Now that we know what ETL is all about, it's time to add another little-known abbreviation to the bunch: ELT. Yes, it almost looks the same, too.
ELT stands for ‘extract, load and transform’, which at its core is a variation of ETL. Instead of transforming the data first and then loading it into a target database or data warehouse, an ELT process first loads the data and then performs any necessary cleaning up processes. While it requires less from the original data source(s) – only the raw data generated by them - it demands more from the target data warehouse as it has to handle the transformation of large amounts of messy, unprepared data.
Two approaches towards the same goal
Essentially, both ETL and ELT solve the same problem: They help professionals get all their relevant data harmonised and ready for analysis as quickly and painlessly as possible - and only then can they start deriving meaningful insights from it. This is a process that should be treated as a high priority.
“Before data can begin to provide marketing intel, it must be collected, combined, updated as new data is generated and analyzed,” writes marketing analytics expert Bill King. “These steps represent 80% of the success or failure in data intelligence projects.” That is 80% that companies need to nail before even getting to read and interpret their data.
The difference between ETL and ELT lies in the how, not so much in the what.
ETL, also known as data integration, is targeted at marketing technologists and IT specialists, who have the knowledge to handle the processes for their business counterparts. They make sure that the piles of raw data are transformed properly and made readily available for decision-makers to use. ELT - or data wrangling, on the other hand, is in the hands of the people who actually need the insights. Those are the marketers, business managers and others that know exactly what data is necessary to make specific decisions. They can access the data from the data warehouse (or data lake) where it is all stored, and start the transformation process from there.
While the output of ETL is clean and ready-to-use data - which, let’s face it, sounds amazing, ELT solutions tend to give decision-makers the freedom to explore their data before transforming it. Both processes have the same end goal in mind, and they both have their strengths. So which one is it going to be?
It’s not ETL vs ELT. It’s the two of them together.
The fact of the matter is that ETL and ELT do not compete with each other, but actually complement each other. So, the right solution for your company is often not one or the other, it’s both.
An integrated ETL solution would be key if you want to run a truly data-driven business. After all, to make sure that you are deriving actionable insights at just the right times, you need to have all your data cleaned and structured in one designed space, much like a data warehouse. On the other hand, however, certain teams in your company (think marketing!) also need the freedom to experiment and be flexible with the data they are using. In that case, it is better if they have access to the pool of raw, unfiltered data, so they can decide how best to use it. This is when ELT comes into play.
While ETL spares you the effort in transforming data from different datasets and provides you with an clean overview from the get-go, it also deprives you of the chance to see all the data and consciously decide what is useful and what isn’t. ELT gives you that opportunity. That being said, it is important to assess exactly what your company needs before choosing the tools you are going to use.
If you want to strike the right balance, however, do consider integrating both ETL and ELT processes into your business strategy. They both have their strengths and together they will take less time and effort to bring you to your end goal.BACK TO POSTS