision = objectives
> How you do what you do = mission = strategy?


Answering Julien and Mathilde, also echoing my answer to Graciela’s earlier question: let’s try not to turn this into a theological debate. The notions of “why”, “how”, “what” are as such ambiguous and can be understood in different ways. I feel Graciela’s structure makes sense, as does Julien’s observation and Mathilde’s also. It all depends how you develop this. What really matters, is that you can detail the difference between working out what your goals are, which is the strategy, measured by defining KPIs, and selecting the tactics that will enable you to implement the strategy, which you’ll measure more with crude metrics, and to a lesser extent with KPIs. This is the essential framework. The way I developed it is considering that to work out the strategy, you’ll find help in defining your vision, mission and values. The way you meet them, the way you put them into practice, the way you transform those vision, mission and values, will drive and constitute your tactics.
I hope this does not confuse you even more. I insist that you needn’t be blocked by the why, how, what as they can be used differently. The safest one is “why”, which is really where you should start from according to Sinek. Even that “why” can be your vision (you’re right, Julien), but you can consider it englobes the whole concept of strategy. Just try to be consistent, bearing the Strategy/Tactics framework in mind. Always!

May 31, 2018 13:52

Hello Sir,

About the concept “Podcasting” I would like to use it in order to illustrate that thanks to voice as an emerging interface, we have observed a huge increase of podcast ads for 2 years now. But I don’t know if it is what you expect from that concept ?
And I have some difficulties to link “Neutral Networks” to the content we’ve seen during this semester..

Thank you for your help!


OK about podcasing – this is actually the come back of podcasting we are witnessing now. It was hyped as an advertising medium some ten years ago, then virtually disappeared except for niche uses, only to deliver new promises now.
Neural networks are one type of deep learning technology. Does that get you started?
Let me know, Anonymous 😉


Hello Mister Ranschaert,

I don’t really understand what the concept « vanity metrics » means.

Can you help me clarify it, please?

Have a nice day


A metric whose value is only to please you, offering some feeling of how good you are (number of visitors to a web page, number of fans, number of followers). Actually, these metrics are useless on their own, because what you are looking for is not visits or fans, but **conversions**. Right?

June 1, 2018 14:24

Thank you Sir !!


Hello again Mr. Ranchaert,
For the concept “the recent evolution of television” I would like to speak about linear, IPTV and OTT. But I am not sure to understand the difference between OTT and IPTV : both use internet, exact? Is the difference with ads (no ad for OTT and ads for IPTV) ? Or is the difference in the installation?
Can you help me ?
Thank you in advance

Anastasia Bousson
May 29, 2018 18:00

I think an IPTV, is a “real” Tv where you have the internet on it and so you have the ability to use your TV as a browser (so you do get ads).

Whereas an OTT (think about Netflix) is opposed to linear television and refers to television content provided via a high speed internet connection rather than cable. And I believe there are no ads on OTT (not so sure).

To me they are opposed : the first one you have internet on your TV and in the second one you have TV on the internet.

I tried to answer your question maybe Mr. Ranschaert can confirm and complete ! 🙂


Hello Anastasia and Roxane,
even if technically and in terms of economic models OTT and IPTV are very different, they both enable the end user to enjoy video content over the Internet. Anastasia’s shortcut is helpful to summarize the difference, because IPTV indeed requires a smart (IP-connected) TV set connected to a (private, operator-owned) IP network, while OTT uses the public internet. As a result, you can get the standard TV offering from your TV operator on your IPTV, while OTT contents (think Netflix, Hulu, Amazon Prime Video) are streamed to your device over the public internet. That device is any screen (a standard television connected to a computer, a computer, a smartphone, a console, a tablet…). Google Chromecast, Apple TV, Amazon Fire TV, Roku stick are different platforms to get OTT streams to those screens. These are the basics.
Now, things can get complicated, for example because you can get Netflix (which is OTT) on your IPTV by installing the Netflix app! Also, one of the components of IPTV, the smart box, can also be added to your computer or console as a virtual box. If all that makes you think there is no longer any difference worth considering between OTT and IPTV, do not forget that they only provide content, which needs to be produced – the distribution of that content rests on very different business models. But this is taking us further than needed for our course. I’m sure some of you will be involved in this professionally, though!
Thanks Anastasia, thanks Roxane! Let me know if this is helpful.


And what about the advertisement on both ?
But yes this is clearer now, thank you!

As IPTV carries the programmes of your television operator, it enables you to get all television advertising in the usual tv programmes of the tv channels offered by that operator (think RTL, TF1, RTBF, … broadcast over VOO or Telenet), while enabling the advantages of digital television (e.g. addressable – or targeted – television ads). OTT (Over the Top) means outside the usual television operator offering, with major actors such as Netflix. As we discussed, Netflix provides an alternative to the usual tandem “free/cheap content supported by advertising”. I do not want to go much deeper into advertising on IPTV, because that is a complete lecture on its own. OK, Roxane?

Camille Gilbert
May 29, 2018 12:32

Hello Mister,
I don’t understand what those different concepts mean and I can’t find them in my notes :
– “fraud detection”
– dashboards
– google trands
– ads tags
could you help me ? thank you.


Oh, by the way, you’re not alone to address me with “Hello Mister” – please remember the use of “mister” requires the name of the person – if you do not mention the name, say “Hello Sir” instead! OK?


Maybe I can help :
– Fraud detection : the goal is to analyze patterns in the practice of click fraud, in order to fight AI. (see click fraud)
– dashboard : it’s a visual display and the purpose of a dashboard is to provide real-time results by aggregating and extracting value from all the data you collect, otherwise known as your key performance indicators (KPIs). It’s also a tool of social listening.
– google trends : a tool from Google Labs to know how often a term was typed into the Google search engine. It is a tool which allows to analyze the popularity of a term on the search engine, for a determined period of time.
– ads tags : a code snippet that has to be inserted within the HTML code of a webpage where an ad is due to be displayed. An ad tag is created and placed for each ad placement on the page.
When the ad tag is uploaded by the user’s browser during a visit on the publisher website, the tag gives instruction to the browser to retrieve the ad from an ad server.


Sorry, I nearly overlooked this!
Fraud detection: carried out by AI to detect fraud (not “to fight AI”). Indeed. Spam detection is another similar use, typical of supervised machine learning involving classification. Here is another example where you can probably not go much deeper into that very concept, but can make use of the conceptual background of this notion.
An example of dashboard is Cyfe. It can actually collect both metrics and KPIs if it is well chosen and well organised.
Google Trends also enable you to compare search volume for two, three or more keywords, for example comparing your brand queries with the competitors’.
Ad tags: exactly!
Thank you Eloise!

Gilbert Camille
May 30, 2018 14:44

Thank you Mister Ranschaert.


Hello Mr Ranschaert,

Concerning the DMP ; I understood that it’s considered as the middleman for all the actors in the programmatic buying process. So for example, if the ATD needs data to deal with the DSP, the DMP will provide those informations. But, could the ATD asks datas about the SSP even if they aren’t from the same side ? And does AI enter in this process ? I can’t imagine such a platform providing so much datas without AI…

Could you clarify this for me once more ?

Thanks you,



Hello François,
not exactly, as a DMP does not do any buying or selling, it only provides and analyses data, it informs the platforms that automate the buying process (DSP) or the selling process (SSP). Both DSPs and SSPs are platforms, they provide media buying or selling interfaces. An ATD on the other hand, is an agency employing people specialising in programmatic media buying for advertisers or advertising agencies, sometimes a specific department within an advertising agency or a media buying agency. The DMP data are used by buyers within the ATD, and the transactions made both through DSP and SSP can be informed by DMP data too. This has a cost, so only big actors are likely to base their media buying/selling on these data.
I understand this is a very complex ecosystem, undergoing constant evolution, with many bits and pieces that can provide different buying processes for different actors.
As we are dealing with automated processes, yes: this is completely algorithmic. You can consider it is AI only – bearing the AI paradox in mind!

May 30, 2018 10:28

Thanks you, this is much clearer now !


I have some questions about concepts that i dont find in my note..
– attitudes towards advertising
– data-driven marcom
– computer vision
– conversational search
– mention
Is it possible to have an explanation?
Thanks in advance


Can I suggest some fellow students give you a few hints first? Anyone?


Hello L,
For attitudes towards advertising, I would talk about the fact that people trust less ads and more recommendations from their friends and family. That’s why it can be interesting to do UCG. It will allow you to engage your audience and to build trust. That leads to influence marketing and the influencers/bloggers.
I would also say that it is important to make your website/content mobile-friendly because mobile is #1 device for the millennials. It is an occasion for the advertiser to get a lot of personal data (more present on the mobile). (//Google ranking)
It is also interesting to use voice in your ads because it is really “in”. Podcasts are really becoming a thing.
There are a lot of things to say about attitudes from customers (much more than what I just said like talking about SEO…)
-data-driven marcom : The adjective data-driven means that in any activity, progress is compelled by data, rather than by intuition or by personal experience. Data driven, also known as Data driven marketing is based on an approach to making strategic decisions based on data analysis and interpretation. The data driven approach allows examination and organization of the data in order to better identify its consumers and customers. Thanks to data, you can also monitor a brand and watch the situation analysis and this SA is the base of any activity. You plan on a basis of a situation analysis than you execute then your measure in order to plan again. Data allows: ad serving, targeting, retargeting, programmatic, monitoring, analyzing and improving your SEO, AI, machine learning and A/B testing.
-Computer vision: the desktop isn’t the first device anymore, it is now the mobile especially for millennials. (not sure about that one at all)
The two other ones I don’t know. Mister, could you confirm what I just said? 🙂


You’re convincing, Sarah, well done too!
I’m not sure, though, SEO is the most immediate answer to customer’s evolving attitudes, until you provide convincing arguments. I suppose you have the mobile-first media usage in mind, with the increase in voice search – this has indeed an impact on SEO.
I invite you to have a look at the other comment to get on your way with the last two questions. This will also correct your approach of the “computer vision” concept.
Let me know if this helps you, Sarah and L., or anybody else!

Maybe I can help for some at my scale:
– Attitudes towards advertising: People expect interaction and participation in brand’s communication, they want to feel concerned and their voice listened. It makes remember the video launched by Microsoft 13 years ago where the consumer wanted divorce with a pushy advertiser where a only one-way communication was possible. Today thanks to the shift from Web 1.0 to Web 2.0 consumers have the ability to take the power on advertisers and communicate, share, participate, leave comments, post reviews and produce their own content. Some advertisers see it as an opportunity (Doritos, Instagram) some others as a threat (McDo). What’s more people trust more recommandations (WOM, reviews,…) more than traditional commercials.
– Computer Vision: As we saw it we’re currently in ANI stage and AI is able to have “human like” behaviors : 1) Physical AI (walk, move, run) 2) Computer vision: to adapt to our environment we must be able to see and look AI is able to do so 3) Speech rec 4) NLP 5) Heuristic Classification
– Conversational search : possible thanks to Google’s Update: Rankbrain (not sure I saw on the Internet that it was actually Hummingbird). This algorithm analyzes different elements (such as geolocation) in the query in order to better understand the question and bring the most relevant answer. For instance if I’m In France and ask who is the president ? the search engine will answer Macron and if I’m in the US it will answer Trump. If I go further and ask who’s his wife, the search engine understands that “his” refers to my previous request and it will be able to answer Melania Trump… this is how a conversational search starts
– Mention : is a media monitoring tool to help businesses grow and protect their brand and reputation


Fine, indeed, thank you Esra.
Let’s just make sure everybody understands the series physical AI – computer vision – speech – NLP – heuristic classification **not** as a time sequence, all that is happening and improving simultaneously. While they are seperate fields, with each AI doing specific activities (playing go or climbing a wall or translating text, …) we are at ANI stage. AGI supposes an integration of all these skills within one AI, exactly like a human person is able to do all of them, at least potentially.

Rankbrain can be considered as a development of Hummingbird, at least for our purposes.

I suggest trying out those tools, e.g. Mention in this question.

May 28, 2018 22:05

Hello, I will not give you long precise definitions but I will try to help by linking them with other concepts of the class.
– Attitude towards advertising: I would link it to the beginning of the course and the introduction video, people want more and more conversation with the brand it is the time of networked communication, there is a shift in power and brand should listen and monitor to create a real conversation with the customers. I would link it with these: “the end of advertising”, “the evolution of the web (1.0 and 2.0)”, “networked communication”, “listening”. But I think you could link it to almost everything because “it helps understand the evolution and specific dimension of marketing that the course focuses on” (to quote M Ranschaert).
– data driven marcom: data-driven means that progress is compelled by data rather than by institution or personal experience. Data driven marketing is based on a approach to making strategic decision based on data analysis and interpretation. So you can also link it to a lot of other concept “science vs creative”, “AI”, algorithmic advertising”, “situation analysis” among others
– Computer vision: is one of the step needed in AI evolution to become AGI (1/physical, 2/computer vision 3/ Speech rec&sync 4/NLP 5/Heuristic). Computer vision is the fact that machine can look and see.
– Conversational search: fact that you can speak to a device and the device can respond. Google made it possible in the Hummingbird update. You can link it with “the voice web”.
– Mention: It is a social media tool that monitors the web and the social media channel to keep you informed everytime somebody mentions your name, brand or target keyword. So you can link that with listening and monitoring

I hope I could help you and I hope I am right with these short descriptions but I am sure M Ranschaert will help us with this.


I have nothing to correct, and not much to add! Well done Aurélie!
Just a few datails I want to add:
– In a data-driven approach, all analyses and decisions are informed by data rather than by intuition or personal experience.
– Computer vision may be used to develop social media intelligence, enabling brand protection for example, or a better understanding of the context of use of a product.
– Conversational search involves AI, i.e. Rankbrain, the sequel of Hummingbird.
Now, answering L., I know you won’t consider this as THE answer, they are indeed convincing elements to build your answer, with other avenues still possible.


Hello again Mr. Ranschaert,
I have some questions for you :
– In my notes, I have explanations about the “Cluetrain Manifesto” but nothing about the “Cluetrain manifesto UPDATE”, is it the same?
– Why is related to the biometrics and our new visual digital identity ?
– I don’t understand what is the growth hacking in the conversion…can you help me?
– And, what is the difference between the private browser and the incognito mode?

Thank you in advance !
Have a nice day.


So, Eloise:
1. Cluetrain
“Markets are conversations” – this was the first thesis of the original Cluetrain Manifesto, published in 1999. The authors were visionary in their analysis and prediction of the state of markets and marketing communication. The whole chapter 2 in this year’s course attempted to present a humble update of those predictions, upgraded to 2018.
In their mode of payment in their supermarkets: your face is your ID is your wallet. Do you remember?
3. Growth hacking
We only touched upon the subject of growth hacking very briefly, to illustrate the mindshift you need to make: the goal of a website is **not** to drive visitors, it is to contribute to business growth. That conversion focus is at the heart of growth hacking too, whose very components are all oriented exclusively towards conversion, as conversion ensures growth. Visitors only generate… visits, which is not sufficient.
4. they are the same, on different browsers – a requirement when you’re working on a website’s SEO.
Is this clear(er)?


Thank you very much! 🙂


Hello Sir,
I am not sure to see the link between private browser/incognito mode and SEO. Could you enlighten me please? Thank you


The search engine you’re using tracks you with a wide number of cookies and fingerprints, in order to customize the answers it returns in the SERP. Accordingly, when you are working on a site’s SEO, you need a more unbiased SERP. This at least requires you to turn off the customization settings of your browser by surfing in incognito mode, aka private browsing, depending on the browser.

Hello mister,
in the concepts, there is “ad impressions”. I don’t really know what to say about that one. Could you enlighten me, please ?
I think it’s the number of times we see the ad but I’m not really sure about it. I found it in the sentence ” Conversion rate increase with more ad impressions. If you have a good campaign you can show an ad up to 6 times”.


Indeed, the ad impression is the fact that the adserver serves (i.e. places) the selected ad inside the (until then) empty ad space on the webpage. The adserver records this as an impression, updating its count to charge the advertiser supposedly on a CPM basis.
This is a situation where you cannot go deeper into the concept itself, but need to look for collateral concepts (serving the ad, CPM, inventory, display ad, banner, …) or higher concepts (adserver, display advertising, banner advertising) to provide a clear view of the context where the concept comes.
This does not mean that there is actually nothing else to say about the contept itself, ad impression, because a more advanced course would for example analyse when and how, technically, an impression is recorded as an impression. This does not concern us here ;-).


Thank you 🙂


Hello Mr Ranschaert,

I have some questions about the concepts :
– When you talk about the #mondayblues, you are referring to social intelligence, but I don’t really get the link…
– I don’t really understand the symbolic/deterministic programming in the machine learning?
– Is rich answer and featured snippets the same thing ?
– Also, I really don’t understand the 3 functions (or, and, not) and 4 operators (attribution, sequence, loop, test) in the bit system…Do we have to go deeper in this explanation?

Thank you for your help !
Have a nice day.


Thank you for asking Eloise:

1. actually, I mentioned #mondayblues to illustrate the need for social listening and the opportunities it offers: memehillstudio was able to spot a trending hashtag on Instagram that they leveraged to garner more engagement around their own Instagram account.

2. The symbolic /deterministic approach in AI corresponds to the elaboration of algorithms and expert systems where no automatic machine learning is involved, but where the algorithms are coded hard-coded into computer programmes: both the algorithm (i.e. the set of rules to treat input and turn it into output) and the computer code are generated by humans. This involves the aggregation of hundreds, thousands and even more different simpler algorithms, which are coded into computer programmes containing millions of lines of code. This approach imitates human thinking because the rules it is based on consists of mathematical symbols, which are manipulated. The advantage is that it is a form of AI that “thinks” in a similar way to humans, making it possible to understand how the AI produced its results (vs. the black box or neural networks in machine learning), while it confronts the complexity monster (space, time, human complexity). Historically, this approach was introduced and researched before machine learning, and is sometimes called GOFAI (good old-fashioned AI) by specialists.

3. rich answers = featured snippets; they is different from rich snippets (which we did not discuss)

4. I introduced this to show that there is no magic or mystique in AI. There is a long mathematical and philosophical tradition going back to Leibniz, that shows that all human reasoning can be broken down into binary choices, i.e. “yes” and “no”. Even complex reasoning, such as demonstrating mathematical theorems, can be carried out in this way. Claude Shannon is the researcher who formalized all this, bringing together the insights and theories from (among others) Leibniz and Boole, and implementing that in the technology of his time (telephone systems with binary switches on/off). His PhD demonstrated that human thinking only needs two positions, 3 functions and four operators). As it happens, not only a secondary school child can carry this out… a computer can do this too. And here we have AI! Christophe Steiner provides a fascinating account of this historical development in his book Automate This: how algorithms took over our markets, our jobs, and the world (chapter 2).

Did you get what you need?


Yes I understand better, thank your for your help!


Hello Sir,
I am not sure to follow you on the “complexity monster” concept. Could you enlighten me? Thank you!


As expert systems, typical of symbolic AI, require working out the algorithm and coding them into machine language “by hand” (as there is no machine learning involved here), as soon as the operation to realize by the AI expert system becomes a little complex, this involves thousands of algorithms that are merged and sometimes millions of lines of code. This gets veeeery complex: in time (to write all this), in space (computer memory) and in understanding: time, space and human complexity!

I have some questions about concepts :
– LoRa: I’m not sure that I have clearly understand this concept. Is it used in IOT to make the communication between connected objects ? Can I have a deeper explanation?
– lead generation : is it the moment when visitor are turned into lead ?
– Brand protection : this is in the part « social media monitoring » but I don’t have any explanation about this concept
– Retargeting vs remarketing : for me, it is the same concept.. I don’t get the difference ?
– SEO quake : I don’t understand this concept…

Thanks a lot for your answer

May 28, 2018 10:53

Hi Roxane, I think I might be able to help you for some of your questions, I have one too about remarketing.
– LoRa: is a Proximus telecommunication protocol, it is ideal for IoT because it is long range and low power consumption.
– lead generation: is a measure us to calculate th amount of leads generated by a campaign. It can be calculate by #SM leads / total #leads (SM= Social Media)
– For brand protection I describe it as the combat against the loss of revenue, reputation and customer trust that occurs when someone esle exploits your brand for their own gain. And indeed listening helps you it that fight.
– I have the same question for retargating and remarketing. Online I have read something about remarketing being about email.
– SEO quake: I found that SEO quake was an add-on for your browser, it provides a SEO analysis.

Mister could you confirm if I am in the right way please?
Thank you, I wish you all a nice week


Hello Aurélie and Roxane,
this is really what I was hoping for: co-creation, or co-study. I’m sure this also happens under the radar, but posting it here enables me to monitor it, for all to benefit from!

“– LoRa: is a Proximus telecommunication protocol, it is ideal for IoT because it is long range and low power consumption.”
It is a communication protocol developed in open source (so, it is not proprietary), where Proximus is indeed a contributor and user. Long range with low power usage is suited to IoT, that is also right. The sort range of Bluetooth for example, is not adequate for IoT, which may carry communication signals over long distances, even between moving objects (a car for instance). Obviously, IoT objects need to be low-consumption items.

“– lead generation: is a measure used to calculate the amount of leads generated by a campaign. It can be calculated by #SM leads / total #leads (SM= Social Media)”
More generally, it is the ability to generate (i.e. create) leads – the ability to turn visitors of a web-site or callers to a call-center into interested prospects (or “leads”). In the Inbound framework (Attract-Convert-Close-Delight) it corresponds to the “convert” stage; in the RACE framework it is the “act” stage; in Kaushik’s framework it is the “think” stage – they are all equivalent or at least very similar).
As this is a possible strategy, you’ll need to define KPI to measure this (you remember: Situation Analysis – Plan – Execute – Measure). Such a KPI may be termed “lead generation” and defined as Aurélie is suggesting. But a KPI may vary from company to company, and “lead generation” may get other ways of measuring it, depending on how you define “lead generation” as a goal. Don’t forget that on top of that you should make sure to make that goal SMART, for example by including a time framework.

– For brand protection I describe it as the combat against the loss of revenue, reputation and customer trust that occurs when someone else exploits your brand for their own gain. And indeed listening helps you with that fight.
Indeed, brands are exposed to many dangers (brand bashing, hashtag hijacking, negative UGC, negative SEO, click fraud, adpocalypse). Listening is a necessary approach to detect this and react swiftly. AI can also help (fight fraud for instance), and a combination of Listening and AI (= social intelligence) can help a brand go even further, by detecting use of the brand’s logo in pictures for example.

“– I have the same question for retargeting and remarketing. Online I have read something about remarketing being about email.”
Indeed, in our course we can use them as synonyms, even if specialists introduce some differences. What we mainly developed is properly called retargeting.

– SEO quake: I found that SEO quake was an add-on for your browser, it provides a SEO analysis.
Right you are, one of the tools needed when working on the SEO of your website, in this case a browser add-on.

So, all in all a very sound basis, Aur̩lie. Do not be afraid of climbing higher in the hierarchy of concepts, in order to mention and describe colateral concepts, showing you feel at ease going from one concept to another Рas long as you make the differences clear.

OK, Roxane?


Yes thank you !

There are a few things that I don’t understand.
-DSP SSP, I don’t understand what it is at all..
-zero search result. Not quite sure, is it just the fact that when you ask for an information on Google (like the weather today in Belgium), you get only one answer given by the AI? And that you have to click on “more results” to get the SERP?

Thanks in advance for your help.


Hello Sarah,
so, different questions!

1. Web-to-web
We used this in the context of analytics, tracking and retargeting. Today’s digital consumption is cross-device, and involves websites as well as apps. Cookies used to be adequate and simple technology to track and retarget visitors of web-pages within one same browser. Now, apps do not support cookies and the mobile web is not cookie-friendly (de-activated by default on iPhones for example). New technologies are now used to track and retarget users across all their digital touchpoints. Web-to-web means the possibility to track one single user across the web on different devices and using different browsers. Detecting, for example, that user X who visited your landing page on Chrome on their desktop is the same person as user Y who visited your homepage on Safari on their iPad of iPhone. I have briefly outlined how this can be done in another answer on this page: known IDs, stable IDs and statistical IDs.

2. Demand Side Platforms vs Supply Side Platforms
These are the digital platforms on either the buyer (advertiser/agency) side or on the seller (publisher) side in the programmatic buying process. You do not need to go much deeper into this, but rather understand the context they belong to: programmatic buying. The videos on the LIVErtising Youtube channel may help you.

3. it’s exactly that – this is an experiment that Google ran in March, which they have not rolled out more generally (yet). The impact is huge, as it completely changes the look-and-feel of the SERP (one result only) and goes back to the need to be “number one” on Google, as the SERP only shows that one result! For now, Google seems to have abandoned it, but it shows that SEO requires constant listening and monitoring what the search engines are doing. This remark enables me to state something I have not had the opportunity of saying earlier: even if Google is the absolute #1 search engine in Belgium, an SEO professional cannot afford to ignore the guidelines of smaller engines.

I hope this helps you, Sarah!?!


Yes thank you very much !


Would you explain me where the adserver is in the model you showed us in class (I tweeted you in case you don’t see which model) please?
Also, what’s the difference between having a Google-friendly website and a user-friendly website? Because Google wants to improve the user experience…
Thank you!


Hard at work, Graciela!
Adserving underlies all digital display advertising (and even more than display), in traditional adbuying as well as in programmatic advertising. They are the two server icons I have included under the media buying agency icon and the media sales house icon, but are used in direct buying/selling, by adnetworks and on adexchanges.
I have already answered the other question, as I explained to you in my twitter reply.


About the two stages of machine learning, can you confirm that first developers need to train the program (to input informations and then correct the output) so then they can input this in a program which is going to be able to predict informations?
So is supervised learning the first stage (train) and unsupervised learning the second one (predict)? Does a program need first some supervised learning to then work with unsupervised learning?
Finally, do you know if it’s possible to bounce in app? Is it possible to measure a bounce rate in app? Because people need to install it, it’s not like a webpage…
I hope you can help me, thank you!


Yes indeed, Graciela, in machine learning there is a training stage first, where the algorithms learn the task at hand. When they are fine-tuned, they can then process new input to produce output, for example in terms of prediction of behaviour or decision (and many other possibilities). That training stage can also produce new algorithms that are then applied to new input.
The second aspect of your question is not correct. Supervised vs unsupervised are two types or training processes – they both belong to the “train” stage. What is the difference then? Supervised learning requires labelled input (for instance a huge database of animal pictures labelled “cat” or “dog”, that may count in millions of pictures) and checks the output (did the algorithms analyse the pictures correctly: did they recognise cats as cats and dogs as dogs – if not, the algorithm is tweaked at each layer of the network in the case of neural networks, until the “cat vs dog” analysis is sufficiently correct. UNsupervised learning works with unstructured and unlabelled input (e.g. a vast database of cat and dog pictures) fed into the neural network, but here the algorithms are going to build their own concept of “cat”, i.e. they are going to detect that there are similarities in all the cat pictures which enable it to determine with a certain degree of probability that a cat pictures a cat and not a dog. These two training approaches involve different methodologies and statistical analyses and their choice depends on many factors, such as the data type and the output you need.
I realize the slide entitled “two stages in ML”, i.e. (1) train and (2) predict, may mislead you into thinking that supervised learning = train; actually: train = supervised or unsupervised – once training is satisfactory in output quality, you can use the algorithms to predict new outputs.
Is this clearer now?

Delbauche Laureen
May 22, 2018 10:30

Hello Mister,
I don’t understand what the concept “clustering” means, I can not find it in my notes… Can you help ?


Hello Laureen,
a quick nudge to help you find it: clustering is a statistical technique which is implemented in unsupervised machine learning. Let me know if you need more help.

Laureen Delbauche
May 28, 2018 16:15

Indeed I guessed it concerns Machine Learning (when I Googled it) but I don’t understand what this technique does?


Clustering, one of the most common techniques used in unsupervised machine learning, is a set of statistical techniques intended to find hidden patterns or groups or categories in data that appears to be unorganised. An example may be feeding medical records into a neural network, making it crunch the symptoms, and provide clusters of symptoms at the output end, pointing to a possible disease.
Here, I can add to a comment I made earlier about supervised vs unsupervised, that it is indeed possible to use unsupervised learning in a first stage for explaratory purposes, and then use these findings to analyse the data with supervised learning tools, both at the training stage, until output results are sufficiently correct to go over to the prediction stage.
OK, Laureen?

May 22, 2018 09:20

I didn’t really understand the difference between user-friendly and Google-friendly. Can you explain me that again, please? Because you said we should first improve our website so it’s user-friendly, but the aim of Google is also to provide a user-friendly experience so I don’t clearly understand the difference.
Then, about retargeting, is bouncing possible in app? So is it possible to measure your bounce rate for your app? I feel like it’s not possible because people have to download the app…
Thank you!


Hello Anonymous (!),

I understand this is a little confusing: the philosophy of present-day SEO is to help you meet marketing communication objectives, **not** to please or cheat search engines. You want to nurture KPIs, not boost visit metrics. This is the foundation. To do so, you need to offer your visitors / leads / converters / purchasers the best possible search and web experience. This is the search experience optimization or search UX. It is focused on the user, not on the search engine.
To do so, your site must at least be findable, crawlable, searchable, indexable. There are technical requirements to meet in order to achieve this: make the search engine’s work as easy as possible (internal linking structure, clean HTML code, straight Hn hierarchy, image and video weight, metadata, …), and removing the known obstacles (flash for example). This is what I called “google-friendliness”. It contributes both to a smoother job for the search bots, indexer and query engine **and* to a better user experience.
An example: you need alt-tags on all your visuals because (1) it enables search bots to “see”, or rather read what the images represent (this is google-friendliness); (2) it enables your visitor to know what is in those pictures in case they do not display properly or are somehow blocked. The latter is UX.

Technically, bouncing is restricted to web page visits. To my knowledge it is not used for in-app behaviour. However, some apps such as ecommerce apps are indeed also confronted to the same behaviour (for example visiting a product page and leaving without buying) and may indeed want to re-capture a prospect’s interest in order to convert them, using retargeting. Also, many people access Facebook as an app, where Facebook indeed offers retargeting/remarketing.
This should answer the heart of your question.
I want to add a technical element which you do not need to go deep into: as apps do not support cookies, which make website retargeting possible, in-app retargeting or app-to-app retargeting or web-to-app retargeting or app-to-web retargeting, or cross-device retargeting more generally, have to use other tracking technologies than cookies. These may be known IDs (such as social media identification), stable ID (such as digital ID tags developed by Google or Apple), or probabilistic/statistical IDs (such as fingerprinting).

Does that help?

Anastasia Bousson
May 10, 2018 17:44

Hello !
I don’t clearly understand the difference between the ad exchange and the ad network. Can you help me clarify the difference ?

Thank you !


Hello Anastasia,

Even if these two concepts sound similar, they are not. Ad networks are actual companies offering a service: they aggregate unsold inventories from publishers, package them and sell them to parties interested in buying them (advertisers, agencies, media buying agencies or even other ad networks), completing this with for example targeting facilities. Through a network, real inventories or audiences are traded between a specific buyer and the network on the one hand and between a specific publisher and the same network on the other hand. The media impressions are traded at a fixed price.

Ad exchanges are places (i.e. virtual technology platforms, not actual companies) where the buying and selling of advertising inventory is automated and carried out using an auction (or bidding) system, where the transaction is mediated by the auction process, meaning that the exchange regulates offer and demand like the stock exchange does for securities, not at a fixed price. On ad exchanges you bid for an audience, rather than buy ad space or time.

Does that help?

Anastasia Bousson
May 16, 2018 17:15

It does, thanks !


Good! Feel free to fire other questions at me and to invite other students to do so as well!

@LIVErtising What’s the difference between long tail and post tail? I hope you can help me. Have a nice day!


They have nothing in common, not even the “tail”. In “post-tail” it actually means “post-retail”, or “post-purchase”. This is a stage in the CDJ (Customer Decision Journey) as developed by McKinsey to deal with the shortcomings of the traditional funnel. It corresponds to “care” in Kaushik’s see-thing-do-care model of communication strategy, or to the “engage” dimension in the RACE model by Smart Insights, or the “delight” phase in the Inbound attract-convert-close-delight model: leveraging the experience people have of the product/service to turn them into advocates, networking for your brand – i.e. enabling your brand to fully use the third stage of marcom: networked marketing.
The long tail on the other hand is a concepts that originates in statistics but became popular in marcom when Chris Anderson published his book “The Long Tail”, explaining “why the future of business is selling less of more”. It has been transfered from sales to SEO to apply to keywords. Long tail keywords (as opposed to head keywords) are multi-word queries (such as “best home insurance company for expats in Brussels”), which provide clear advantages in terms of SEO (lower competition, more targeted, more conversational). Neil Patel provides a clear explanation of how some brands leverage long tail keywords in this blog post – in case you want to dig deeper!
Is this clearer Graciela? Let me know.


Yes thank you!


Leave a Reply

Your email address will not be published.