Segment, increment and stand out with a new breed of marketing tools using machine learning

Marketers want to be in a position to confidently say: “If we spend X then we’ll increase sales by Y.” But while the need to understand performance and optimize strategy is there, many just don’t have the right data and tools to tell them what’s working and what’s not – especially when they’re bidding on users in real time.

This is what SAP Hybris Customer Attribution, formerly Abakus, was set up to do. To help people drive revenue and optimization through multi-touch measurement. The fuel for this engine? Machine learning technology built on Nobel prize-winning collaborative game theory. This patented technology
creates serious firepower in the form of advanced analytics, and it’s brilliant.

Rather than aiming for gains you can already achieve, the point is to understand the impact of all marketing activities on your customers – and to use that insight to enable dramatic improvement.

From its start-up origins, Abakus has now become SAP Hybris Customer Attribution. It takes us astep closer to our end-goal: closed loop marketing. A hands-off, intelligent marketing system where everything is automated and you have a known single view of the customer across devices.

Play the game: How machine learning can help you find the right customers

Making a good impression on the right customers is hard. The marketplaces we operate in are increasingly crowded and disrupted. Every day the competition comes up with a new idea or a faster, more agile rival appears. It’s noisy, too. Your banner ad might be one of over 1,700 your customers are exposed to in any given month. Perhaps this noise is a contributing factor behind a near 10% year-on-year drop in Facebook’s active engagement.

Then there’s the fact that word of mouth still rules. Research from Forrester shows that a third of online consumers will trust a stranger’s opinion on a blog or public forum more than they trust branded advertising. And 47% of people go out of their way to avoid online ads altogether.

Your tactics need to reflect the fact that inundating consumers with messaging will not work. To make it harder, much of what many companies sell has been reduced to a product, with heritage, loyalty and brand promise counting for less than ever. Mass marketing is costly to you and tiresome for the consumer. Today, you need to be more targeted than ever to reach the right people and you need to know what will reach them.


Or, to be precise, collaborative game theory. In its original format, it’s a way of ascribing value to the players of a game, enabling them to share the winnings. The winnings are proportional to each player’s value – the classic example is of two card players who are breaking even until a third arrives. Working together, the three clean up. So how do they split the winnings? Game theory works out the value each brings to the table and that determines how much money they get. This value is called the Shapley Value (see panel below).

In marketing, game theory can be used to ascribe value to your marketing efforts by looking at their impact. This is a data driven approach, not a model. It calculates the value of each touch point on the customer journey, giving real detail about their worth and effectiveness. That allows marketers to figure out what really works and, just as crucially, what doesn’t. It also means you can work out who would have bought your products anyway, without needing to be persuaded, and who purchased because they were influenced by your campaigns. It’s a much more sophisticated method than last click or first click attribution, which ignore most of the customer journey. Game theory attribution looks at the whole journey creating a much more accurate picture of marketing performance, what customers like and what makes them buy.

When you know the true value of marketing you can make data driven decisions. Imagine looking at 100 new customers, knowing that without marketing you would have acquired 35 customers (i.e. fractional value from each of the 100 customers). That means you know that your marketing had an impact on the remaining 65 net new customers. And you know the value each marketing touch point had on every one of those 100 customers. That represents the true incremental value of marketing, and now you can focus on those marketing efforts – the ones that are making a difference.


If this all sounds too good to be true then you’re on the first steps towards making it a success. Because while attribution software will increase your knowledge, you and your organisation need to know what to do with it. Attribution software is the means. You need to find the end.

It’s not uncommon for organisations to buy into new technology and start using it without knowing what to do with the data and knowledge it provides them with. The intention is there, but there is no plan of action. You buy attribution software, you figure out what works and what doesn’t – and then what? Carry on collecting data and analysing it? That, in itself, is not going to increase revenue.

Taking the next step is one of the harder moves, but here’s an idea, and it goes beyond the immediate norms of attribution software. Use the opportunity
to work out what sort of experience you want those people who have come to you because of your marketing. Because incremental sales are the growth. You need to keep them coming and they require a combination of marketing channels and tactics. To do this will require looking at your entire customer experience and to do that, you cannot treat any part in isolation. You have to see it as a whole.

Such a reassessment is likely to be a major undertaking. But it should have a positive effect on the way you operate, making every part of the customer journey complementary and giving every customer a more relevant experience.


That’s just the beginning. The potential for attribution software to disrupt the normal way of doing things is huge. You could use it to redefine your relationships with your media partners. If, for instance, they’re serving ads that have little, or no, value, then you can stop rewarding them and start working towards incremental sales. You may well find that the partners who you thought were top performers are not driving value. They will have to change the way they work, and that will benefit your business.

But it won’t just be your partners who will feel the effects, and you need to be prepared to deal with the ramifications within your own organisation. That’s not to say this is a destructive process – in fact, it’s just the opposite, affording a chance to improve the way you work– but navigating the correct path may require care, thought and negotiation.

Redefining the cost of sales is a good example. Traditionally cost of sale is your total investment divided by the total revenue attributed to those tactics – but what about an ‘incremental cost of sale’? Same investment but you’d only take credit for that percentage of revenue that was incremental, and that’s going to be lower. Suddenly the cost of sale has increased and although it’s a more accurate figure, it’s not necessarily a positive message. When you start to deal with these sort of metrics, tact and patience may be required to make the message appealing rather than depressing.

You may need a method to introduce these ideas to your organisation, to get them accepted as part of the culture. Describe how they will be used. Talk to people about the effects. Once you have people on side, you can start to build exciting campaigns and experiences that are aimed at growing your business.


Convincing people in your organisation that this is a good idea is going to take more than anexplanation of how it works. And most people in your organisation probably don’t care how it works. They want to know about the results it gets and how it will make the business better. Marketers exist in a specialised world that can be tricky to explain, so you need to imagine getting this idea across to someonethat isn’t directly involved with using attribution software. What’s the KPI that will appeal to them and demonstrate its worth? So you will need to find a solid result, something that software has changed and improved, and use that to prove its efficacy. It could be the number of new customers that your all new and improved customer experience is attracting. It seems obvious but the irony of marketing is that while we’re very good at talking to customers, we’re often less proficient at convincing our own people of our worth. And if one of the fringe benefits of attribution software is to improve how we talk to each other inside our organisations, then that’s money well spent.

Making better marketing decisions:
The four values of disruption

Insights are one thing – but it’s when you put them into action that you are able to disrupt an industry and drive positive change. Hot on the heels of theory come case studies and examples. Here, the founding team behind SAP Hybris Customer Attribution (formerly Abakus) walk us through their four key values of disruption – and offer a real world example for each one that shows the power of SAP Hybris Customer Attribution at work.


People don’t buy what or how you do, they buy why you do it. As the creators of SAP Hybris Customer Attribution, the ‘why’ is our motivation. For us, the ‘why’ of our product was that marketing measurement was broken. It’s inaccurate and backward looking. We wanted to fix that.

Example: One of the largest car rental providers, this company adopted global multi-touch attribution, but while it was effective at getting customers to come back, it struggled with net new. We were able to show it that by using measurement software that segmented for new customers you can decrease cost per new customer and drive more net new.


Culture is critical. The mindset of the team has to be focused on big steps, not incremental gains, so look for audacious goals.

Example: We implemented daily attribution of marketing on customers for this home insurance company, setting bold improvement goals for their display marketing. Over two months they were able to grow sales by 84% while reducing display spend by 48%.


Any task that is repeated more than once is a candidate for automation. But that requires a closed loop system that can include machine learning and intelligence to identify and resolve issues. This is something we have delivered with a system that is automated end-to-end, and these new tools can be used to optimize marketing hands off.

Example: As an online retailer, is not burdened by legacy systems. Its marketing technology stack represents the latest best of breed solutions.

We organize their data into different taxonomies, such as mappings for different products, bring in cost data and automate measurement and optimization. It’s
a great way to take unstructured data and make real improvements to marketing efficiency. Jet delivered a 23% increase in new customer activations, along with
a 24% reduction in cost.


If your goal is disruption, you need to hit the ground running with a minimum viable product (MVP), launching it as soon as customers can get value
from it. It’s a difficult value to achieve in practice. There’s two sides to an MVP. You might do it too early and have it fail, or on the other hand, you might
over design the product and so spend too long on it, missing the opportunity to get feedback.

Example: This global hotel group wanted to focus on engaging existing customers, but with nine brands and more than 3,000 hotels around the world,
previous measurement attempts fell short. We set up proof of concepts to show an increase in bookings from display remarketing to existing customers. We
increased bookings 40% and decreased cost per booking 12% over six months.