In our previous article we discussed the role & purpose of data within modern day business environments. The role of data is to be transformed into insights, in order to build better products & experiences which in return lead to business growth. We then further clarified how we could turn data, also known as the new oil, into its end product being the insights. In order to do so, today’s volumes of data require additional computing power & intelligence. This comes in a form of Artificial Intelligence known as Machine Learning. The cloud serves as the pipeline in order to make data accessible by these ML algorithms.

We concluded that if data is the new oil, AI would serve as its refinery and the cloud is the pipeline connecting both.

An architecture blueprint

In order to make this setup tangible, we’ll start by drawing a blueprint. Note that this is not a representation of what things should exactly look like in all cases. This blueprint serves as a reference in order to develop the overall idea and map the roles and interactions between the key elements of the infrastructure from a marketing perspective & purpose.. The following blueprint example shows a reference infrastructure. As indicated in the model representing the role & impact of analytics, it represents an infinite iteration. Full lines represent an influx that collects all the data and brings it to the end-point being the Cloud solution where ML is deployed. Dotted lines represent the enriched activation track where enriched signals are being ported back into the respective platforms to act on the derived insights.

The main pillars within the blueprint are:

1. Paid & owned (media) channels: These are represented by SEA (or), Search360 & DV360 as well as other web sources (read, Facebook, Twitter, etc). They need to be part of this ecosystem, otherwise we can’t act upon the insights derived from the raw data which blocks us to act upon them.
2. A web data ecosystem: Linked through elements such as a datalayer, tags, a Tag Management System and Data Quality Assurance Platform. Data Quality Assurance Platforms such as ObservePoint & HubScan monitor & safeguard your data quality through tag deployment & data streams on a permanent basis. A web analytics system, such as Google Analytics 360 completes this setup.
3. A unification engine: The role of a unification engine is to combine all data points around a single identifier. This could be represented by Tealium AudienceStream, Ysance or Segment. These platforms are known under different names such as Customer Data Infrastructure, Customer Data Platforms & Retail Data Platforms.
4. A cloud solution: The role of the Cloud solution is to centralise all data storage (data lake like) required for analysis and make it available to Artificial Intelligence / Machine Learning engines. Moreover it allows organisations to own their data as such outside of the platforms they are originally collected in.

From idea to architectural blueprint

Now let’s move from collection to refinement.

Step 1 - Data Collection

The elements depicted in blue in the below blueprint represent all elements linked to the data collection step. Data is picked up on by the Tag Management System. The Data Quality Assurance Platforms monitors the fluxes in place & warns the infrastructure owner when something is off. Data is piped through to the respected platforms & centrally collected by the analytics platform after which it is enriched within the unification engine by external sources. 

Step 2 - Data analysis

Analysis of the data happens in two area’s. For web-related matters only the analytics platform can be consulted directly. This is where questions regarding visitor counts, average order value & traffic sources are answered. More advanced analysis takes place within the cloud platform after the unification engine came into play. Here analysis regarding Recency Frequency Monetary modelling, churn & consumer classification is handled based on a unique identifier (typically a globally shared userId). In a first stage this is a purely descriptive & exploratory model which can evolve towards predictive modelling. The cloud serves as a gateway to pre-trained Machine Learning models which allow for further analysis than capable by humans. 

Step 3 - Better products & experiences

Based on the insights derived from the analysis the organisation can start building better products & experiences (in the digital space). This could be represented by improved website funnels, website personalisation, more qualitative media audiences or better segmentation clusters upon which teams can act more precisely. 

Start building

As the need for personalised & tailored experiences increases, the volume of data generated by users increases, the pressure on ROI increases, the cost of data storage & advanced analysis decreases, blueprints as the one explained in this article certainly show the way towards more customer-centric solutions. This again is an end-goal (& new starting point) for business growth. The most important element is to build the ecosystem step-by-step, keeping the overall objective and blueprint in mind. This shows the role of ML & Cloud versus marketing is no longer theoretical, but ready to be put into practice.

{snippet glennvanderlinden-en}

Data is the new oil. A tagline used over and over again to indicate the value & potential of raw data points as a resource for better decision making within business (& other) environments. While this might be true, the misconception with this tagline is that data as such is a product ready for use. Mind you that cars do not run on oil, they run on a refined form of oil - gasoline. That raises the question “how do you turn oil into gasoline” & what does this mean for the world of data?

The purpose of collecting data

First let’s ask ourselves the most important question before solving this riddle; if the role of gasoline is to power cars, which they in turn take us places, what is the role of data, what is its final product & what purpose does it serve?

The goal of collecting data is fairly simple.From raw data insights are derived in order to build better products & customer experiences. This in turn will grow the business. A growing business in turn allows for even more data collection & the cycle restarts. It’s a never ending self-feeding optimisation process.

How to turn data into gasoline?

Simple; take it to a refinery. How does this apply to the world of digital business intelligence? 

Insight from data comes through analysis. Where analysis used to be fairly simple as the number of data signals used to be fairly small in the past, today’s world brings us a heap of data impossible to analyze using only human capabilities. Today’s need to refine insights from data requires increasing computing power. This comes in the form of Artificial Intelligence. More specifically Machine Learning algorithms. Think of this as the refinery. Where the refinery turns oil into gasoline, AI will help turn data into insights. What should you imagine in terms of insights derived from running your data through these AI/ML algorithms? When it comes to analysing for marketing purposes a couple of examples come to mind. ML can help predict churn rates on top of your customer base, build conversion prediction models for customers based on website visits, classify users in Recency Frequency Monetary classifications (RFM) and as such pre-qualify 3rd party data.

But how do you get oil to a refinery? Through pipelines. Just as miles of pipelines are bringing raw oil into refineries in order to be transformed into gasoline, the cloud acts as a pipeline in order to get data into AI’s perimeter. Here data points are transformed into insights ready to be put into action. Cloud serves as the gateway to Artificial Intelligence & allows organisations to leverage existing pre-trained Machine Learning (and other) models without going through a huge sunk cost of building their own infrastructure.

Once refined, the insights are ready to be put into use & improve products & experiences across the board. The end-game of all of this? Growing your business.

Welcome to the new way of running analytics

Just like cars need oil to go through a pipeline & to be transformed in a refinery in order to move forward, businesses need their data to be transported into the cloud & analysed by AI in order to build better products & grow their business as a whole.

After having explained you in the Article 1 what is Smart Shopping, how it works and what it could bring you in terms of performance optimization. We’ll now guide you step by step in the set up of your Smart Shopping campaign. We’ll also share some recommendations that we hope will help you leverage the full potential of this new format for your business.

Requirements

On top of the basic requirements for standard shopping campaigns, in order to start running Smart Shopping you’ll need to fulfill the following requirements:

• Get at least 20 conversions over the last 45 days across existing Shopping campaigns
• Add a global site tag to your website and have a remarketing list of at least 100 active users associated with your account

Find detailed info here. For those who start with Google Shopping, it’s recommended to use first a standard shopping campaign with ‘maximize clicks’ bidding strategy to meet the conversion and remarketing requirements before running Smart Shopping campaigns.

How to set up a Smart Shopping campaign?

Setting up your first Smart Shopping campaign is quite straightforward if you are already working with standard Google Shopping (read below). If not, you can find all detailed steps here.

Step 1 - Configuration

In Google Ads interface after clicking on create a new campaign, under the ‘Sales’ objective, you first need to select your campaign type (Shopping), your Google Merchant Center account ID, the country of sale and check as campaign subtype the option ‘Smart Shopping campaign’.

Step 2 - Budget and bidding strategy

Afterwards, you need to define your daily budget and bidding strategy, and specify a target ROAS only if relevant to your approach.

Step 3 - Product groups

Step 4 - Upload your assets

Finally, you need to provide some assets, mainly a visual, two headlines and a description. Those will be automatically combined and turned into different responsive ad variants to be shown across a variety of placements, and based on Google’s machine learning those with the highest performance will show more often.

Note that these assets will be combined and shown to users who have visited your website but have not yet expressed interest in a specific product. Once user interest has been indicated, relevant data will be pulled from your product feed to create an ad.

For your information, text assets will be used in various combinations and ad formats as follows:

• The short headline (first line of your ad) appears in tight ad spaces where the long headline doesn't fit. It may appear with or without your description.
• The long headline (first line of your ad) appears instead of your short headline in larger ads. It may appear with or without your description.
• The description adds information to the headlines and should invite people to take action.

Recommendations

In terms of overall optimization and performance, it’s recommended to :

• Create only one Smart Shopping campaign targeting your whole product feed, as the more data the algorithm has to rely on, the easier and faster the optimization will be.
• Segment your campaign at the product groups level. This way afterwards you can still analyze your results in more details and evaluate your campaign performance across the different product groups.
• Evaluate performance after a minimum of 15 days, the approximate time the algorithm needs to learn and adapt.
• Use a pre/post testing framework by evaluating performance during similar time periods before and after you launched Smart Shopping, excluding the 15 days learning phase from your evaluation.
• Target ROAS vs Maximize Conversion Value strategies: if you have an extended product feed with many different products and you decide to opt for Target ROAS strategy, then it might be relevant in this case to segment your products in distinct campaigns based on their different ROAS targets. This to avoid missing out sale opportunities on products that clearly return a ROAS far from your target. It could be for example that some products are more competitive and then show a higher CPC, which returns lower ROAS for an equal sales volume. Other products might show a lower CPC or may be more expensive and then return a higher ROAS for an equal sales volume.

Conclusion

Smart Shopping is the new fully automated shopping format of Google that helps you optimize campaign management and performance. If you haven’t had the opportunity yet to test it, it’s not too late. Just define a budget, provide the required assets and Google algorithm will take care of the rest. Advertising your products on Google has never been easier. Let’s see what it will bring to your business!

{snippet morganederoy-en}

Showing up on top of the search results, visually appealing and displaying key product information in one shot, Google Shopping ads proved to be one of the most successful channels for brands to attract valuable consumers online at the very spot where most of them start looking for the products they want.

Beginning of 2018, it was reported that Google Shopping ads were driving 76.4% of retail search ad spend in the US, and 82% in the UK – accounting for more than 85% of clicks on Google ads between January and February 2018. Today, Google Shopping is a clear must for retailers aiming at boosting their online sales. For this reason, we expect this format to become even more competitive and more time- & resource-consuming in terms of optimization.

This is where Smart Shopping comes in. This new shopping ad format is leveraging the power of Google’s machine learning to automate your bidding and ad placement across the broader Google network and to optimize to get you the most value within your defined budget.

What is Smart Shopping and how does it work?

Smart Shopping is the latest Google shopping ad format that uses machine learning to completely automate your campaign. It lets the Google’s algorithm optimizing your bids, creatives and placements in order to maximize your conversion value at your given budget.

That way, this smarter format allows you to save time in campaign management, optimizes automatically towards your objectives (maximum revenue) and increases your reach by showing your best performing ads across different Google networks including the Google Search Network, the Google Display Network, YouTube, and Gmail.

Practically, it combines under one campaign 3 types of Google ad formats that you might have already used :

1. Standard shopping ads based on user’s search queries and your product feed
2. Dynamic remarketing based on which products users have seen on your website and your product feed
3. Static display remarketing targeting the users who have been to your website but haven’t yet expressed interest in a specific product. It will be based on the assets you will upload to set-up the campaign.

Unlike standard shopping, Smart Shopping uses the available budget as a starting element and intends (by testing & learning) to maximize your conversion value within that specific budget. Consequently, it will not be constrained due to budget limit and instead it will adjust your bidding based on various signals in order to maximize your return taking into account that threshold.

See below the set of signals Google’s machine learning system will take into account:

In the end, automation will help in a more efficient way to show the right product at the right time to consumers who are more likely to convert and spend more.

Performance - More than 20% increase in revenue on average

Based on Google data, advertisers using Smart Shopping campaigns have seen on average a more than 20% increase in revenue for the same cost. Based on our clients who started using Smart Shopping, we also observe similar positive trends in performance with a significant increase in revenue and decreasing CPA for a stable investment.

Conclusion

Shopping is today one of the most performing and most-used paid channels for retailers in order to increase their online sales & revenue. New Google Smart Shopping format offers a much more efficient way to manage your shopping campaigns. By leveraging Google’s machine learning and automation, it automatically optimizes your bidding, creatives and placements to meet your goal and maximize your revenue within your chosen budget. It saves you time, increases your reach across Google networks and is already showing great performance uplift compared to standard formats. Ready to test it ? You will find more details on how to set-up Smart Shopping campaign in our second article.

{snippet morganederoy-en}