The secret to SEM success: machine learning
The need for strategy
The pressure for paid search to drive growth has never been greater. Stakeholders are demanding a more strategic approach regarding how their programs are managed. Organizations expect their digital marketing managers to understand how paid search ties into their overall business plans, how attribution influences the performance of other marketing channels, and all the while, continuing to prove that PPC is directly impacting revenue and profit.
Machine learning is critical for gaining deeper performance insights and using those insights to automatically optimize key account functions. Through thorough real time analysis of big data, machines can automatically adjust bids, create unique and powerful ad messages, and pinpoint target audiences based on factors such as devices their consuming content on, geo targeting, time of day, and even the operating system their customers are using.
Machines can collect and process information quicker than humans. However, to collect the right insights and optimize accordingly, humans must create the right strategies. In short, deploying machines without proper direction will lead to failed performance, wasted budget, and unhappy stakeholders.
What is strategy, and how is it created?
To create effective strategy, it’s imperative to understand what it is. According to the dictionary, strategy is defined as “a plan of action or policy designed to achieve a major or overall aim.” Strategy is not tactics, it’s bigger. Strategy is the path you’re going to take to reach goals.
Strategy can be broken down into a three-step process:
Step #1: Assess the overall situation
Assessing the overall situation will uncover the biggest challenges your business faces. The S.W.O.T. method is a popular process to use for uncovering what the overall situation is.
The S.W.O.T. method helps uncover what your business’s strengths, weaknesses, opportunities, and threats to it are. Understanding what your business is good at, not good at, what threatens its standing, and opportunities for growth provides clarity. Understanding the overall situation provides the necessary background and context to develop proper goals and objectives. Sharply defined goals and key performance indicators are the cornerstone of any successful PPC program. Every advertiser must have a clearly defined understanding of what success looks like to them.
Conducting the Situational Assessment
When conducting a situation assessment, perform the following steps:
Have a business download meeting
I meet with key stakeholders to learn everything I can about the business and the industry they compete in. Is the business itself growing or struggling? What about the organization’s performance compared to the overall industry? Use this time to learn what has worked and not worked from a PPC perspective and to determine if paid search goals are realistic and achievable (or if they exist at all).
Conduct a thorough account audit and gap analysis
Audits are time-consuming and tedious, but they’re necessary. Use the account audit to understand underlying performance drivers to determine whether work being done in an account is in alignment with business goals. Uncovering an account’s strengths, weaknesses, and opportunities provides the critical information needed to form a guiding principle for account management.
Competitive analysis
Stakeholders often misidentify PPC competition incorrectly. Competitive analysis helps uncover who the real competition is and what they are doing to be successful in terms of bidding, keyword targeting, and creative messaging. The competitive analysis helps inform the strategic direction needed to win in the marketplace.
Step #2: Develop an Account Management Governing Policy
Once the situational assessment has been completed and a firm grasp of the overall situation has been obtained, it’s time to develop a policy for managing the PPC account. Paid search management is a very iterative process. PPC professionals are often trained to manage accounts to “common practices” and in the process, hopefully achieving ever-improving results along the way.
Completing common practice tasks, such as negative matching and keyword expansion, are the right things to do for an account. However, there’s a proper place and time to do them. Without a policy governing how to manage the PPC account, we’re simply “stringing tactics together,” which creates a high probability that the work completed will not connect to results.
What should an account management policy look like? It can look something like this:
- We’re going to use PPC as a growth driver, and we’re going to manage the account accordingly.
- Paid search is going to be a profit driver, and we’re going to manage the account accordingly.
- Paid search is going to be a touch point leading to conversions through other channels, and we’re going to manage the account accordingly.
The policy should be written down for all stakeholders to see, agree, and abide by. Now that the governing principles are in place, we can build an action plan that reflects the governing policy.
Step #3: Creating an Effective Action Plan
The next step in the strategy development process is developing a coordinated action plan that’s designed to overcome the biggest challenges a business faces.
The action plan is higher-level in nature. For instance, let’s say you’re managing a PPC account for a college or vocational school. Schools are always in need of students. Based on this information, it can be surmised that the primary challenge is to find more potential students. Therefore, the policy could be to use paid search as a lead growth driver.
The action plan regarding this example would support the policy by focusing on market expansion and remarketing. The specific action plan for your account will most likely be more comprehensive, based specifically on your stakeholder’s overall situation and the policy you’ve decided to pursue.
The tactics used will now stem directly from the action plan. For instance, tactics for market expansion might be to set up Facebook and Twitter campaigns. The tactics for remarketing might be “set up an RLSA campaign focused on shopping cart abandoners” or “implement GA remarketing.”
The point is, action plans are designed based on a guiding principle, which is based on an assessment of the overall situation. Connecting action plans to your account management’s governing policy ensures work completed has the highest probability of helping performance reach established goals and objectives.
Evolve from PPC tactician to PPC strategist
Machines will not make the PPC professional’s job obsolete. However, the arrival of machine learning and its ability to automate key tasks will change the nature of the role. The value of the PPC professional going forward will be tied directly to their ability to be strategic. Those who can see the big picture, fully understand the marketplace the business operates in, and connect PPC to the overall growth of their organization’s entire digital marketing program will win the day. Machines can crunch data, spot trends, and execute routine tactical tasks much faster than humans ever could.
Listed below are a few tips that will help the PPC professional become more strategic, so they can create the strategies needed to drive success for their stakeholders.
- Have a firm grasp on the metric most important to your stakeholders and focus on it like a laser beam. For instance, if an account has dual goals like lead volume and CPA, strive to understand which goals are most important to the client. Doing this will reduce misalignment and also provide the understanding needed to collect the correct machine learning data and execute it properly.
- Understand all obstacles that could impede meeting or exceeding your stakeholder’s most important metric or goal. Is it competition? High CPC’s? Low conversion rates? Having a firm understanding of the obstacles will help guide planning.
Identifying technology resources & defining requirements
Now that the strategy has been created, it’s important to determine the resources needed to achieve goals. Depending on the account size, complexity, and budget, you’ll need to determine whether you need more people, technology, or a combination of both. Since the purpose of this eBook is to discuss machine learning and automation, we’re going to discuss how to best deploy technology to execute strategy.
Machine learning and automation currently play a big role and will play an even larger role in PPC going forward. Paid search and paid social functionality is more complex than ever. The number of platforms and networks that need to be managed is constantly expanding, and stakeholders are demanding deeper analyses and cross-channel insights. These dynamics put a strain on PPC managers who need to efficiently manage their accounts manually.
The right technology stack provides an ability to both automatically analyze large data sets and automate routine tasks to quickly solve complex problems. Deploying machine learning and automation to PPC accounts frees PPC managers. It allows them to focus on strategic planning and implementing strategic initiatives that lead to new growth opportunities.
What kind of technology should I use?
The answer to this question is, “It depends.” There are dozens of technological solutions on the market, ranging from lightweight reporting platforms to complex machine learning and automation solutions. Additionally, the advertising platforms offer automated bid management functionality and provide the ability to pause keywords, ad groups, and campaigns based on specifically defined criteria. Scripts can be deployed through Google Ads that allow paid search accounts to be integrated into an organization’s inventory or CRM system.
Budget and account size certainly play a large part in deciding whether to utilize a third-party technology solution or free tools the advertising platforms offer. It’s important to weigh cost vs. time saved to focus on big strategic initiatives.
When deciding whether it makes sense to deploy technology, consider the following:
- Can free automation tools help meet account goals and execute strategy effectively?
- Do paid tools offer the specific functionality that provide deeper performance insights that can’t be gained from other tools?
- Will you save a significant amount of time using technology vs. taking a manual approach?
Having defined criteria for using technology will make it easier to decide whether it’s worthwhile to invest in technology.
Selling the machine learning approach & demonstrating its value
A difficult obstacle is selling the machine learning approach to stakeholders. For instance, you’ll need to convince executive level decision-makers that machine learning technology will have a positive impact on revenue and profits. From an operational perspective, there might be a perception that introducing machines into the PPC program will make PPC managers’ jobs obsolete. With that perception, it’s more likely they will fight the need for machine learning and hamper its adoption. Therefore, as a digital marketing manager overseeing the paid search program, it will be incumbent upon you to sell the benefits of machine learning in a non-threatening way.
To successfully sell the machine learning approach, it’ll be important to demonstrate the value of it to your stakeholders. A few ways to explain the value of machine learning in the paid search space include the following:
- Deployment of machine learning would elevate your PPC team from a tactical team to a strategic force. It would connect paid search to master digital marketing and business strategy.
- Decisions would be made faster. Removing the “downtime” of pulling and organizing data would provide more time to spend on interpreting the data and planning the next steps for your organization.
- You could tap into the “invisible” part of PPC. Advertising platforms track thousands of data signals that can be leveraged for more precise targeting, bidding, and messaging. Only machines can track these signals and make higher-level optimizations needed to gain superior results and a competitive advantage over the competition.
Value is best demonstrated through results. To win over your stakeholders, it will be important to show that machine learning makes organizations smarter and faster, which results in superior performance. Having these arguments prepared and ready to go will help support the case for machine learning.
Listed below are a couple of methods for overcoming objections:
Educate your audience
Ensure you understand the big problem your stakeholders face, and educate them as to how machine learning will solve it.
Have a strong point of view
Many pitches fail when the presenter does not have a convincing point of view regarding the subject matter. The best way to develop a strong point of view is to do extensive background research, make solid projections on potential outcomes, and consult others to gain feedback about your plans and pitch. Based on that information, develop a strong point of view that can be communicated in way that garners trust and confidence in your plans for moving forward.
Overcoming objections is all about building credibility. Being fully prepared and confident in the information being presented can help reduce objections and lead to successful outcomes.
Approach for testing machine learning
Once you’ve earned approval to proceed with machine learning, the next step is testing. To properly use machine learning to your advantage, you’ll need a solid testing philosophy.
Examples of testing philosophies include the following:
- Where am I stuck and not advancing performance? Can machine learning offer the opportunity to test and learn in different ways that can lead to improved performance?
- Where are the knowledge or data gaps, and how are they affecting our ability to innovate? Machine learning affords the opportunity to provide answers to questions that can’t be answered through a manual approach. It leads to uncovering new insights, which can answer key questions that help move performance forward.
- Where is the PPC team bottlenecked and unable to scale? Where are they using a manual approach that is inhibiting their ability to perform? Machine learning is tied directly to automation. With the complexity of PPC, employing automation can free up time to focus on bigger projects that can have a bigger impact on performance.
Building a ppc strategy with machine learning included
Who should take care of keyword research?
64% Humans
36% Machines
Who should take care of PPC strategy?
93% Humans
7% Machines
Do you think machine learning is too costly for the value it delivers?
16% Yes
38% No
46% I’m Not Sure
Making machines work for ppc
Smart SEM, powerful ppc performance
Better performance begets even better performance. Marketers using artificially intelligent algorithms for campaign management find that performance improves continuously. That’s because machine learning algorithms can, and do, actually get smarter and inject that learning back into their performance. It’s that continuous learning aspect that is at the heart of an SEM marketer’s advantage, allowing those who leverage intelligent martech to be increasingly competitive.
Machines have an advantage over humans
We asked marketers who they thought could process more data at one time between machines or humans – 89% said machines.
Who can process more data at one time?
11% Humans
89% Machines
Humans are complex beings capable of processing immense amounts of data at all times within the human brain. Yet we cannot dedicate all of our neural connections to continuously improve PPC performance. Even if we could, we likely wouldn’t want to, so we created machines that could. Machine learning algorithms process more data than any human can at one time, using more data faster than ever before. Another advantage machine learning optimization has over the human counterpart is that machines never get tired. They go to work 24 hours a day, continuously bringing home results. From this perspective, machines are an obvious choice for certain aspects of campaign management.
Machines get great results for SEM
Bid and budget management is an area of campaign management that is often tedious; hence, the creation of Google Ads scripts and third-party management tools. The dynamic marketplace combined with changing campaigns amongst competitor actions makes for a fluctuating landscape. It would take constant attention to every detail of a PPC campaign. Also, not to mention, it would take an ongoing memory of the patterns within those details to execute the necessary changes continuously to optimize at the level of AI today.
We asked marketers who could manage bids and/or budgets best – 85% said machines.
Who can manage bids and budgets best?
15% Humans
85% Machines
When advertisers begin using machine learning in their accounts, the algorithms will learn from the historical data as well as all of the data in real-time moving forward to make decisions about bids or budget management. The results we’ve seen so far demonstrate that all of the data involved is definitely helping machines to make smarter bid and budget decisions within SEM accounts.
Acquisio TuringTM, Acquisio’s machine learning suite of algorithms, reduced cost-per-click for two-thirds of the sample, while increasing clicks 59% on average across the 30 000+ SEM accounts. While not all accounts were optimized for conversions, those that were saw an average increase of 71% in conversions! PPC advertisers were also better able to pace and spend their budget as intended, and the lifespan of customer accounts was longer compared to those not using machine learning.
Machines are great at moving money around in real-time to get improved results for SEM marketers. Algorithms take over the nitty gritty of bid and budget management. That can mean optimizing bids and budgets across campaigns, across devices, by location, by time of day or week, by placement, and more. For example, the algorithm could decide to move data from Google Ads over to Bing during a time of day where it would make more money than being in Google Ads.
Bid and budget management emerged as the first area controlled by machine learning under the PPC umbrella because it’s so important – it’s the main area that money is involved within campaign management. However, there are other tasks that machine learning can take care of for PPC marketers.
Machine learning algorithms are also great for standardized tasks in SEM such as recommendations for account audits, keyword selection, ad copy, ad testing, audience segmentation, campaign structure, and even identifying missed opportunities.
Machine learning does indeed present an opportunity to PPC marketers to gain an advantage over their competition; however, it’s based entirely on our ability to trust and adopt these technologies.
Barriers to machine learning adoption in marketing
Marketers haven’t adopted machine learning and AI technologies en masse just yet. But at the rate of technological innovation in this area, we know that this revolution is underway now. Google has unleashed machine learning algorithms into the Google Ads marketplace. That has caused mixed reactions among advertisers, since others are applying machine learning to their marketing via third-party tools. But do marketers trust machines to do at least some parts of their job for them? What about in everyday life? Do we truly trust machines to make decisions for us?
Since there is an “invisible” element to machine learning, humans have difficulty conceptualizing and trusting machines. Marketers have to trust “the invisible.” For example, if a machine bids $5 on a campaign, we don’t really know why. We can see it in the system, but the system won’t talk and explain why it made that bidding decision based on 30 million data points. So, marketers are left to just trust that the machine learning system made the best bid for them.
During our webinar, we polled the audience on what they would trust a machine to do for them. They responded with the following:
I would trust a machine to…
29% Serve Me Food
32% Drive Me to Work
2% Create My Website
24% Run My Ad Campaign
12% Do Surgery on My Body
A quarter of the audience was willing to trust a machine to run an ad campaign for them, while almost no one liked the idea of a machine creating a website for them. Interestingly, most people have warmed up to the idea of self-driving cars. However, that is perhaps the most controversial example when it comes to trusting a machine to make decisions on a person’s behalf. As a hypothetical example that we list in The Marketer’s Field Guide to Machine Learning, what decision will a self driving car make if it has to choose between hitting a person on the left or hitting another person on the right? It’s critical that we as a society continue to ask and resolve these types of ethical scenarios. But not all examples of AI have lives depending on them.
In the world of PPC and search engine marketing, machine learning and AI become less of a question of ethics and more one of adoption. Marketers face practical challenges adopting machine learning and AI technology because they’re new. There is no precedent for these types of systems, and no budget has been specified for them. Marketers are unsure what those look like. The truth is that while examples of intelligent technology, like IBM Watson, have been given human names and faces, machine learning simply looks like code. Just like “the cloud” is conceptual, machine learning can be described with a similar dialogue.
Whether marketers are ready to adopt these invisible algorithms, they are here to stay within Google Ads and third-party martech. Additionally, more examples will permeate the landscape on the way. Despite reluctance, marketers who can enhance their PPC strategy by leveraging machine learning technology will get better performance results than those who choose not to.
Ultimately humans and machines work better together
Machine learning algorithms don’t mean much without humans to make, monitor, and interpret them. Humans and machines work together to achieve better performance results. They are mutually inclusive. With this understanding, marketers don’t need to fear the loss of their jobs when adopting machine learning technologies. In fact, they should embrace technological innovation, which can only enhance their performance. It can result in improvements even if performance is exceeding expectations, campaign structure is complex, or if budgets are small. In conclusion, marketers should feel comfortable exploring machine learning, and expert tips and tricks for adoption can help them transition more smoothly.
