8 min.
Blindfold: the Adviso secret weapon that will revolutionize your SEM campaigns
1L’art de la gestion de projet2Un projet à succès commence par une bonne gouvernance3Cascade, agilité, demandes de changement?
SeriesPart 1

Blindfold: the Adviso secret weapon that will revolutionize your SEM campaigns

Paid Media & SEM Innovation Secret weapons

Performance metrics have changed little since the emergence of the paid search channel nearly 20 years ago. SEM specialists want to maximize their volume of clicks as a proxy metric to conversion and at the same time, they want to decrease the cost per click; it's been the same fight since the very beginning.

For example, if your annual budget on AdWords is $1,000,000 per year and you manage to reduce your costs by 20% for the same volume of clicks, that represents a considerable savings of $200,000. It is in this constant quest for performance that Adviso develops and deploys innovative tools and solutions. This article explains how we have developed one of these internal tools capable of maintaining this high level of performance for our customers despite the degree of difficulty which has increased considerably over the last few years. A secret weapon, integrating data science, allows us to maximize the quality and precision of targeting, all in record time!



Since the year 2000, the difficulty curve has increased considerably to obtain such results. Competition is fiercer, auctions are saturated and Google's algorithms are more finicky. In fact, acquiring even 10% savings in CPC is now quite an achievement, especially for highly optimized accounts and on quota keywords.

In the past, the primary way to gain competitive advantage was to adopt bid management software like Marin Software, Kenshoo, or Adobe Media Optimizer (now Adobe Advertising Cloud). These tools allowed us to manage much more complex and granular bid formulas on huge volumes of keywords.

Certainly, these platforms remain relevant today, as do the good old instincts of specialists. However, this is no longer enough for the high-caliber specialist to bring significant added value. Any advertiser or agency with enough budget can afford a Google Marin or DS3; we therefore need to push innovation further.


Extracting the full potential in the data from your AdWords accounts, even with Marin or Kenshoo, requires a lot of specialist time. And time is precisely what we lack in a competitive environment in SEM.

This is why at Adviso, we have decided to develop another approach by seeking the contribution of science, data science to be more exact, in order to remedy this time issue.

From this exploration into the universe of APIs, Big Data, statistical models and programming in R and Python,  we have developed a series of tools that now help us to do analyzes in a few clicks that in the past not so far, would have taken us hours and hours of work.

The vast majority of the accounts on which we have carried out audits have obtained the same conclusion. We must move away from targeting broad queries towards exact queries in order to maximize the quality and precision of targeting.

It's obvious. But, the real issue you will encounter is the time you can invest in building and maintaining your structure. Using broader targeting allows you to cover a larger volume of queries while minimizing campaign creation time. Time that can thus be invested in the analysis and the search for new media opportunities.

The R&D team decided to attack this problem and we set out to create a tool that would attempt to do the laborious task of selecting the best exact keywords to add in the fastest way possible. he be.

The tool therefore needs to be able to traverse in real time the hundreds of thousands of user queries currently displayed in the query report for your entire account. Then, it must identify the most interesting words to add exactly while simultaneously offering to stop broader targeting that will no longer give you new interesting clicks.

So the manager can now focus on finding new targeting opportunities and hand over the hard work of prioritizing and learning about current targets to a series of algorithms.

We affectionately call this process the  blindfold You will discover by observing the following screenshot of the  Search Term Report  why this algorithm bears its name so well. 


1- The “Other search terms” line: a well-kept secret

I direct your attention to the last line of the report, it is very often ignored. And yet so important! The “Other search terms” line here contains 87% of impressions and 0% of clicks, so it is a banner that hides the most useful information for your optimization, namely: What are the less interesting queries that my targeting currently covers? .  Until now, the only solution used is to  manually add  interesting targeting in exact and thus slowly get rid of these less attractive impressions. 

2- The SKAG technique on steroids

In fact, if you only keep the top 7 keywords from the example above in exact targeting in separate groups, your new campaign would get 31% CTR! A superb result. This technique is known by the acronym SKAG,  Single Keywords Ad Groups,  but unfortunately it suffers from a limit. She asks for an expert who performs the task of adding the targets one by one, full time. Especially in large structures.

The blindfold therefore tries a similar approach, but with a slightly more  lean process . This requires combating the fact that if a query is searched for a lot, but isn't clicked on because it isn't relevant, you'll never be able to tell. It is therefore necessary to have a maximum of “visible” impressions (outside the banner) and always avoid broad targeting which generates large volumes of invisible impressions.

Simply, you can add the Keyword column in the Search Term Report. Then you combine the keyword level report and the query report into one. To do this, you can use a  Vlookup  in Excel or  Merge  in R Studio. Above all, keep in mind that several requests can accumulate for the same targeting. Doing this job manually can easily become a Herculean task, and if you have multiple campaigns and multiple accounts, feel free to use Google's API. Once the data is associated, you will get a decision grid similar to this:

You are now in a position to make an informed optimization decision. Here you leave the work of instinct based on the experience of the senior specialist and you get a working tool. It only remains to democratize it for all levels of experience and all customers, from the local pizzeria, to the international multi-MCC account.

3- The use of R statistical language: when data science meets SEM

To do this, we opted for R as a statistical language, using the Shiny R library to support the visual interface and the AdWords R library to communicate with the API.

We have created the following interface which allows to:

  1. Estimate the impact of the change
  2. Play with the selection criteria
  3. Get a patch file ready to import
Video player

In short, the Blindfold allows:

  • To obtain a more precise and better quality target 
  • Reduce the time specialists spend on tasks with less added value on current targeting
  • Give specialists more time to explore new targeting opportunities that can generate incremental performance.
  • Significantly increase the performance of your campaigns

If you want to build this kind of tool by yourself, I invite you to take inspiration from the screenshots and the video, as well as from the different libraries presented a little earlier. You have everything in hand to succeed.

For advertisers who would like an example of Blindfold to measure the potential gain this could have on their AdWords accounts internally or managed by your agency, our team is available to help you produce the analysis with our tool.

Happy SEM revolution to all!



For a real customer, does this automated method have a measurable impact?

We opted for a  Bayesian factor test  using the Causal Impact library to verify the three advanced hypotheses.

Hypothesis 1:  Reduction in print volume

The black line tells us that there was an 83% reduction in impressions on the test group. The dotted line tells us that the control group remained stable during the test period. The probability that the observed result is not due to chance is 99.9%, or P<0.001.

This confirms the hypothesis that putting the keywords in exact form greatly reduces the queries covered and drastically reduces the volume of impressions.

Hypothesis 2:  Growth of CTR

Here, we see a 32% improvement in CTR and a 99.9% probability of impact, thus confirming that the method reduces impressions and improves the average CTR of targeting.

Assumption 3:  Stable level of clicks


Result: the number of clicks varies very slightly.

Be careful, to succeed, we nevertheless had to improve our estimation service in order to select which keywords to work on as a priority, with some trial and error. We also realized that some broad keywords still contained interesting queries not yet visible in the report, which led us to adjust the algorithm accordingly so as not to lose these potential visits.

If you do this work over a long period of time, you will notice that as the  Search Exact Match Impression Share  increases, the average cost per click decreases. Thus, more precision increases your quality and reduces your overall cost without losing potential visitors! In the following example, the cumulative savings for the 14 weeks following the first 7 weeks of testing was $2,500. In other words, with $3,500 you get what would have cost the original cost per click $6,000.

The level of effort required?  Verify an interface and upload the revised change document. The time thus freed up can be used, among other things, to focus on the optimization of ad texts, too often neglected in the management of small accounts.

Search Exact Match Impression Share VS CPC

Investment VS Clicks