Artificial Intelligence is a hot topic in PPC but until the machines fully take over day-to-day account management, there are a few key areas where human PPC pros can still add a lot of value.
Use Business Data for Bid Management
Bid management can be one of the most repetitive and boring tasks of managing PPC because after a model has been built, you are left with an ongoing task executing on the plan and this may involve downloading the data, putting it into the correct format, and then running it through your formulas to determine the new bid. For machines, this might sound like the perfect dinner on a caribbean-beach sunset, but for us humans? Not so much. Repetition is dull, and as a dull task, we tend to become a bit less thorough with our analysis as time goes on.
This is why both Google and Bing offer automated bid management solutions. There are also many third party bid management solutions which aim to improve on shortcomings of the bid management solutions from an engine. Though it is a well-known fact that the engines can do amazing bid management work, their solutions are generic and can ignore aspects that the business owner knows will impact their online conversions.
There are four clear advantages to using the engine’s bid management systems:
They are free to use.
They are based on best-in-breed algorithms.
They have access to a lot of auction-time signals that advertisers don’t get (e.g. who is the user, what did they search before).
They can set bids in real-time based on auction time signals.
But there are several things these automated bid systems cannot do:
Know the context of the performance that is measured through conversion tracking (e.g. conversions were slow yesterday because there was an issue with servers in one of the data centers).
Understand the factors that impact the industry (e.g. a plumber with 15 vans will be better able to service a distributed customer base than one with just 3 vans).
The ideal bid management system combines the algorithms from the engines with data from your business. To this end, advertisers should calculate their own CPCs based on in-house data and then submit these bids to the engine as an Enhanced CPC, so that Google or Bing can adjust the bid up or down based on what they know about each auction.
This is a reason why tools like Optmyzr are popular for managing bids. They can help automate bid strategies that use a combination of data from the search engine (like historical conversion rates for individual keywords) and business data (like phone sales data, ecommerce returns data, or even how the weather impacts sales).
Optmyzr’s Rules Based Optimizations for bids are also ideal for agencies who want to add value beyond what the engine’s own bidding system can do, but who don’t want to build complex technology in-house that they need to maintain as Google and Bing go through their frequent updates to the API. Prebuilt recipes can be installed in seconds to help advertisers reach goals like target CPA, target ROAS, or target position. These recipes can be enhanced over time as more is learned about factors that impact performance, whether they be based on Google’s data or internal business data.
Use Keywords to Target Shopping Ads
A second area where PPC pros should take back some control from the machines is with managing keywords for Shopping Ads. While shopping ads are automatically targeted to relevant queries that match the product in an advertiser’s feed, there is always the option of adding negative keywords.
In a rather extreme, yet interestingly practical way, you could actually target a specific keyword not by the inclusion of that term, but rather by exclusion of all other terms.
This is the foundation of “Query Sculpting”, a PPC technique that deploys negative keywords to drive traffic to the desired target. And because negative keywords are much more explicit than positive keywords, they are the main tool.
Even in search campaigns, query sculpting is done with the addition of negative keywords. And while this makes a strange sort of sense, our logical side is still asking “why can’t it be done by simply adding exact match positive keywords?” Because ever since Google’s latest change to the algorithm, exact match no longer truly means ‘exact’.
Query sculpting for shopping campaigns was invented by Martin Roettgerding and later refined by various entities including SmarterCommerce. Martin’s technique requires maintaining 3 parallel shopping campaigns and proactively adding certain types of negative keywords.
But proactively adding extra campaigns and unnecessary negative keywords can really eat into an account’s allowance for number of keywords under management. Optmyzr, taking into account the pros and cons of both sides, has a solution that uses recent performance data to sculpt queries when it is clear they could perform better elsewhere in the account: The Shopping Negatives Tool.
The Shopping Negatives Tool analyzes the performance of the same search queries across different ad groups in a shopping campaign, and finds the ad group in which the query is not performing well to recommend adding it as an exact match negative.
Using this technique, advertisers can run as many shopping campaigns in parallel as they want or keep everything in one campaign, and Optmyzr’s analysis will make suggestions for how to sculpt the traffic to drive more sales at a better ROAS.
Create Better Ad Tests
Googler Matt Lawson has recently covered the new ways to think about A/B ad testing. Thanks to Google’s improvements in Machine Learning, there is less need to manually cull underperforming ads from an account. The premise is that the worst ad in an ad group could actually perform quite well with a subset of users hitting that ad group, which means that removing a slightly losing ad could actually be counterproductive.
But he also says “Delete stuff whenever an ad stops seeing a large fraction of the impressions and therefore generates minimal to no clicks. Then add a new ad to the mix. It’s better to have options.”
To help with cleaning up ads who are seeing a minimal share of impressions in an ad group, you could use AdWords Scripts, like some of the ones that are part of Optmyzr’s suite of tools.
While Google is removing the need for manual testing of ads, and though they’re even doing some automatic generation of new ad text challengers, this remains an area where the human expert – someone who is close to the business being advertised – will have a leg up on automations.
You’ve heard the story that if you gave 1,000 monkeys typewriters and an infinite amount of time, they’d eventually write a Shakespeare novel. But monkeys eat lots of bananas and tend to prioritize climbing trees before writing those award-winning soliloquies, so they’d most likely take forever. And though the concept of novel-writing monkeys does seem very attractive, advertisers aren’t willing to wait for an infinite amount of time. That’s why we still need tools that help us write great ads in the least possible time.
Tools like Optmyzr can help with the ideation for new ads by highlighting ad text elements that have performed well historically.
Frederick Vallaeys made the point that the PPC agencies of the future will be the ones with the best process for testing. Machine Learning means computers can figure out the winners and losers, but conclusive test results will happen more quickly when using human insight to prioritize the most valid hypotheses for testing.
Exciting, and perhaps scary times are ahead for all sorts of professions where AI will take over a plethora of tasks that used to require human intelligence. There’s a slight fog surrounding the future of human intelligence in the workplace, and though it isn’t thick enough to cover us just quite yet, it creates a bit of unease in many circles. What will happen when machines take over?
It’s an inevitable passage, but the more human input we give these machines all throughout this transition period, the more effective they will be at helping achieve the shared goal of improving PPC performance. And in the meantime, human PPC pros have many opportunities to transform their current day-to-day, into something that will endure over time, and set a solid foundation for working in an AI-first world.