Unlike Search Network, GDN Quality Score is not communicated in any way by Google and can’t be visualized in any custom columns, but nevertheless they use it to determine which advertiser will win a display placement and to set the cost of the click.
As it is impossible to see this KPI in our display ads we can only have an indirect and partial indication of it looking at:
- Click Through Rate (CTR);
- Total cost of the placements & Cost per Click (CPC);
- Conversion Rates (CR);
I will come back to these metrics, but first of all if we want to optimize a GDN campaign (and we definitely want to), we have first to eliminate or limit “junk sites and applications” from our placements (i.e.: such poor in quality or relevance to our campaigns that might even affect our brand).
Many of these sites/applications should actually be blocked from the beginning, if we exclude the most potentially problematic topic categories in the campaign additional settings (look at the image below).
GDN sites and apps are many millions and grow every day. These category filters are not enough to significantly improve Quality Scores of our ads, but at least will prevent our ads from being shown in the worst Caracan Bars 😉
Moreover they are obviously less important in “pure remarketing” campaigns, because in this case we are following users, not content context.
Last but not least, Google recently announced that starting from September “adsenseformobileapps.com” placement exclusion will stop working. It would seem like this is bad news, but actually it comes along with enhanced filtering possibilities in the Display Network campaign additional settings by devices.
In this example you see a complete block for our ads in the Display network for apps.
I am not suggesting to do it in all campaigns since the beginning, because I’ve seen several cases in which these placements performed very well in terms of both CTR, CPC, BR and CR. This mainly depends on what we are advertising. The more the promoted products/services are free or low-cost, or have low user commitment (so-called emotional purchases), the more these placements might work. Take this decision only when you have strong signals that your campaigns are not going to perform when shown in applications.
Regardless campaign settings, to evaluate the situation we have to check the “placement” tab from time to time (the higher your daily spend is, the more frequent we need to control it), looking at “where ads showed” and starting from the most potentially dangerous sites/apps:
- Sites/apps that cost us a lot but do not convert at all, whit very high CTR or billions of views and many clicks even with indecently low click rates.
- Sites/apps that do a lot of impressions but very few clicks and conversions, in the eyes of Google these are the verified proof that our ads are poorly working (despite the fact that content matching is often done automatically by the platform itself!).
To find them faster just order results by decreasing cost, then by impressions and finally by CTR (adding some impression or click filters and selecting long enough time windows to retain statistically significant data), as shown below.
In the examples above there are not app placements because they have been blocked, but it is instead quite likely you will see a lot of them.
If not relevant or not bringing conversions, these sites/apps are serious candidates to be excluded. In the most evident cases the “offending” site/app could be added to a placement exclusion list (the GDN equivalent of negative keyword lists, available in the “tools” menu) to be connected to all campaigns (in order to block them once for all). On the other hand, if we find non-themed or low CR placements at group level, we can simply pause them.
Unless we are promoting a multinational campaign (as in the above example), it’s quite likely that blocking mono-language sites with country specific extensions outside our target areas will result in increased efficiency of campaigns.
In both these cases, before making any decision it is absolutely essential to have at least a few thousand views (I would say at least 1,000 impressions or at least 100 clicks), otherwise we will make choices based on statistically irrelevant numbers.
Gmail or “anonymous” placements (those in which Google and the GDN partner site agreed not to disclose their data to advertisers) are the result of a mix of many different sites, so it makes no sense to consider other parameters different from cost per conversion. In other words, if we see that they are worth the money we spend on, we hold them, otherwise we pause them or add to exclusions (given that it’s not yet possible – by now – to see disaggregate data and block only the poor performing ones).
As for all the other placements, unfortunately there isn’t a quick way to evaluate them other than paying them a short visit. If a Google representative is reading, a preview of the site directly in the control panel would be highly appreciated.
The sites on which it is absolutely not convenient to be present are quite easy to identify, even at first glance, but for the vast range of other sites not so relevant, but also not dangerous for our brand, what to do?
This is where the 3 parameters I initially identified (CTR, Average CPC and CR) come into play.
For Click Through Rates there are no fixed attention thresholds, but we need to do some more complex considerations.
The same banner, displayed on the same site, may in fact have a very different CTR depending on the targeting criteria we are using. A completely unrelated site can work very well if our target users already know us, or very badly if not.
This means that we must first distinguish between campaigns in which we are targeting groups of specific users (remarketing), compared to those in which we’re choosing users to show our ads based on the content they are looking at, or which they have surfed/searched previously (keywords, topics, interests or in-market audiences) or some of their general characteristics (gender, age, geographical area, etc).
In campaigns based on remarketing, it’s obvious that click rates and the subject of the page our banners appear on are almost completely irrelevant (since we have filtered the users who will see the ads), in others they are fundamental, because they define the user’s own affinity with our target.
For example, if we’re promoting discounted cooking pots with users who already visited our site it’s irrelevant whether the banner appears on Allrecipes.com or on a solitaire card site. Instead, if we want to intercept potential users interested in the offer based on what they are surfing, Allrecipes.com will be fine, while we should immediately block the solitaire site.
Having said that, it’s worth considering that it’s normal for campaigns in the GDN to have click rates largely below 1%, but a very close to zero CTR means that the placement is not suitable (or that our ads are merciful ;-). As for the Search Network, creating new variations of ads can definitely help. Otherwise, you have to seriously consider whether to pause placements with the lowest click rates, unless its absolutely essential to safeguard the site/app in question.
On the other hand, a very high CTR can be a symptom of two things: sites very much related to the campaign, or “click-bots” (mechanisms that generate clicks fraudulently), which AdWords is not always able to intercept.
The American Associations of Advertisers estimated in 2015 that about 10% of display clicks were fraudulent (and almost 25% of video clicks). If you aren’t ready to accept these numbers, may be display campaigns are not for you, since it’s technically impossible to completely eliminate the problem, although probably Google has one of the best platforms at it.
In the case of potential automatic clicks, it’s better to block the site/app through a negative list linked to all campaigns.
On the contrary in the case of very relevant sites we could consider to increase the placement CPC or isolate it in another group or specific campaign which is absolutely a best practice in my opinion, because enables us to consider also Bounce Rate on the landing site (which is available only at group level).
BR can be imported automatically from Analytics, if accounts are connected on both sides and the source View has been selected.
To link AdWords and Analytics, you need to first select the Google Ads account in the Analytics Property (accessing it with the same admin account used for campaigns) and then, in AdWords, select which View you want to import data from (you can choose only one of them). Of course it is better to pick the most comprehensive raw data source available (if you didn’t modify it, the “all web site data” default view will be fine), not exclude data which may be filtered out from more selective views.
Once we import BR and add the data in the custom column, it should be quite easy to judge whether the site or the app on which our ads are running is bringing us users who are really interested in our offer or not (another little prayer of the evening for Google is to let BR data be available at placement level).
If a placement CPC is unreasonably high, we should consider mainly its results in terms of cost per conversion, rather than the general efficiency or QS of campaigns. If it has a positive ROI it definitely deserves to live and prosper, otherwise it deserves to survive only if it is vital for our brand positioning or its budget spending is acceptable in a more global consideration.
The last suggestion regarding the optimization of efficiency is to verify the distribution of clicks by type of device (desktop/tablet/mobile) and by geographical areas (to evaluate the possible exclusion of under-performing areas).
To optimize a GDN campaign, besides having implemented all additional settings to limit junk placements, we will need to regularly check the most expensive placements, the ones with more impressions and with the highest or lowest CTR.
There is a strong difference in campaigns targeting users and content context, we have to optimize them in different ways, looking mainly to conversion costs in the first case and to campaign’s efficiency in the second.
The most granular group setting is, the best data insights we will have, because this way we can evaluate specific Bounce Rates too.
In any case, as for all the other parameters being taken into consideration, no decisions should be made regarding any pauses or exclusions from the campaigns when we haven’t seen statistically significant numbers (at least a few hundred visits).