PPC – PPC forecasting with Google Sheets and Vertex AI

PPC - PPC forecasting with Google Sheets and Vertex AI

As the PPC landscape continues to evolve, having the ability to predict future campaign performance is invaluable. 

This article will cover some of my favorite PPC forecasting techniques using Google Sheets and Vertex AI that paint a clearer picture of the future and empower clients with actionable insights. 

While no tool or technique can provide a 100% accurate picture of the future, the methods outlined here can show us a glimpse into the potential trajectories of PPC campaigns.

Google Sheets FORECAST function: The basics

Google Sheets offers an easy-to-use and reliable forecasting function using the formula:

=FORECAST(z, known_y values, known_x values)


  • z is the data point for which you want to predict a corresponding y-value.
  • known_y's is the range of dependent data points (usually your past results or outcomes).
  • known_x's is the range of independent data points (usually the variable you think might influence your outcomes).

This function is a great tool if you only have two dimensions.

However, it uses linear regression, which is fine for a quick forecasting sneak peek but nothing too advanced to account for external circumstances or other data sources.

Let’s say you have historical data from the last year and want to forecast future budget predictions to have some numbers to plan with. 

In this example, we have the current year’s sales data up to August and want to forecast future sales from September to December.

If we visualize those forecasts, you’ll quickly see the drawbacks of using this method.

Visualized Google Sheets PPC forecast

The blue line represents the known sales data up to August, and the red line represents the predicted sales data. 

The prediction is not more than a trendline, which might help to get a high-level look at something but is nothing compared to the blue line, which is basically how real business data will look like.

Supercharging the Google Sheets FORECAST function

To fix the issue of linear regression, there are multiple ways to approach the forecast formula with advanced methods. 

Instead of just using the linear =FORECAST() function, you can add a little twist by adding trend data or other market predictions into the forecast formula, as in:


You can grab trend data from public sources like Google Trends, Google Keyword Planner, Dataset Search by Google or industry reports (from PwC, EY, McKinsey, etc.) and export it into a CSV or any other format you’re used to working with. 

Clean up those datasets to match your original sheet’s structure, like data on a day-to-day, week-to-week or month-to-month basis,

Next, supplement the FORECAST function to get a more realistic prediction rather than just a straight line going up or downwards.

Google Sheets - FORECAST function for PPC with trend data

In this example, we have used additional trend data, which shows an increased trend toward Q4 of the year. The numbers are, therefore, different from the predicted sales without trend data.

If we visualize those new data, we can see that the trend data gives us better insights and more detail compared to a flat trend line.

Visualized Google Sheets PPC forecast with trend data

As a rule of thumb, it is almost always a good idea to support those forecasts with as much data as possible and supply data on more detailed timeframes like day-to-day or week-to-week.

Get the daily newsletter search marketers rely on.

Advanced forecasting with Vertex AI

If your PPC campaigns involve large datasets with multiple variables, turning to Google’s Vertex AI could be a game-changer for your forecasting needs. 

Unlike simpler tools, Vertex AI allows for more complex models that can take numerous factors into account, such as seasonality, different ad platforms, or even global market trends. 

To get started, you’ll first need to upload your historical PPC data to Google Cloud Storage.

From there, you can access Vertex AI’s AutoML Tables to build a machine learning model tailored to your dataset automatically. 

After training the model, you can evaluate its performance using built-in metrics to ensure it meets your forecasting requirements. Once you’re satisfied, deploying the model is easy. 

Now, you can use this model to predict future outcomes, such as clicks, impressions, or conversions, based on varying levels of ad spend, ad placements, or any other variable you deem important. 

And the best part? You don’t need to be a machine learning expert to do this. With a bit of setup and some fine-tuning, you’ll be on your way to more accurate and insightful PPC forecasts.

The capabilities of Vertex AI are sheer endless, but let’s look at a simple framework to start with.

Once your Google Cloud account is set up and you’ve created a project within Vertex AI, you should start by creating a data set.

Google Cloud - Vertex AI training data creation

A data set is basically the collection of data points you want to use for your forecasting. 

The data set contains a time dimension and some budget and revenue dimensions. Depending on your goal, the data sets might contain different data points.

Google Cloud - Vertex AI data set options

Name your data set, select Tabular as your data type and Regression or Forecasting as your objective.

While regression is often used to understand relationships and can be applied to various data types, forecasting focuses more on predicting future points in a time series. 

Both are essential tools in data science and are used for different types of decision-making and analysis. You will be fine with forecasting in most of your cases.

Google Cloud - Vertex AI training method

Now it’s time to train a new model. For beginners, the AutoML training method is always a good choice. Next, you must set some settings on the forecast period, the target, and data granularity. 

Once done, set the training duration and budget, and you’re ready. The model will now learn, and you’ll receive a notification once it’s finished. 

The last step is to get predictions from the ML model. This option is only available once the training is done. 

To create a forecast, you need to submit data that the forecast will be based on. It’s ideal to use more recent data. 

The model will predict future target values learned on the training data set and based on your forecast data set.

Depending on the data volume, the job will take a while. But you shouldn’t wait more than 5-10 minutes for PPC tasks. 

Once finished, Vertex AI will supply an output file that contains new columns with predicted values that you can use for further decision-making.

Vertex AI might seem a little much for some forecasting tasks, but remember that you can submit years of historical data, inventory insights and more to train the model. 

With Vertex AI, you can build a machine learning forecasting model tailored to your business, which is much stronger than any static forecasting formula.

PPC forecasting for better campaign performance

Ultimately, the choice between these tools depends on your goals and the complexity of your PPC campaigns. 

Google Sheets offers a straightforward and accessible way to dip your toes into PPC forecasting. While it may have limitations, it’s a valuable starting point for many advertisers.

On the other hand, Vertex AI takes your PPC forecasting capabilities to the next level with its ability to handle large datasets and complex models. You can now factor in seasonality, global trends, and various variables to make accurate predictions tailored to your business.

Whether you opt for the simplicity of Google Sheets or the sophistication of Vertex AI, forecasting in PPC advertising is no longer a guessing game.

You can now equip yourself with actionable insights and make data-backed decisions on your PPC campaigns.

Dig deeper: A guide to effective PPC projections

Opinions expressed in this article are those of the guest author and not necessarily Search Engine Land. Staff authors are listed here.

Leave a Reply

Your email address will not be published. Required fields are marked *