Supporting a World-class Documentation Technique with Atlan
The Energetic Metadata Pioneers collection options Atlan clients who’ve accomplished an intensive analysis of the Energetic Metadata Administration market. Paying ahead what you’ve discovered to the following information chief is the true spirit of the Atlan group! In order that they’re right here to share their hard-earned perspective on an evolving market, what makes up their fashionable information stack, modern use instances for metadata, and extra.
On this installment of the collection, we meet Tina Wang, Analytics Engineering Supervisor at Tala, a digital monetary providers platform with eight million clients, named to Forbes’ FinTech 50 checklist for eight consecutive years. She shares their two-year journey with Atlan, and the way their sturdy tradition of documentation helps their migration to a brand new, state-of-the-art information platform.
This interview has been edited for brevity and readability.
Might you inform us a bit about your self, your background, and what drew you to Knowledge & Analytics?
From the start, I’ve been very excited about enterprise, economics, and information, and that’s why I selected to double main in Economics and Statistics at UCLA. I’ve been within the information area ever since. My skilled background has been in start-ups, and in previous expertise, I’ve all the time been the primary individual on the info workforce, which incorporates establishing all of the infrastructure, constructing experiences, discovering insights, and many communication with individuals. At Tala, I get to work with a workforce to design and construct new information infrastructure. I discover that work tremendous attention-grabbing and funky, and that’s why I’ve stayed on this discipline.
Would you thoughts describing Tala, and the way your information workforce helps the group?
Tala is a FinTech firm. At Tala, we all know as we speak’s monetary infrastructure doesn’t work for a lot of the world’s inhabitants. We’re making use of superior know-how and human creativity to resolve what legacy establishments can’t or received’t, in an effort to unleash the financial energy of the World Majority.
The Analytics Engineering workforce serves as a layer between back-end engineering groups and numerous Enterprise Analysts. We construct infrastructure, we clear up information, we arrange duties, and we be sure that information is simple to search out and prepared for use. We’re right here to ensure information is clear, dependable, and reusable, so analysts on groups like Advertising and marketing and Operations can deal with evaluation and producing insights.
What does your information stack appear like?
We primarily use dbt to develop our infrastructure, Snowflake to curate, and Looker to visualise. It’s been nice that Atlan connects to all three, and helps our strategy of documenting YAML information from dbt and mechanically syncing them to Snowflake and Looker. We actually like that automation, the place the Analytics Engineering workforce doesn’t want to enter Atlan to replace data, it simply flows via from dbt and our enterprise customers can use Atlan immediately as their information dictionary.
Might you describe your journey with Atlan, up to now? Who’s getting worth from utilizing it?
We’ve been with Atlan for greater than two years, and I consider we had been one among your earlier customers. It’s been very, very useful.
We began to construct a Presentation Layer (PL) with dbt one yr in the past, and beforehand to that, we used Atlan to doc all our previous infrastructure manually. Earlier than, documentation was inconsistent between groups and it was typically difficult to chase down what a desk or column meant.
Now, as we’re constructing this PL, our aim is to doc each single column and desk that’s uncovered to the top person, and Atlan has been fairly useful for us. It’s very straightforward to doc, and really easy for the enterprise customers. They’ll go to Atlan and seek for a desk or a column, they will even seek for the outline, saying one thing like, “Give me all of the columns which have individuals data.”
For the Analytics Engineering workforce, we’re usually the curator for that documentation. After we construct tables, we sync with the service homeowners who created the DB to know the schema, and after we construct columns we set up them in a reader-friendly method and put it right into a dbt YAML file, which flows into Atlan. We additionally go into Atlan and add in Readmes, in the event that they’re wanted.
Enterprise customers don’t use dbt, and Atlan is the one approach for them to entry Snowflake documentation. They’ll go into Atlan and seek for a selected desk or column, can learn the documentation, and might discover out who the proprietor is. They’ll additionally go to the lineage web page to see how one desk is expounded to a different desk and what are the codes that generate the desk. The very best factor about lineage is it’s absolutely automated. It has been very useful in information exploration when somebody is just not aware of a brand new information supply.
What’s subsequent for you and your workforce? Something you’re enthusiastic about constructing?
Now we have been wanting into the dbt semantic layer previously yr. It’s going to assist additional centralize enterprise metric definitions and keep away from duplicated definitions amongst numerous evaluation groups within the firm. After we principally end our presentation layer, we’ll construct the dbt semantic layer on prime of the presentation layer to make reporting and visualizations extra seamless.
Do you might have any recommendation to share along with your friends from this expertise?
Doc. Positively doc.
In one among my earlier jobs, there was zero documentation on their database, however their database was very small. As the primary rent, I used to be a powerful advocate for documentation, so I went in to doc the entire thing, however that would stay in a Google spreadsheet, which isn’t actually sustainable for bigger organizations with tens of millions of tables.
Coming to Tala, I discovered there was a lot information, it was difficult to navigate. That’s why we began the documentation course of earlier than we constructed the brand new infrastructure. We documented our previous infrastructure for a yr, which was not wasted time as a result of as we’re constructing the brand new infrastructure, it’s straightforward for us to refer again to the previous documentation.
So, I actually emphasize documentation. Whenever you begin is the time and the place to actually centralize your data, so every time somebody leaves, the data stays, and it’s a lot simpler for brand new individuals to onboard. No person has to play guessing video games. It’s centralized, and there’s no query.
Typically completely different groups have completely different definitions for related phrases. And even in these instances, we’ll use the SQL to doc so we are able to say “That is the method that derives this definition of Revenue.”
You wish to go away little or no room for misinterpretation. That’s actually what I’d like to emphasise.
Anything you’d wish to share?
I nonetheless have the spreadsheet from two years in the past once I seemed for documentation instruments. I did lots of market analysis, taking a look at 20 completely different distributors and each instrument I may discover. What was necessary to me was discovering a platform that would connect with all of the instruments I used to be already utilizing, which had been dbt, Snowflake, and Looker, and that had a powerful help workforce. I knew that after we first onboarded, we’d have questions, and we’d be establishing lots of permissions and information connections, and {that a} sturdy help workforce can be very useful.
I remembered after we first labored with the workforce, everyone that I interacted with from Atlan was tremendous useful and really beneficiant with their time. Now, we’re just about operating by ourselves, and I’m all the time proud that I discovered and selected Atlan.
Photograph by Priscilla Du Preez 🇨🇦 on Unsplash