What is Predictive Performance Marketing?
It’s always good when companies base their decisions on solid data. However, it is better not only to rely on historical inventory data, but also to consider realistic future scenarios. In short: it is good to make decisions not only according to the data of the past, but also before the data of the future. A simple extrapolation is not enough, because such an extrapolation takes neither the development pattern nor the seasonality of a dataset into account. For realistic forecasts, regression analyzes with seasonality values should be used to calculate exact values for individual weeks or months.
Forecasting with IBM Cognos Analytics and Microsoft Power BI for Desktop
Different metrics in online and performance marketing can be predicted for such forecasts: clicks from organic Google searches on websites, executed goals of online users from paid advertising, the statistical duration from the first contact to a purchase from a customer or the development of UX -Values, such as the bounce rate or the conversion rate (percentage of users who fulfilled a goal on a website). In other words: With forecasting dashboards, a company can monitor all channels in online and performance marketing, which acts like a radar six to twelve months in advance. This also creates a new way of working in online and performance marketing – namely predictive performance marketing.
Detect downtrend from a specific metric with forecast from Microsoft Power BI with regression analysis and factored in seasonality value.
This is what an uptrend based on seasonality might look like in a Microsoft Power BI for Desktop forecast.
Which data is suitable for forecasts in online marketing?
All conceivable data from online and performance marketing is suitable for forecasting, and with the right technologies it can be realistically forecast for around six to twelve months. This is possible, for example, with environments such as Microsoft Power BI for Desktop or IBM Cognos Analytics.
This in turn makes it possible to recognize and prevent downward trends and to take advantage of opportunities in foreseeable upward trends. You can also use Forecasts to determine the approximate date of a calculated KPI target achievement for projects, which makes resource planning decision-making much more effective. This way you know what a certain cost point per deal in an online advertising campaign might reach or when your website exceeds a threshold of organic Google clicks per month.
Different metrics in online and performance marketing can be predicted for such forecasts: clicks from organic Google searches on websites, executed goals of online users from paid advertising, the statistical duration from the first contact to the purchase of a customer on the website or the development of UX values, such as the bounce rate or the conversion rate (percentage of users who fulfilled a goal on a website). In other words: With forecasting dashboards, a company can monitor all channels in online and performance marketing, which acts like a radar six to twelve months in advance. This creates a new type of performance marketing: predictive performance marketing.
Information and action advantage through realistic predictions
Forecasting dashboards create an information advantage on the one hand and an action advantage on the other: You don’t have to wait until a downward trend has occurred, you can avert it before it occurs. So if a website is likely to fall below a certain threshold in terms of its performance in organic Google clicks in six months, a marketing team in the company can preemptively mobilize resources to place more content for organic reach. Many problems that would otherwise have arisen do not develop in the first place.
This is what a forecast downtrend would look like in a report using IBM Cognos Analytics.
How forecasts work: regression analysis and seasonality values
But which technologies should be used for forecasts? First, it should be able to do at least two things: It should perform a regression analysis from the dataset’s values, taking seasonality values into account. The background is that a linear regression in itself only gives the statistical direction of development, but not a development pattern. Only a seasonality value ensures that the regression analysis is based on a causal development pattern, which means that fluctuations and trends between weekdays, calendar weeks or months can be calculated from the inventory data.
The data sets for forecasts in the common business intelligence applications must be in tabular data, such as a Google Sheets spreadsheet, an Excel or CSV file, or a BigQuery database. The table headers provide the labels for the dimensions and metrics, while the row entries contain the actual metrics. Such a tabular data record must always contain a time axis, i.e. a column that contains time information, such as days by date, calendar weeks, months, years, hours, minutes or seconds. That would represent the dimension in the forecast. Another column should contain the metrics of the forecast, e.g. website clicks from organic or paid traffic sources, conversion achieved (goal on the website), UX values,
The seasonality value mentioned at the beginning reflects the number of the most recent data points in your data set along the time axis. So if you want to forecast with a tabular dataset containing 700 days of clicks and you choose a seasonality of 200, the pattern of the last 200 days would be calculated for your forecast.
Environments for forecasts
Two particularly popular forecasting technologies are Microsoft Power BI for Desktop and IBM Cognos Analytics. These carry out regression analyzes from tabular data sets, taking into account seasonality values. According to the same scheme, you can also set up a specially trained machine learning model with the Vertex AI on the Google Cloud . The following is about the forecasting functions of Microsoft Power BI and IBM Cognos Analytics. Both applications excel in allowing high-performance data joins, connecting and modeling data from disparate sources, and drag-and-drop chart, dimension, and metric creation in an easy-to-use report editor.
Forecasting with Microsoft Power BI for Desktop
Microsoft Power BI for Desktop is a decoupled program from Microsoft Power BI, which can be used offline, but which can also import online data from various sources – such as from Microsoft Power BI itself. Microsoft Power BI for Desktop is a free download from Microsoft. The application offers the possibility to connect different data sources and, if necessary, to transform the data if there are compatibility problems, such as an incorrect date format, wrong decimal places or unrecognized headers for the labels of dimensions and metrics. Here you can upload a tabular data set as an Excel or CSV file or link Google Sheets tables and BigQuery databases online. The dimensions (time axis) and metrics (e.g. clicks) of the data set are then displayed on the far right in the menu bar. From the toolbar, you can drag and drop chart types into the report editor, and then import your dataset’s dimensions and metrics.
In Microsoft Power BI for Desktop, you can use the two right-hand menu bars to drag and drop visualizations, dimensions, and metrics into the report editor.
The forecast function is located in the “Analysis” menu tab with the magnifying glass icon in the right-hand toolbar. Here you can enter your forecast period in data points. A data point corresponds to a table row in your tabular data set. So if you have 700 days as rows in a row in your table, a forecast period of 200 would mean exactly 200 days in the forecast. According to the same scheme, this applies to hours, minutes, seconds, calendar weeks, months or years. You can also enter your seasonality values here. You will also see an adjustable confidence interval for possible deflections.
In Microsoft Power BI for Desktop, you can use the two right-hand menu bars to drag and drop visualizations, dimensions, and metrics into the report editor.
Forecasting with IBM Cognos Analytics
In the area of forecasts, IBM Cognos Analytics works according to a system similar to Microsoft Power BI for Desktop and calculates regression analyzes taking seasonality values into account. Again, there is a report editor that works with drag-and-drop. IBM Cognos Analytics has some exciting chart visualizations that are not always available elsewhere, such as the forecast for decision making as a structure chart. The application is also capable of connecting a massive variety of datasets online to explore and visualize connections and relationships between multiple datasets. However, IBM Cognos Analytics is not completely free: In addition to a free 30-day trial version, you can purchase the program for as little as $10 a month.
Also in IBM Cognos Analytics, it is possible to enter forecast periods and seasonality values using data points that correspond to measures of the time period dimension of the tabular data set. The forecast function is located in the report editor at the top right in the “Analysis” menu item.
Forecasting dashboards as a radar
When making decisions, companies should not only judge according to past data, but should consult a realistic forecast before data developments occur. In this way, downward trends can be prevented, upward trends expanded and the achievement of KPI targets terminated. Basically, this applies to different data sets in performance marketing: for the performance values of individual advertising campaigns with Google Ads or Meta, but also for the organic ranges from search engine traffic, either entire websites or individual subpages. When a UX team makes continuous improvements to the website’s usability, UX metrics such as bounce rates, conversion rates, session lengths, and scroll depths can also be measured. Online shops and B2B companies can use their analytics systems or CRM systems to forecast the average duration of reaching customer lifecycle stages in order to see whether customer retention strategies or newsletters actually get users to buy faster. Ergo: There are countless use cases for which it is worth using forecasting dashboards to find answers to events before they happen.
Marketing & Lead Gen
SEA, paid media, conversion rate optimization, market and trend analysis in the Berlin-Brandenburg area.
Telephone: +49 176 588 744 04
Email: info@slavawagner.de