We are very happy to be here. This is Jannette and I’m Miriam and we’re gonna talk to you about creating the future digital marketing today. So, talking a little bit how Google thinks about what’s going on in digital marketing and how you can get yourself ready for these changes that are happening today. But before I’m gonna talk a little bit about the future, I want to go back into the past to the year 2002. In 2002, we launch the paper click model. That was actually 14 years ago. And not only did we launch the paper click model but we also started thinking about something else, and that was the last click attribution model. I think most of you will be familiar with the concept which is looking at the last interaction or the last click before a conversion, you have a click out conversion an app download or whatever happened. And this has been the model since then, since 2002. We have quite a bit of legacy there. 14, 15, 16 years of one way to look at digital marketing and at search advertising in particular, and what we gonna talk about today is how this might change and how you can think about using these new ways of attribution modeling in your digital marketing efforts. First, what has change since 2002? Just a quick show of hands, who has been in the business in 2002 already? Okay, not too many actually. But I’m sure the ones that have been in business in 2002 would agree with me that a lot of stuff have changed since then. I wanted to go through some of the changes on the consumers side now. Actually, three. So, the first one, I think it’s that we can see that consumers become less loyal to a brand. They are more likely to change and switch between different brands, between different providers, so we picked an example here from retail but we see this across the board for different industries. So, that’s very easy to explain. We have the internet coming in. People research more there’s more transparency around pricing, around offers. So, you have a lot of touch points where you need to convince the consumer to go to your website and take the required action there. And don’t go to a competitor website or any other offering in the market. The second change is about the vice usage. And I’m sure you’ve heard about this many many times. This is not really new. This has been going on for a couple of years. But we really see the effect kicking in now, also when it comes to campaign optimisations. Users use different devices. They have lots of different touch points across different devices. And a logic where you only look at desktop and at a single cookie space is not really helpful at the moment because you’re missing out on some of your potential. And lastly, if you go a step further, and you look at the whole customer journey, you can see that it expands across online and offline. You have offline touch points. You have digital touch points. Many shapes and sizes and you have different devices that are relevant across that customer journey. This is the consumer side. I think this also has a lot of implication for marketers or advertisers or anybody who is trying to communicate with their customers in some way or another. A way we think about this at Google is this. This is not research. This is a conceptual chart that is suppose to illustrate how value is driven today in online marketing or in the digital space versus how it’s gonna be driven in the future. And a lot of these changes are already happening. Let me talk a little bit about this. What you can see here today that a lot of time and a lot of value is created in the campaign management space. What do we mean by that? We are talking about setting the right bids, the CPC bids, max CPC bids, about getting your creatives right, about getting you targeting right, to get the most performance out of your marketing campaigns. In the past, also, when we talk to very large advertisers, this has been where lot of the success lied in the past. So, if you were really good at this and you had a great team or great agency supporting you in this, then you were good and you could be really successful. Another big time investment went into reporting. Actually, informing those decisions that you do on the campaign management side via reporting. That can also be quite time consuming. The big shift that we see right now is that this campaign management and reporting aspect becomes smaller and smaller and less of a unique selling point, also when you think about your online marketing skills. Why is that? This is largely due to automation. So, we see a lot of this more monotonous work like campaign management, setting your bids, hitting your targets is being automated by automated bidding, by different tool providers but also by a functionality built into AdWords, for example. The same is true for reporting. This is happening on the one hand, this blue bubble becomes much smaller. But then, on the other hand, we have some other areas that are really evolving. Those are more strategic areas. Things like the data management. What kind of data you have about the consumer? How you can leverage that data. How you can sense of that data. Think, if you go to Google trends and you look for data analyst, you see a curve that is going up a lot because people need that skill set, people need to make sense of the data that they have and that they own. Another aspect is attribution. And attribution is what we want to talk about a little more today. Not only measuring touch points but also crediting them with the exact and the right value that they are driving for a click out, for example. Now coming back to a more operational perspective, what we want to talk about today… Two distinct challenges that we see in the market: The one is about measuring interaction across different devices, across different touch points on your mobile phone, your desktops, your tablets, your second and third mobile phones, whatever it is. And then, two. Talking about attribution in a setting and obviously we’re gonna talk about search for most but this is also a concept that can be applied to other media as well. With that, I’m gonna hand it over to Jannette who’s gonna walk you through the cross device piece. Thank you very much Miri. Miri just introduced us to those big challenges for measurement and attribution. Let’s have a quick deep dive into cross device measurement in AdWords for instance. Just to get everybody on the same page, who in this room is aware of cross device conversions or using cross device conversions? Ok… yeah like, 1/3, 1/4, good to know. Just for everyone to be on the same page, cross device conversion are conversions that start with an ad click on one device, for instance a mobile phone. But they end with the conversion on another device. For instance desktop computer. Cross device conversions were not visible in our reporting, so far. And now that we launch cross device conversions, it’s super important for our advertisers or affiliates to also consider those conversions. As you can see, it’s real conversions. Ad click has just happened on another device. Now that technology makes this possible. We absolutely advise our advertisers to take advantage of that. Let’s have a look at our ecosystem and how we measure cross device conversions. A lot of our advertisers ask me, what is the user base based on how do you measure that cross device conversions? And I always say, that’s the fact. We use our logged in data as a basis. Then the next question is always…How many logged in data does Google have? And as I see quite a lot of people in the room, I’d like to test this with you guys. Quick show of hands if you do have an Android phone or if you use Google Play. Keep your hands up. Alright. Who in this room uses Gmail, for instance? More hands. Wow, that’s almost half of the room. Keep your hands up, please. Who in this room went to this conference with Google Maps, or in general uses Google Maps? Please, keep your hands up. Another thing, who uses YouTube? Alright. I think now we have almost all hands. Please, keep them up. Who uses Chrome, Google Chrome? Now, who uses Google search? The big big bomb. Keep your hands up. Who is logged in with one of those products? Keep your hands up. I think everybody in this room. Now, who believes that Google has the right cross device solution? As you can see… So, Thank you. This is how we just explain we do measure cross device conversions. And when we launch this five years ago, we did not want to show you any number. We wanted to show the right number. That’s why we hired a lot of mathematicians globally and we wanted to have a real valuable methodology. Let’s have a look at this. For cross device conversions, we use a lot of signals such as the device that you are using. The conversion type. Such as app installs or leads. We use the date and the country. For all those signals, we kind of calculate the cross device conversions and then we extrapolate. And for that extrapolation, we only show the cross device conversions if we have a confidence interval that is at least 95%. That again shows you that we only shows cross device conversions if we are very sure that they do exist. In order to have a look if everybody understood what cross device conversions are, here is a little exciting quiz for you guys. And in order for that, you need to know how cross device conversions are reported in our AdWords interface. This is actually how it looks like. You will see a column that shows cross device conversions. And in this case, for instance, you have six cross device conversions attributed to mobile. So, you will see the split of devices. In this case you have computer, mobile and tablet. For the quiz that we’re going to do, you need to ask two questions. First question is – Where was the last ad click? On what device did the user do the last ad click? Because this is the device where we would attribute the conversion tool. The second question that you need to ask yourself in a couple of minutes, is – Did the user do a device change after that before he converted? Was there a change in devices and then, a conversion. So, that’s the second question. And if the answer is yes, then it is a cross device conversion. And if the answer is no, then it’s no cross device conversion. Did everybody understand it? Nodding. So, please do stand up. It’s a little energiser as well. Let’s start with the quiz. First, this is our first scenario. We do have a search ad click on a mobile phone and then the user changes devices and with organic search click he goes to the website, and then he converts on the computer. The question is: A.) is this a cross device conversion attributed to computers? or is it B.) a cross device conversion attributed to mobile? I’ll give you 10 seconds to think about it. Show of hands who thinks it’s A – it’s a cross device conversion attributed to computers. Some people like 1/4, 1/5. And who thinks it’s B – it’s a cross device conversion attributed to mobile. A little more people. Everybody who said A can sit down. This is unfortunately the wrong answer. These are cross device conversions attributed to mobile. Why is this so? Because the last ad click as Miri said, we are in a last click model. The last ad click has taken place on a mobile phone. Now the next one. It’s as tricky. Let’s have a look at the second example. Now we do have a display ad click on a computer device. And then the user changes device. And the second display ad click happens on a mobile device. And then there is a conversion. So, scenario one, is this a same device conversion attributed to mobile? That’s A. Or is this a cross device conversion attributed to mobile, which would be B. I give you 10 more seconds. Who thinks that it’s A – Same device conversion attributed to mobile. Half of the people. Who thinks it’s B – it’s a cross device converion attributed to mobile. The other half. Everybody who said B can sit down. In this case, it’s same device conversion attributed to mobile. Why is it same device? Because after the display ad click on desktop, the user change the device and then there was another click on a mobile device. So, the last click happened on a mobile device and then the user did not change devices before converting. So, as you see, it’s a topic where you really need to think about how the measurement works. Thank you and congratulations to everyone who are still standing. Who of our advertisers did this? One the first big advertisers in Germany? We do have Zalando, they are a huge retailer for fashion, shoes, and they do a lot of online advertising with us. They really wanted to see the value of mobile. What they did is, they included cross device conversions into their measurement and they even went one step further. They bid up accordingly. They modify their bids according to the cross device conversion uplift and at the end of the day, they manage to see 30% more conversions that were attributed to mobile. We do see this industry wide. To have a look at other industries, of course, we always observe that industries where we do have long user path, those are the industries where we usually tend to see high ratio of cross device conversions. One example would be travel or automotive. And this is how it looks like in our AdWords interface. The advertiser or the affiliate just needs to click a tick box. And after clicking this tick box you will be able to see those cross device conversions and they will be included in your conversions column. If you, for instance, use auto-bidding, then the cross device conversions will be included in that auto-bidding algorithm. As we mentioned before, these are real conversions that need to be considered. Talking about AdWords auto-bidding, of course we would not launch cross device conversions if it would not be compatible with our powerful auto-bidding. Cross device conversions are compatible with eCPC, so enhanced CPC, with target ROAs and with target CPA. And that’s definitely what we also recommend to our advertisers and affiliates use this technology because automation, I think we don’t have to talk about why automation is important but it saves you time, money and a lot of nerves. What signals does our auto-bidding system determine and use while taking those cross device conversions into account? We use again, time, language, what device are you using what creative are you using. Last year, we also included the remarketing lists so we know the user has been on the advertisers or affiliate websites, so now we bid higher on those users. But one signal that is specially important is the search query, and if you ask me, that is pretty unique with Google. Let’s have a look at this example. Let’s say you are an affiliate or someone who sells Samsung Galaxy phones. In this case, the user could type in different queries, such as Samsung Galaxy, some smart phone high end, or Samsung Galaxy phone plan. Our algorithm knows that those different queries have different conversion rates. In this case, the last query Samsung Galaxy phone has a conversion rate of 9%, so our system would automatically bid a lot higher on this query. And for what advertiser is this relevant? For instance, if you have broad match keywords or if you’ll have dynamic search ad campaigns, Google shopping campaigns. That really helps and to me, this is super super powerful. Well and with this, I’d like to hand over to Miri. Maybe, a quick question, who’s already using cross device conversions for their bidding and for Google search? Anybody? One hand? Yey! It’s really something you need to get your head around to understand it and know how it works. But I think, Jannette, as she explained it, this is a last click model so we are not double counting anything. It’s just a simple way of showing you something that we couldn’t show you before. Jannette talked about cross device and the important thing here is she talked about the measurement side of things. Are you able to measure all the relevant touch points that are happening before somebody does a click out on your website, for example. Measurement, and that’s an important distinction to make is not the same as attribution because attribution is about giving the right credit to a touch point. And this is what we are gonna talk about now. I’ve touched on this briefly in the introduction, the most prevalent common model, also, Google introduced that model back 14 years ago is the last click model. You are giving all the credit to the last interaction that happened before conversion. Who’s running their campaigns based on a last click model in here? Ok, not everyone. Maybe 2/3. This is how it works and obviously this is not the reality. Specially not in a world where consumers research intensively in the digital space before they convert. You have lots of touch points that are happening maybe with your AdWords search ads before somebody does a click out. There are two types of attribution models that can be helpful here and that are currently being used in the industry. One of them are rule based models. And those have been around for a couple of years already. This is an example. It’s a linear model. It distributes the credit evenly across all touch points. If I say all touch points in this scenario, I’m talking about Google search ad clicks. This is what we are working on right now. There are lots of different rule based models. There’s a position based model also sometimes called the U shaped model and obviously that’s first click. That’s also a model. And there’s last click which is the most common model. But all of these are rule based meaning they have a fixed percentage how the credit is distributed. And of course, you can click out at closer to the reality but they will never be perfect. The second type of models are data driven models. And those have been coming up over the past, I would say a year and a half, really being used in the industry. What is a data driven model? With the data driven model, the credit is distributed by an algorithm and it’s distributed dynamically. Meaning that one day it could look like this and the next day this could look completely differently. How does this happen? And I’m speaking for the Google AdWords solution right now. What we do, if we have enough data in a given AdWords account, is we look at different conversion paths. We are comparing conversion paths that are very similar to each other except for one touch point. This is also marked here. This one touch point was in the first conversion path and it was not in the second conversion path. We do this to determine what is the incremental uplift you are seeing for your conversion rates based on the difference. Just say, this was an important touch point while another one maybe wasn’t. And this is very important to understand the meaning of each touch point in this customer journey. If we have enough data, we can do this for you in the AdWords account and you can also look at the data driven attribution model compared to a linear model or to a last click model. Jannette talked about cross device and of course we didn’t leave cross device out of this. This also works for cross device. Here’s an example. You have a path again, you have AdWords ad click, another AdWords ad click on mobile. You have a switch to a desktop device, you have another ad click there. You have a website visit and a conversion. Based on the quiz, all of you should know how this looks like. But I’ve already shown it briefly, so I’m not gonna ask you now. But in the last click model, you would just measure a very simple last click conversion for desktop. This is where the last ad click happens and it was also the converting device. Those two things. In a data driven model, this could look like this. You could have point zero attributed to the first ad click on mobile then another point one to the second ad click on mobile and point six to the last ad click on desktop. We are getting closer to the truth. Also, a data driven model is imperfect but it’s based on your own data on you own AdWords account and it’s taking into consideration changes that are happening. Like changes in the auction, for example. Because it’s dynamically adjusting the model on a continuous basis. This is why we recommend using a data driven model but you are also free to use any of the other models in AdWords. If I say use a model, there are two components to that. You can look at it, there’s a nice set of reports. You can find them at AdWords under attribution. You can play around with it and just understand what the impact would be if you were switching to a new model. And then, the second thing is take action based on that. In the past this was really lacking. We’ve launched this now in AdWords to include the new attribution model, also into your bidding and also into your budget allocation, for example. So you can just create a new conversion column based on a new model and you can also use this for your automated bidding, for example. Who’s using automated bidding? Target CPR, conversion optimiser, ECPC… Okay, about half of the room. You guys, you can simply switch your attribution model and bam your bidding on a new type of model that’s closer to what your consumers actually doing. But I recommend you look at the data first. I think that already brings us to the end of this. I briefly like to summarise what we’ve talked about. I think the first thing would be check if you have cross device data in your account. It’s really exciting to see it. Just look at the uplift you are seeing, maybe compare it to the industry averages we’ve just shown. Those were Germany’s specific numbers but you can also find international numbers on our blog. Second thing is, use AdWords automated bidding. It’s just the most straightforward way to include all of these changes without causing a ton of work load. And lastly, check if you have a data driven model available in AdWords. I mentioned you need a certain data threshold that you need to meet in order for us to calculate that model for you. So, check if you have it, if not, you can use the other models as well. And then, as a last thing tested, see if it adds any value to how you look at your campaigns, if it helps you make the right decisions where to put your investment, etc. I like to finish with one example. Who have you have heard of the game of Go? Okay, good. I can say, tell a new story at least to some of you. The game of Go is an ancient version of chess. I’ve been told a much more complex version of it. We have a computer that we trained to play the game of Go, it’s called AlphaGo. AlphaGo played one of the biggest most famous game of Go players, Lee Sedol, in March, in South Korea and actually won. That was a surprise for many many people, I think, also for us. What it shows is that computers and machine learning in particular, can actually solve very complex tasks such as setting the right bid for you CPCs but also something like playing the game of Go. We just wanted to share this with you to give you an idea of what can be done if you apply these technologies in the right way and that it can save you a lot of time and also, hopefully, money if you use them.