Kees Smit

More than 100% growth in omnichannel revenue

View the case

Smart shopping made even smarter

Kees Smit Tuinmeubelen has been leader in the market of garden furniture for quite some time. With its click share of 60%, Smart Shopping is a big driver of growth within the entire ads strategy. The control over search queries, product visibility and insight into revenue generating products has become more difficult with the use of social shopping. That is why we started looking for a solution to take smart shopping to a higher level. The objective was boosting revenue volume and improving the ROAS for running smart shopping campagnes. We approached this by combining our own unique product data with Google's machine learning algorithm. The results were overwhelming as omnichannel revenue doubled in one year's time. How we managed to accomplish that? Read it below!

This case was nominated for
Nomination Dutch Search Awards 2020 - Beste PPC B2C campaign Nomination Global Search Awards 2021 - Best use of search - B2C (PPC)


The power of machine learning within Smart Shopping .

Smart Shopping campaigns can no longer be ignored. Campaign management is greatly simplified as machine learning takes over the coordination of campaigns. Conversion value goes up, visibility within the search network improves and range is extended as a result of an increase in the number of views within the Google Display Network, Youtube and Gmail. For Kees Smit's growth strategy, smart shopping campaigns were the obvious option to go with as oppose to standard shopping campaigns. 

Smart shopping does have some disadvantages. Search query and audience data are not available and control over product visiblity is being limited to a minimum.  

Google recommends targeting all products in a single campaign in order to generate as much data as possible for smart bidding. This is quite a short-sighted approach because both low-value and high-value products get equal visibility in this situation. 

An innovative solution to this problem is combining the power of machine learning within smart shopping with control over product visibility. With the use of data modeling, we connect our own unique product data to Google's data in order to coordinate the smart shopping campaigns. 


Our approach

Feeding Google with our data .

Google collects a lot of data which is used to optimize smart shopping campaings. We are convinced that results can be much better when we feed Google with our own unique product data. The goal is to create an automation layering on top of existing smart shopping solutions. 

We retrieve all collected campaign data and then store it in a data environment. All available additional product data is then added from our own system. This method of using enriched data creates a more realistic view. Based on all available information, scores are calculated for every single product. Products are then assigned to product groups based on their score. This is done according to the statistically proven normal distribution and standard deviation, which is considered to be the arithmetic measure of the spread of numbers around the mean.

This standard deviation is used to determine the category limits for three category types : low performing products, average performing products and high performing products. This makes it easy to distinguish well-performing products from products which have not yet had the chance to perform. These results are then fed back to the Smart Shopping campaigns where products are automatically assigned to the correct campaign with the right target. 

Always dynamic product classification

Thanks to the algorithm .

Feeding back the results is done by using custom feed labels which are also used to assign the smart shopping campaigns to the different product categories. What makes this case distinguishing and innovative, apart from enriching Google's data with our own, is the dynamic product classification made possible by the algorithm. Product performance is constantly reassessed which could lead to products being assigned to a different campaign. That makes this method very scalable. Once the data environment, analysis and the data feedback process are set up, campaigns are automatically updated on a daily basis. The succes of this data modelling also manifests itself in the results of the different campaigns, which will be discussed in the next section.


in numbers.


Users +69%


e-commerce conversion percentage +28%


Transactions +101%


Omnichannel revenue +104%*


Omnichannel* ROAS +62%

Growth in perfomance

with succesful smart shopping campaigns and more control over product visibility.

This approach led to a significant performance growth. We compared results of the Smart Shopping campaigns with those of the year before. The period from which the results are taken was about 10 months (nov 19th until aug 20th). The results are exceptionally high and exceed set targets amply.

* Both online and offline revenue are included in the campaigns. That means that smart bidding strategies optimize on omnichannel revenue. 

Side note : this project took place during COVID-19. Many markets were turned upside down during this period, with one industry growing exponentially while the other collapsed. With 104% growth, Kees Smit's Smart Shopping campaigns performed exceptionally well as a result of the chosen strategy.

The most important objective was getting more control over product visibility in order to further optimize revenue and ROAS of the Smart Shopping campaigns. That has been achieved which resulted in a very satisfied client.  

Marijn ten Bulte Adwise

Want to know more about smart shopping campaigns?

Joris Kroes, digital strategist at Adwise, will tell you all about it.

Contact us
Emerce100 Best Fullservice Digital Agency '24

Best Fullservice Digital Agency '24