Artificial intelligence, or AI, has been a top buzzword over the last year.
In fact, AI was even named the “Marketing Word of the Year” in 2017 by the Association of National Advertisers (ANA).
There are so many opinions and prognostications about the future of AI – with industry leaders and others weighing in – that it can be difficult to understand.
Facebook’s Mark Zuckerberg recently shared his optimism over the rise of AI technologies like deep learning and how they could lead to breakthroughs in areas like healthcare and self-driving cars.
Space X founder Elon Musk contends that “AI is more dangerous than nuclear weapons.”
On the other hand, Bill Gates asserts that “Artificial intelligence is good for society.”
While the late Stephen Hawking declared that “The development of full artificial intelligence could spell the end of the human race.”
What do we make of these statements?
Like many hot topics, AI is often described in hyperbolic extremes that can sound apocalyptic or utopian. But many applications of AI – and machine learning, which enables computers to learn from vast datasets – are practical.
This is particularly true in the field of advertising, where AI has been mostly uncontroversial in its applications. Search advertising is a great example.
AI in Search Advertising
Search advertising involves billions of data points and decisions, from keywords to inventory sources – and additional dimensions such as audiences, device, geo and time, with multiple campaign settings.
In all, it is too much data and too complex for humans to make the optimal decisions manually. AI is uniquely suited to help humans make trade-offs to drive the best possible performance.
Leveraged properly, AI can help advertisers drive better performance and ROI. It can take complex data sets and build predictive models to make decisions on bids and budget allocation to best meet advertisers’ objectives.
But for many marketers, AI is not something they want to simply check a box to activate – or “set and forget.”
To increase the impact of AI and machine learning in search advertising, a number of best practices have emerged that blend the best of human and machine.
1. Set Clear Goals
AI is not taking over your job. It still needs humans to make strategic decisions and set business goals. It is important to take the time to set the right goals, which are unique to any business.
For some, optimizing to a cost per action (cost per sale, lead, registration) is best while others might want return on ad spend.
AI alone will not determine strategy or goals, but it will enable advertisers to efficiently achieve them.
2. The More Data, the Better
The more data that AI has to learn from on a searcher’s intent, the better it can optimize campaigns.
Even if searchers for some keywords don’t convert, AI can leverage other data like actions on a website or audience data to optimize for performance.
Leveraging a marketer’s full data footprint – including site engagement and conversion metrics (time on site, number of pages viewed, registration, shopping cart status and product page visits, etc.), customer CRM data or data from a data management platform (DMP) – is key to maximizing performance.
3. Apply Expertise
Setting guardrails, like campaign settings and budgets based on an advertiser’s expertise and historical successes, helps AI stay in its lane. But advertisers should be careful to not get carried away.
Since AI works the best with few constraints, going overboard can be detrimental since the AI system will look for the optimal solution among a smaller set of possible values.
By focusing on setting the right goals as guardrails, marketers can then let the algorithm decide how to best achieve them.
4. Run Simulations
Simulations allow a marketer to do a low-effort test to see estimated impact without investing real dollars.
Before making big changes to campaign settings or budgets, simulations can help a marketer understand the impact on business goals – then, when ready, AI can take over on performance optimization.
5. Trust, but Verify
Regularly reviewing model accuracy reports offers much-needed peace of mind that AI forecasts and optimizations are working as expected.
Reviewing forecasts for impressions, clicks, and revenue versus actual performance verifies that AI is making the right decisions to most efficiently meet performance goals.
With these best practices in mind, AI can do what it does best – and so can you.
More AI & Search Marketing Resources:
- The Role of AI and Automation in Paid Search [PODCAST]
- Why Machine Learning Is Key to the Search Marketing of Tomorrow
- How to Integrate AI Into Your Digital Marketing Strategy