ProgSol, Programmatic Solution, s.r.o.
Karla Engliše 3201/6
Praha 5, 150 00
There are huge changes happening in the online marketing world. There is an ongoing boom in programmatic buying in display campaigns, a sizable movement in the automatization of PPC campaigns, a big jump in Facebook advertising effectiveness, marked growth of video campaigns and the reach of media ads, and many other developments. All these changes are improving the precision of targeted campaigns and their results.
However, measurements of effectiveness are much as they were a few years ago. Commonly, we use very accurate metrics, but these metrics consider only one device – one cookie. But this does not account for the fact that many campaigns have results in the offline world as well. Customers are not only buying online; they go to brick-and-mortar stores as well. In addition, many conversions take place on mobile devices. With a broader approach, we can catch the attention of a customer on their desktop and retarget them the next day on their mobile when they are near our shop.
Do you realize that today’s technologies can take these kinds of conversions and measure them? There has been a rapid increase in the number of people with smartphones in recent years (2/3 of Americans now own a smartphone). Many of these users have geo-location information permanently active, and most use map and navigation apps (74% of users used navigation services in 2013).
Exact, real-time location data has great relevance in advertising. We can seriously say that visiting some place and staying there for some time can be considered a conversion. Customer visits to stores are now trackable through digital means for the first time.
On the Internet side of things, it is very common to target users with ads based on geo-location data. This targeting is only approximate on desktop, but targeting mobile users is very exact.
In addition, there has been very strong progress made in cross-device algorithms, which can merge users’ profiles across devices. These algorithms allow us to target a user on their desktop and continue to track them in the real world by use of the smartphone in their pocket.
These cross-device combination solutions for measuring offline conversion will be an important topic over the next two years. I expect that we’ll see some interesting solutions. Already we see Google taking the first steps, as they are now testing offline conversion in AdWords and are beginning to use cross-device algorithms as well.
And what if we go even further? It is possible to track people’s movements on the streets and to compare their locations to the locations of billboards. Or, once more users own smart TVs connected to the Internet, it will be possible to track whether or not someone was targeted by a television ad and then to make the same action later—for example, when visiting web pages or brick-and-mortar stores, or simply when searching a relevant keyword on Google.
The future will bring many new opportunities to the marketing world. Are you ready?
If we use our resources to buy banner ads and want to have a unified evaluation in Google Analytics, it is necessary to have uniform UTM tags in links.
If we use similar UTMs in all campaigns and for all directed paid ads, we can evaluate everything in Google Analytics. For directed paid banners we recommend using AdServer. By using AdServer, we receive more information than we would merely from Google Analytics. We will thereby gain more functionality for banners, as well as for story-telling.
A final evaluation could be for a retargeting campaign. For example:
Or, similarly, for a prospective campaign in which Post View parameters are more important:
For Post View parameters, it is usually necessary to use AdServer.
We are using UTM tagging for this model:
Some variables can be set automatically. For example, in AdForm we use:
AdWords can use an automatic tagging system; but be careful when naming campaigns to keep data in Analytics consistent with other DSPs.
We use AdForm AdServer for all directed paid banners from publishers.
Very often I am confronted with the objection that RTBs have a higher cost per click (CPC) than PPC systems. Yes, this is true – not always, but often. But is a higher cost per click really what you want? Well, usually not. In fact, we really want relevant / high-quality clicks – users who are truly interested in our content – users who fill orders or search more pages on our site.
Just as when you want to win an F1 race, the only criterion you will be interested in will be the maximum speed of the car. But that result will depend on the experience of the driver, the quality of the service team, the tires, etc.
There is a huge difference between PPC ads in search results and banner ads in content. In-search ads go to users who have searched for something: these users want to visit a relevant site. Content ads are shown to users who are doing something else entirely.
In these two cases, effectiveness must be measured differently – CPC will be different. It is not possible to compare the two types of ads. You should use display ads after you’ve already used search ads to maximize your ads’ effectiveness.
However, the click is not the final parameter. We don’t want just any click – we want users who are actually interested in our content.
Always measure the quality of clicks. The simplest measure is in an eshop campaign. Usually we want to sell – in this case, a conversion is a filled order. What is important is not CPC (click cost); far more important is CPA (cost per action). And a higher CPC often means a lower CPA.
On content sites (with no simple conversion), we can measure how many pages users visit, how long they stay on the website, what the bounce-rate is, etc.
Evaluations can be presented as examples for eshops:
Or for content sites:
It’s been shown that cost per click is important, but it’s not the only (nor the most important) parameter. Some sources with higher cost per click may result in inferior conversion.
For content sites, we can set up a formula to evaluate conversion based on higher quality parameters.
It is very common to purchase display ad (banners) via different methods and different systems (DSP). Many clients are buying banners through Google AdWords and also through some DSP (AdForm, DoubleClick, AppNexus, Dataxu,…). E-shops use very often black box retargeter systems (RTB House, SocioMantic, Criteo, AdRoll…) too. Some clients buy direct placements through their chosen publisher as well.
In these cases it is inevitable to get in trouble with evaluating results. Some of these issues we will describe below and show how to fix them.
If we use banner ad systems that are based on auction model (PPC or RTB), there is a problem with self-increase of auction price.
For example: we have 1000 cookies, which visited our website during the last 14 days. If we start bidding for these cookies in two or more auction systems (f.e. AdWords and DoubleClick), as a result we will buy impressions at a higher price. This is a simple example, in reality it is more complicated due to the different bidding models and different system ranges.
It is almost the same as if you want to buy some art and give this task to two different art-agents, who do not know about each other. During auctions they would make the price go higher and at the end they would be bidding only against one another. So they would eventually get the item for you, but at a price higher than necessary.
Let us imagine that you want to show ads maximum three times per day. If you use more DSPs, it is very hard to coordinate them so that they do not exceed this frequency.
If we retarget one cookie in three DSPs and in each we the frequency set up to max. 3x per day, is possible that we receive a frequency 9x per day in the end because each DSP would fill up his own frequency.
Another relatively common phenomenon (if you use retargeting in more DSPs) are two or more impressions of the same client on one page at the same time. This is a result of the fact that the page is using more systems for selling impressions (SSP). If you use more buying systems (DSP), there are many combinations which can happen. For example, one DSP can win an impression on a page and fill its own frequency, bud second DSP can win a placement on the same page as well.
Because more SSP can be used on one domain it is very difficult to fix this problem if you use only one DSP. If you use more DSP it is almost impossible.
On the other hand, there are DSPs which can buy all impressions on one page. These systems are called RoadBlock format and can be used for banding effect.
If we buy banners in direct sales from publishers, we usually buy certain number of impression in specific time at a fixed price. But usually it is not possible to choose which impressions we will buy. In reality we are buying all impressions that the selling site is offering us.
But we usually want to show different ads to cookies we are retargeting and different to which we are prospecting. Or we want to buy only retargeted or only prospected cookies, or only prospected cookies with some characteristics (women at age 25-40). It is very helpful for evaluation to separate these two models to have comparison with other buying strategies in our DSP.
From the programmatic poit of view it is better not to buy ads in direct sales, but to set up a private deal with the publisher and use only those cookies we really want.
For reduction of these problems we recommend to use the below described procedure.
At first it is necessary to set up the plan or strategy for targeting the users. This is generally important for the whole marketing, but we will be writing only about banner ads.
We need to convert the strategy into the buying system/systems (DSP). We will choose DSP in accordance with the strategy, not setting up strategy in accordance with the used DSP. This is very often the case in PPC – we automatically buy banner ads in PPC and do not coordinate it with RTB because there are different persons controlling these systems. In reality, many clients use more people to buy display ads:
But today we have the technology to be more effective and to use only one system or one person to coordinate this according to our strategy and need. So if you do not have special reason why to keep the “old” system of dividing the activities across more systems or people, we recommend not to use PPC system but DSP which has the functions we need (there are important differences between DSPs and it is necessary to know how to choose the right one).
Why not to use PPC systems:
On the other hand, if we need to buy only cheap clicks and other parameters are not so important – PPC systems are a good choice.
Very frequent argument for using PPC systems to buy display ads is cheap price of a click. That is true – generally a click in PPC is cheaper then in DSP. But is really the price of a click the only thing what you want?
In reality there are other parameters like conversions, time spent on the site, number of visited pages,… which describe the quality of a click. As a result it can happen than a click with higher price has a better quality than the cheap one and means a lower price of conversion.
ProgSol, Programmatic Solution, s.r.o.
Karla Engliše 3201/6
Praha 5, 150 00