Return on ad spend (or ROAS), is a topic we come back to time and time again at Upp., and for good reason. Having control of your ROAS is the biggest determiner of commercial success in retail.
Machine learning technology can drastically improve ROAS for retailers, and this technology already exists and can be bought today.
Google began embedding ML into its software twenty-two years ago, making Google Shopping a potent tool for retailers to find customers for their products. The platform is the best in the world at matching products (or results) to search queries and is getting smarter every day.
The problem for retailers is that everyone has access to Google technologies – there is no inherent competitive advantage to using the tool. Instead, the gains are to be found on the other side – your side, the retailers’ side – and what data you give to Google for bidding on products. The gains rely on how intelligent and organised that data, from you to Google, is.
In this article, we delve into the detail of Google’s real-time bidding (RTB) and machine learning (ML) technology, and how additional machine learning technologies can supercharge your ROAS on Google’s Shopping platform to give you that competitive edge and improve profit margins.
Understanding real-time bidding
Real-time bidding (RTB) is when advertisers bid for ad placements on a per-impression basis, all in the blink of an eye. Similar to financial trading markets, it’s governed by algorithms and high-speed technology.
In the case of Google Shopping, this means that each time a buyer searches for a product, Google evaluates advertisers’ bids to determine which products get displayed most prominently. The goal for a retailer is clear: secure prime placements whilst minimising wasted ad spend.
The relationship between machine learning and real-time bidding in Google Shopping
Whilst real-time bidding might be a lightning-fast process, speed isn’t everything. It needs to be smart and fast.
Google’s machine learning technology is some of the most powerful in the world. Within Shopping, it analyses an enormous volume of data in real-time, considering factors like user intent, product relevance, historical performance, and even sales velocity to improve the likelihood of a bid turning into a sale.
The algorithms make highly informed, complex decisions instantaneously, ensuring that the right bid is placed for the right buyer at the right time. But it can only use the data it has available – your target ROAS, or tROAS.
If the ROAS data Google is using isn’t connected to your business objectives, and is purely looking at a siloed advertising number, the impact can be catastrophic.
For example, if you have a product in a Google Shopping campaign which has a high return rate, but is also in high demand, Google will likely put more advertising spend behind that product than your business can afford, and you’ll end up making a loss on that line, despite selling a large volume. Your ROAS might be on target, but the impact on your business is completely misaligned.
If you could tell Google where your break-even point sits for every product, considering factors like real-time returns rates, delivery and warehousing costs by SKU, you’d prevent this from happening and protect your margins.
Google Merchant Centre (GMC) does allow you to input much more detail than a standardised ROAS for your inventory, but it falls short of allowing the above – and most retailers don’t spend the time it takes to input the data as they know that Google does not (yet) take bidding action on specifics, such as items on promotion.
How additional machine learning tech can supercharge your ROAS
Imagine a world where Google was built just for your business. It knew all of your operational costs, your sales and marketing data, and your business revenue and growth targets.
Imagine it could take that information and buy/sell as needed, increasing Google Shopping bids when demand spikes (i.e. your back-of-the-shelf product goes viral on TikTok overnight), and not wasting ad spend on products that are out of stock.
Imagine you could take the cap off your marketing budget – because you have a true ROAS calculation that you can trust – and keep investing in positive growth.
That’s exactly what Upp.AI does.
Upp.AI knows that no product is equal and that supply and demand change constantly. So it takes an entirely different approach to Google Shopping campaign set-up, management and optimisation.
Custom-programmed for retailers, the platform calculates sales velocity, assesses ROAS data, and all the other factors that affect how much return a retailer sees from a product. Using this data, Upp.AI then groups each product into Google Shopping Performance Max campaigns according to its performance potential – not product categories as is the usual practice.
If a retailer gives Upp. the contribution margin needed for products in their inventory, Upp.AI is able to allocate the optimum budget for each product and assigns each one with an intelligently calculated ROAS target – aka iROAS™.
The product grouping, budget allocation and iROAS™ target-setting are all dynamic and change in real time in line with supply and demand.
Budgets are moved based on real-time auction dynamics and inventory intelligence, which prevents much of the over- and under-spend seen in Google Shopping campaigns today.
Time to upgrade beyond the limitations of Google Shopping
Despite advancements in machine learning technology by advertising platforms, the ROAS feedback loop is physically broken. Google is blind to what’s actually going on in your business. Its powerful ML is learning based on a small amount of data – almost entirely consumer data, plus your target ROAS – but not what is actually happening inside of your business and where your costs are attributed.
If your machine learning tech stack is restricted to Google Shopping, you’ve got no competitive advantage. You’re likely stuck in a bidding war over products that produce very little profitable return for your business.
However, by layering additional machine learning technology like Upp.AI, you will supercharge your Google Shopping campaigns, gain a significant edge over your competitors, and improve profit margins – essential with the peak sales period around the corner.
Today, Upp.AI analyses 7bn data points every day to make intelligent, split-second decisions that maximise Google Shopping ROAS for brands including Charles Tyrwhitt, Poundshop and Roman. Now that’s a machine worth learning about.