Morphio blog Post

4 Reasons Why Every Agency Needs to Use Anomaly Detection in 2019

By Eric Vardon

It wasn’t too long ago that you would hear the term AI and think of a futuristic concept. Now it’s a household word.

In fact, 28% of society agrees that machine learning has had a significant impact on their day to day life.

One industry that has been transformed the most by AI and machine learning is business itself.

Artificial intelligence is capable of improving conversion rates, detecting unseen errors, and helping agencies find the best possible solutions for their problems.

This is exactly why 43% of marketers are using machine learning to reap these benefits. Are you?

If not, don’t worry. 

Today we will be teaching you how one of the main concepts in machine learning, anomaly detection, can help your agency achieve higher conversion rates.

Let’s dive in.

What is anomaly detection?

Anomaly detection is a process in machine learning that detects errors and unusual events within sets of data it analyzes. Both historical and predictive data is used to discover anomalies that may be worth your interest. Anomalies can be both good and bad.

Often these are bugs and problems an agency to can fix to increase revenue or improve their client’s experience, but often they go unfixed until found manually. Not good.

Most data that machine learning software will collect falls into standard deviations like so:

When a particular piece of data stands out from these deviations, users are alerted so they can begin investigating.

Anomalies can be found in anything from website metrics, ad spend, organic search terms, and more.

Imagine being able to find which ads are generating you the most revenue or what keywords drive the most traffic to your website and you didn’t have to lift a finger. That’s possible thanks to anomaly detection doing all of the heavy lifting.

This brings me to my next point.

How anomaly detection can increase conversion rates

Anomaly detection helps agencies discover subtle events that are affecting their revenue, advertising, marketing, and funnels, and more. 

It isn’t always easing come up with new content ideas, predicting website bugs, and finding what advertising campaign deserves the most ad dollars.

That’s where machine learning steps in to bridge the gap and help you with these areas.

The following are some of the main ways it will help increase your agency’s conversion rates.

Find glitches before they create trouble

Have you ever discovered a glitch on your website by accident or from a customer informing you? Who knows how long it existed and the potential clients it lost you. And the potential loss of revenue they may face. 

Software bugs cost companies in the United States $1.1 trillion dollars in 2018. It’s no joke.

Anomaly detection, however, can find glitches in your website, code, advertising campaigns, and other assets which may hurt conversion rates and revenue. Finding these ahead of time allows an agency to boost performance and prevent further issues.

You can think of the process like this:

For example, you may notice that a certain ad set is taking up most of your ad budget out of the blue. This seems strange, so you investigate and find that the settings weren’t optimized and you were spending too much. You may have wasted hundreds to thousands of dollars without anomaly detection to warn you.

On the bright side, machine learning can also find properties which are excelling

That means that not all anomalies are bad. In fact, many of them are signals that certain networks, ads, or keywords are outperforming the rest. You can use this information to improve the performance of different strategies that you may not have discovered otherwise.

Take Dell for example. Back in 2016 they were struggling to figure out why email engagement was declining. Do you know what they did to solve their issue? Dell used machine learning to find out why.

Thanks to AI software, Dell was able to enhance the copy for each of its segments through automated testing to achieve:

  • 50% increase in click-through rate.
  • 46% increase in conversion rates from email marketing.
  • 77% increase in how many customers added to cart.
  • 22% increase in page visits.

Don’t you want numbers like that?

Filter out poor performing ads

The average business makes $2 for every $1 they spend on Google Ads. You’d be crazy not to take advantage of that! However, finding the perfect combination of copy, creatives, and targeting is easier said than done.

Since machine learning integrates with display ads, anomalies in advertisements, ad sets, creatives, and other components can be found to help agencies double down on the best performing ones.

The ads which are receiving low amounts of clicks, conversions, and impressions can be filtered out to leave the high performing assets to shine.

You can also dissect copy, audiences, and schedules are making these ads perform so well to implement those elements in future ones. 

A great example of this is Hiscox, a business insurance company, who were experiencing costly PPC campaigns they knew had to be improved. 

Hiscox wanted to remain competitive in their industry while increasing PPC conversions and decreasing cost per click. They opted for Google’s very own machine learning based products like Dynamic Search Ads, Smart Bidding, and Smart Lists.

The results? An impressive 8% increase in click-through rate, 62% increase in conversions for Smart Lists, and 39% cost reduction.

The Head of Acquisition Marketing, Stuary Mahoney, commented on this by saying “Machine learning-based tools have proven particularly invaluable in keeping cost per action down, whilst also giving us great insights into our own users’ search behaviour.”

It’s not just about campaigns, though. Personalization is on the rise and users want advertisements which are tailored to them specifically.

In reality, 4 out of 5 consumers want to see search ads which are customized to their exact location.

Machine learning will be able to help your agency achieve this by analyzing what cities and countries make up the bulk of successful campaigns.

You can then refocus campaigns on these geographics and use tools like Google’s Dynamic Search Ads or others to continually refine them.

Focus on the highest yield social networks

Do you use social media and digital marketing for your agency? Of course you do. 90% of businesses use social media to engage with customers, build loyalty, and drive sales.

But, do you know which networks are driving the most traffic, revenue, and results for your agency? If you said no, you’re essentially marketing in the dark.

For example, LinkedIn, Twitter, and YouTube have been found to be the most effective B2B platforms.

Information like this is very useful, but you’ll never know if Facebook or Instagram out perform the others for your individual business without testing.

While anomaly detection normally involves finding bugs and improvements, it can also open the door to discovering brand building opportunities like where your agency receives the most social mentions, social shares, etc.

Machine learning software will integrate with the networks you’re active on, discovering which offers the most engagement, traffic, and conversions.

Agencies can then effectively put more ad spend and marketing efforts into the platforms that are worthwhile versus investing into lackluster ones.

Boost your organic results

Did you know that leads from search engines convert at a rate of 14%? That means that optimizing your website and content for search engines needs to be high priority for your agency. Who doesn’t want high quality clients coming to their doorstep naturally?

Nonetheless, plenty of agencies use SEO but they miss out on search term and content opportunities. It’s not their fault, though. They simply don’t have the means to identify them.

That’s where anomaly detection comes in. It’s capable of finding which search terms are driving the most results for your website, allowing you to target similar ones in the future.

Similarly, you’ll be able to discover which search terms and topics aren’t moving the needle.

This data can be also used to improve title tags, meta descriptions, content, and other elements that affect search engine listings.

Wrapping up anomaly detection

Excelling in business is all about staying up to date on trends and machine learning is the biggest wave you need to catch.

It will save you heaps of time and energy trying to debug ads, social networks, and campaigns. In fact, anomaly detection can actually increase conversion rates by presenting you opportunities you would’ve otherwise missed!

It’s capable of doing this by collecting data on your marketing campaigns and organizing them into different deviations. Data which acts out the normal range is singled out and you get notified.

Agencies can then investigate why the anomaly was detected, allowing them to fix problems before it’s too late or double-down on revenue generating opportunities. 

If you’re interested in seeing how anomaly detection can grow your agency firsthand, try out Morphio’s 14 free trial

Eric Vardon

CEO, Co-Founder @ Morphio