Richard Couture and Jason MacDonald

Interview: Managing PPC Seasonality with Machine Learning

With the summer months upon us, advertisers of sandal companies, sunscreen and garden accessories are all increasing their keyword bids – it’s their season! Marketers of all kinds are affected by seasonality – not just paid search marketers. But what does that mean?

Seasonality could refer to; the climate changing throughout the year, different holiday times of year or even when a business’ customers buy. We wanted to clear up any confusion about seasonality and paid media and give you the best advice to manage those peaks and valleys. So we turned to not one, but two of our smartest employees to get their take.

Richard Couture, our CTO and Co-Founder, as well as Jason MacDonald, our Head of Optimization and Data Science, are featured in the interview below focusing on seasonality.

Richard Couture and Jason MacDonald

What is the concept of seasonality in paid advertising (Holidays? Times of year? The times your customers buy? Market fluctuations? All of the above?)?

RC: All of the above and in the context of our clients (mostly SMBs), we’re talking about the cycle their business is going through. Exterminators (pest control), landscapers, pool maintenance services, and snow removal are obvious cases. Auto repair, contractors, taxis, and moving companies are more subtle to understand and detect.

The concept of seasonality in paid advertising is the ability to understand the characteristics of time (time of the year, time of the day) on the ad campaigns performance/results. With proper predictable knowledge (pattern detection), we can adjust the campaigns and budget to maximize their return.

JM: Seasonality is the fluctuation of search volume and the price to advertise in a specific vertical or market and location at a given time. Depending on the specific business, seasonality can present itself in many ways, including one large spending period around a major holiday or a change in strategy through the year.

When it comes to seasonality, it is important to understand that there are a lot of factors related to the business that you have to understand to properly plan for the variability. The products or services offered by the business and their location are going to contribute to the seasonality. Understanding them will be important to be able to identify and react.

For example, a plumber may seem like a very straight forward business with rather uniform demand throughout the year. In actuality, they can have distinct seasons where HVAC installations and repairs will be a big part of their business in the warmer months and home renovations will be more needed in the fall depending on the climate of the area. You need to plan your campaigns to target the right service in the right season.

How are paid media accounts typically affected by seasonality and how do campaign managers navigate through these ups and downs?

RC: In the obvious cases, we can expect peaks and valleys. During their high seasons, demand is strong and we usually don’t have to be too aggressive to get clicks (we pay less). It is the complete opposite situation in low seasons, where clicks are so rare that in order to get some, we usually have to increase the bids.

It is extremely difficult for humans to detect when to adjust the bids to leverage these market changes due to seasonality. Seasonality is not only associated with market conditions that recur time over time, seasonality may also slide from one period of time to another. Pool season for Canadians may vary based on the warm or cold weather in April and May, same situation for air conditioning vendors. Detecting seasonality patterns and adjusting campaigns accordingly is an active thing that campaign managers need to consider.

JM: Seasonality is a campaign manager’s biggest challenge to overcome and adapt to. We are all familiar with the bigger events tied to holidays like Christmas, where we can clearly identify dramatic increase in volume leading up to the holidays. These are easy to see coming and to plan proactively for.

In many cases the seasonality is very subtle and extremely hard to detect. You have to be looking for the right signs to detect and react quickly. The typical signs of seasonality are a drop in clicks or calls that leads to challenges spending the budget. This can be due to either entering a low season where search volumes have reduced or entering a high season where the competition can be much fiercer.

How does machine learning improve a campaign manager’s ability to manage and prepare for seasonal fluctuations in campaigns?

RC: Machines are able to easily correlate time series with experienced data such as impressions, clicks, cost, conversions, and conversion value. ML can identify patterns in time series that recurs over time in order to suggest adjustments that campaign managers can apply. It can also compare its prediction with real life data and learn/adjust smartly from actions that drove improved results from the ones with lesser or negative impact.

JM: Machine learning solutions like Bid and Budget Management can help with many aspects of seasonality. It is actively reviewing your campaign’s performance throughout the day and when it detects a change in performance it can adjust your bidding automatically in small increments to stay on target.

Primarily it can make sure that you are bidding the right value to stay competitive in the auction, maximizing your clicks or conversions within the budget allotted.

What kind of results can marketers expect from automating bid and budget using machine learning in the context of seasonality?

RC: Better meaningful results (which is usually more clicks and/or more conversions). Because ML adjusts the budget and bids based on the seasonality, the budget varies across the year making sure we pay the appropriate price for each click on high and low seasons. Campaign managers can also expect to free up a lot of their time to do other tasks like writing great creatives.

JM: This can vary depending on the transition. If you are transitioning into a high season you will see stellar results. With the added search volume you will see an increased volume of clicks at a better price point than the recent history.

If the season is leading to a drop in your traffic the results will not look as good. With the decrease in interest clicks will be impacted and you will see less of them. With the decrease in clicks the account will be challenged to spend the budget and the automated bidding will increase the bids to ensure the account is able to spend as much as possible, resulting in higher CPCs.

If campaign managers are seeing this scenario, they need to increase visibility and find a large audience. There are a number of ways they can achieve this depending on the scenario. For example: refine their campaigns to other service offerings, adjust budget distributions or widen the reach by extending geo-targeting or simply lower the budget to adjust for the decreasing demand.

What advice can you give to campaign managers who are trying to manage seasonal campaigns manually?

RC: Be patient and like your job because you will work long hours. Seasonality applies at various levels. It can be at the account level (pool maintenance is a good example). It can also be at the keyword level where some keywords may be impacted more by seasonality than others. Predicting changes and applying the right adjustment can easily become cumbersome when handling many entities. If those entities are many client accounts, the amount of work required for campaigns managers become gargantuesque. ML algos are made to solve this kind of problem and learn by crunching tons of data, no matter what day and time it is. This allows the system to identify patterns and apply the appropriate changes, while continuing to learn.

JM: The best strategy is to be aware of it and react to it. Like those using machine learning technology, campaign managers who are managing campaigns manually should focus on increasing the visibility in a low season. Additionally, they will need to adjust their bids and budgets to the appropriate value.

The higher season is as equally important to react to, but we will often overlook it because the performance looks good and the client is happy. Reviewing the bids and ensuring they are set at the right value will go a long way to driving even more clicks.

Just for fun, what is your favourite season? Spring, Summer, Autumn or Winter?

RC: Winter, essentially because it’s hockey and ski time but also because it’s good to sit by the fire with a gin (from Quebec) tonic in hand!

JM: Without a doubt, Spring. It’s such a positive time of year with the transition from cold winters to the warmth and sun of summer.
 

If there is a question you’ve had about seasonality and paid media that did not get answered in the interview above, feel free to ask in the comments below and we’ll make sure either Richard or Jason get to it post haste!
 

Image Credits

Feature Image: Taken by Cassy Trussell at Acquisio

Chandal Nolasco Da Silva

Chandal Nolasco Da Silva

With nearly a decade of digital marketing experience, Chandal has created content strategies for both the biggest and sometimes the most unexpected markets, while developing strategic relationships with editors and publishers. Chandal contributes to some of the highest authority industry publications, has been featured in industry events and is thrilled to be Acquisio’s Content Director.

The First Machine Learning Marketing Platform
Built to Scale Search for Local Resellers & Agencies

Automate, optimize and track more campaigns, more profitably.