Data wrangle your way to more accurate forecasts

Tyler Robinson - Consultant

It goes without saying that demand forecasting is a mission-critical operation of your business. Staffing, inventory management, production decisions, and financial planning are just a few of the areas highly dependent on accurate forecasting figures. Without a reliable finger on the pulse of how your business is going to perform in the coming days, weeks, months, and years, you are blindly making crucial decisions that can impact your bottom line and stunt company growth. Predicting demand is not an easy task itself, so it’s vital that the data you are using to make these predictions is clean and intelligently formatted.

What is the Problem?

There are many tools on the market today that can help you create machine learning models that predict demand with a high degree of accuracy. Depending on the technical skillset of your data science or business analyst team, and the quantity of data you are working with, it can be difficult to select the right one. While it is a cliché, the idea of “trash in, trash out” could not be more true when it comes to any of these programs. You are going to get out exactly what you put into them. And while you might discover a very high R-squared value or see a beautifully clustered scatter plot, deriving financial gains from these outputs requires data of the utmost quality. Otherwise, the output numbers are simply just numbers. The key is to find a signal in the noise, and with too much noise resulting from poor data, you will not be able to identify the signal, or you will misidentify it completely. As a result, data integrity and cleanliness should be your primary concerns before inputting your data into any kind of business intelligence or statistical tool. These programs aren’t going to stop you from inputting flawed data, so it is your job to provide them with the best data possible, because with each extra percentage point of model accuracy, better business decisions can be made.

What is Trifacta?

This is where Trifacta enters the equation. Trifacta is a premier data preparation tool designed to automate and accelerate your data cleansing processes, as well as provide valuable insights into the overall structure and quality of your data. Trifacta visualizes your data, as well as its associated meta-data, and allows you to manipulate it in a more user-friendly way than can be done with Excel, SQL, or other data preparation tools. You no longer need to write complex macros or SQL queries to conduct your data preparation and combination efforts. Trifacta leverages an intelligent system that provides recommendations to you based on common themes it notices in your dataset. For example, if some date records in your data are stored as MM-DD-YYYY, and other records as MM-DD-YY, it will suggest changing all the records with one format to the other. You can then choose to accept this modification and add it to your “recipe”, thus standardizing the field and automating this step, so when new data gets loaded into the workflow, this recipe step always is conducted. Trifacta’s intelligent system can help you solve data quality issues that you may not even know exist. For instance, another common thorn in data preparation is how your end tool reads and distinguishes null values, 0s, and empty text strings. Trifacta will quickly discern these differences and allow you to standardize these values into the output that is best for you and your systems.

Another way that Trifacta cements itself as uniquely powerful data preparation tool is its seamless ability to connect to your dynamic data sources such as Microsoft SQL server, IBM DB2, Oracle Database, PostreSQL, and many more. Therefore, your data can be pulled directly from the source and your transformations can be exported as any number of formats such as a CSV, XLSX, JSON, whichever format is best for your data’s final purpose. You can even export it as a QVD so that it can be optimally read by Qlik Sense, expediting your data load.

How Trifacta Can Help

With Trifacta, your dataset modifications are demonstrated to you in a before-and-after presentation as you add steps to your recipe, so you can see the effects of your changes in real-time. This instant feedback is a great improvement upon using Excel and SQL for data preparation, because you no longer need to make sweeping changes without knowing the exact outcome of your decisions. Furthermore, because data preparation in Trifacta uses a recipe format, it is easy to jump back to older steps to modify or delete them in order to sculpt your data to your exact export specifications.

Combining multiple data sources is a notoriously difficult task due to the variability of naming conventions, field data types, formatting, etc. Currently, you may be constantly cycling and manually combining various Excel files to gain an understanding of what lies ahead for your business. Joining datasets in Excel is not an optimal solution and can only be done by copy and pasting (not automated) or writing a macro (not easy). With Trifacta, it can be done with just a few simple commands. Trifacta uses its built-in intelligence to match field names, even if they are not perfectly aligned, and provides you the ability to manually map them. Even better, once the recipe is in place and Trifacta is connected to your dynamic data sources, it can automatically pull from the source, combine the files, and clean the data. A good use case for this is you can union your daily reports on a scheduled basis, so when you want to export the data, you will be able to trend daily data over longer periods of time, without the necessity to manually combine files via copy and paste or other labor-intensive methods. Now, your data is in one place and is ready to be utilized in your preference of data science tool or business intelligence software.

If your organization is stuck in the rut of reporting from a seemingly countless number of Excel spreadsheets, or struggles with the accuracy of business intelligence and data science tools, Trifacta could be the program you need to take your data preparation to the next level. With clean data, more accurate insights and predictions can be made from your choice of business intelligence and statistical prediction tool. According to a recent survey, 80% of a data scientist’s time is devoted to data preparation, yet 76% of data scientists believe that data preparation is the least enjoyable part of their work. This makes a lot of sense, as cleaning data in Python, SQL, or Excel is incredibly tedious, but it also creates an enormous opportunity for improvement that Trifacta capitalizes on. Trifacta simplifies and expedites the entire process, meaning your data scientists and business analysts will have more time to spend deriving valuable insights, rather than cleaning and preparing data for analysis, making them happier and more productive. If you are interested in learning more, please reach out to us and we would be happy to discuss further.