the importance of first-party data and lookalike audiences – Econsultancy

Fuelled by the availability of large data sets, cheap cloud computing power and growing confidence among marketers, machine learning is radically changing the marketing function and the way advertisers can reach their audiences.

As with any new technology, there are early adopters and those that follow. So how are forward-thinking brands putting machine learning to use in their organisations today? How can the long tail of marketers learn from their peers and how can the industry separate high performance from hype?

Fresh is best

Machine learning models are only as good as the data that powers them. The ruthless ability for machines to pursue the goals set for them presents a huge potential scale that humans can’t match. Machines are able to tirelessly improve, iterate and adapt to a vast set of signals in order to achieve the best possible outcome. Manual campaign-targeting techniques requiring people tweaking bids and targeting behind the scenes simply can’t keep up.

This dramatic increase in the power and speed of campaign adjustments means more accurate and effective online advertising.

But that also means that marketers need to be sure of the quality of the data that goes into the machine. Supercharged models guided by bad data may end up in a completely different end-point than a marketer intended. Just as machine learning can drive incredibly positive results in campaign accuracy and effectiveness, it can have an equal and opposite effect if the wrong data is used.

The most important thing marketers can do to guard against this is to leverage first-party (owned) data. The most successful marketers today are investing in building their own first-party datasets, partnering with providers who are able to inform their campaigns with real-time data, or both, which cannot be achieved with static third-party data.

From announcements of industrial action by airline staff and unusual weather patterns to surprise album drops by best-selling artists and breaking political news, there are a huge and growing number of forces exerting themselves on consumer behavior.

For example, the announcement of industrial action on the London Underground may lead to a rapid increase in demand for conferencing software among commuters looking for a means to stay connected with the office remotely. This emerging trend can be fed into campaign targeting immediately and automatically offers a tremendous opportunity to serve similar users an ad and win new business. However, this insight two weeks later – when bought from a third party data broker – is close to worthless and would lead to inaccurate targeting and wasted budget. Just ask anyone who’s been followed across the internet by a pair of shoes they bought last month.

READ ALSO  Google AdSense Native Ads Could Ultimately Lead to Better Browsing

A Marketer’s Guide to AI and Machine Learning

Ensuring that ads are precisely targeted and relevant not only boosts campaign effectiveness, it also allows brands to serve more ads for the same budget while ensuring each one is more likely to reach a relevant audience with the right frequency. This helps avoid a negative impact on brand perception due to ad bombardment.

Beyond retargeting

There’s no shortage of companies out there able to help brands reach their website visitors across the internet through ‘retargeting’. However, if a potential customer has visited your site there’s a good chance they’ve also visited those of your competitors (shopping around is quick and cheap for online consumers). This means that your retargeting campaign is likely to be competing with the campaigns of your competitors who are also scrabbling to draw the consumer back to their sites to complete the purchase.

Taking this retargeting-only approach also automatically restricts your reach to a fraction of online consumers (for most brands there will always be more prospective website visitors than existing ones).

The real power of machine learning is to use your existing site visitors and customers to create models to discover lookalike audiences to find other consumers who haven’t yet visited your site. For example, by understanding more about the consumers that are completing purchases on an online retailer’s website, machine learning can identify key patterns in browsing behaviour among the cookie IDs of those customers to map the wider online audience.

Combined, these patterns provide a data-driven way to identify which internet users are more or less likely to become customers, statistically speaking, and then that information can be used to automatically optimise campaign targeting and bidding levels.

READ ALSO  News Roundup 22nd February 2019

Take internet-user Hannah, for example. She may start her day by getting up to speed with the news on her phone during her commute into work. At this point, a machine learning algorithm wouldn’t necessarily prioritise her cookie ID to begin bidding to serve impressions for, say, a gym membership. But later that evening, when Hannah gets home and starts looking for healthy recipes and sportswear on her phone, these signals could align with the customers who recently purchased memberships to the gym, enough to make sense for the algorithm to begin serving her ads with a membership offer.

By using these techniques brands are able to reach a far larger potential audience than if they simply targeted their existing website visitors.

Harder, better, faster, stronger

By leveraging fresh data and looking beyond existing site traffic, machine learning is able to make a marketer’s ad spend work harder, achieve better results faster and deliver a stronger return on investment. Lowering cost-per-acquisition (CPA) for each new customer allows marketers to spend the same budget but win more customers.

Automation can undeniably drive efficiencies. Yet one of the challenges faced by brands and the agencies they work with is learning to trust in the technology’s decision-making capability. Understandably, marketers are reluctant to relinquish control over their campaigns entirely to automation. It is therefore incumbent on data and tech providers to not only educate their clients on how their technology works, but also offer enough insight into live campaigns to allow for trust in the technology to grow. As ever, the way forward is a compromise: between appreciating the power of the machine on the marketer’s side and investing the necessary resources to keep the client in the loop on the tech side.

What’s more, by embracing the opportunities that machine learning provides, brands can automate a considerable amount of time-consuming tasks in their day-to-day, enabling them to plug back into the ‘bread and butter’ of their role and focus their efforts on strategy, planning and creativity.

Six case studies of machine-learning powered email marketing

Source link

WP Twitter Auto Publish Powered By :