What are you able to do with a reverse-ETL?
Hey everyone! In today’s post, we’re going to dive into reverse-ETLs. They are one superpower you can use to leverage your data for Growth.
Here’s what we’ll cover in this article:
What the f**ck is a “reverse-ETL”?!
[Sales use-cases] How Notion, Retool, & Qonto empower Sales teams with a reverse-ETL
[Marketing use-cases] How Canva is personalizing marketing for 55M users
Replicating this for Ulysse
[Customer support use-case] How Loom is prioritizing Zendesk tickets with product data
What the f**ck is a “reverse-ETL”?!
The role of a reverse-ETL is to send data from your data warehouse (or database) to your marketing & sales tools, such as Salesforce, Hubspot, Customer.io, Braze, or Facebook & Google Ads.
This is their main job. Because with a reverse ETL, you can actually send data to any tool you want, including also customer support tools (Zendesk, Intercom, Front…) and Analytics tools (Amplitude, Mixpanel…).
Basically, if you download data from your database as a CSV, to import it back to your mailing tool, you are doing the job of a reverse-ETL, but manually.
When companies start to scale, this becomes unsustainable, so that’s why you need a reverse-ETL!
(BTW, I am not going to explain what “ETL” stands for, for today. If you know what that means, please drop a comment below 👇).
The main reverse-ETL on the market are Census & Hightouch.
I dug into their case studies to let you know how the fastest-growing companies in the world are growing thanks to their data (and also handpicked some examples from my experience 😉).
Let’s go!
[Sales use-cases] How Notion, Retool, & Qonto empower Sales teams with a reverse-ETL
Notion: sending product usage data to Salesforce & Facebook
People often say Notion grew mostly thanks to virality and product-led growth.
That’s true, but when you dig a little around the web, you discover they were also pioneers in other stuff, such as… Using data to empower the sales team.
This is what I found from Notion ↔ Census case-study:
We wanted to export data from our warehouse to different destinations, such as Facebook or Salesforce for some set of users like, ‘people who have been active in the last 90 days’.
While the article doesn’t explain precisely what flows were live at Notion, here’s what we can imagine:
1. Notion sent those “active in the last 90 days users” to Salesforce, so the Sales team could upsell them:
Let’s say you are working in a big company, using Notion with some colleagues from your team, but not everyone in the company.
The Notion salesperson could outreach you and your colleagues to encourage you to invite as many colleagues as possible to use Notion, showcasing the benefits of having a “shared wiki for centralizing information”.
→ More users from your company in Notion = more revenue for Notion. 🤑
2. Notion sent those “active in the last 90 days” users to Facebook, to create custom Lookalike and retargeting audiences:
🍪 Cookies are dead (RIP). So how to still create relevant ads/targeting in this world?
By using your 1st party data! That’s what Notion probably did here.
They synched those active users to Facebook so they could target similar users to the ones who were already paying for their product, using Facebook Lookalike audiences.
One recurring use-case is also to exclude active users from your retargeting campaigns (as you know you already converted them).
→ Less ad spend on already-converted users = more ad spend on relevant targeting = more revenue for Notion. 🤑
How Retool increased Sales by 32% by sending product data to SDRs
If you don’t know about Retool yet: it’s a tool for internal tools:
You can easily plug your database to Retool, and you are able to create interfaces for your team so they can interact with your data.
The most common use-case is for customer success teams: you can create an admin panel so customer success people can create, read, edit and delete data.
So what did they do with Hightouch?
As we have just seen, connecting a database is a key action Retool’s users can do.
The simple fact of informing Salespersons about which database the users plugged helped improve the answering rate by 32%.
Here’s the flow:
Retool’s user plugs database (let’s say PostgreSQL)
The salesperson from Retool will see it in his Hubspot dashboard (thanks to data brought by the reverse-ETL)
He’s able to send a personalized message with some documentation about PostgreSQL: how to set it up, specific tips…
That sounds basic, but numbers are here: personalization works. You should put effort into doing it at scale.
Qonto: prioritizing sales based on collected company information
This last example is one project I had the chance to work on, back in my days at Qonto.
Qonto is a B2B neobank focused on SMEs. The registration process is self-serve, and most of the leads are coming from inbound.
That means the biggest lead generation machine is… The product! And users coming by themselves on the website.
Now, as you can guess, there are many steps in the registration funnel: Qonto has to collect legal documents, for both the owner and his company (KYC/KYB).
Many of the leads are actually dropping and not finishing the process down the line, often because they don’t have the documents with them.
Sometimes, companies’ owners don’t even know where are those documents. They need help to finish this process.
That’s where Qonto’s sales team comes in: to outreach the leads to help them finish the registration process and give them some advice on how to use their account/cards with their team.
The problem: there were actually too many leads to handle. I know, that’s rather a good problem. 😁 But still, it was a problem.
Calling leads is time-consuming, and human resources were limited in the sales team at this time.
So that’s when we decided to leverage data to prioritize leads:
High potential leads (enterprise, big SMEs) were sent to Salesforce so Qonto’s sales team could outreach them by phone
Low potential leads (freelancers, small companies) were sent to Customer.io where we were sending educational email campaigns with tips to help them finalize the process.
Everything was handled by Segment there.
(Yes, a customer data platform can also play the reverse-ETL role… But this will be for another post 😉).
The sales team is able to focus on the most important leads, without even having to think about it.
[Marketing use-cases] How Canva is personalizing marketing for 55M users
Let’s have a look at how Canva is thriving with a freemium model, data, and 55 million users (wow!).
Canva (as Notion) mainly grew organically, with users testing the product freely, and coming back whenever they want to create designs.
So one of the keys for them was to boost user engagement as much as possible, to make sure users stick to the product.
This didn’t happen by magic. But by efforts from data and growth teams.
As you can imagine, Canva’s users generate a lot of data:
What type of design do they build? (Presentations, logos, prints…)
What features do they use? (Templates, background remover, stock photos…)
How often?
For what purposes? (Work, personal…)
The list is gargantuan.
Betting on personalization (yes, personalization, again!) was a game-changer for Canva’s growth & data teams.
As they say themselves:
We realized we were sending users the same thing, and we really needed to start segmenting more.
Before implementing their reverse-ETL, they simply didn’t have a way to send data to Braze (emailing tool).
Now, Canva uses its data warehouse as its single source of truth for analysis. And Census to push data to operational tools.
Let’s say you’re using Canva on a daily basis, for presentations at work (similar to PowerPoint).
Canva will know it, from the data you generated in their product
They will send this data and modelize it (transform it) into their warehouse
Then, data will be sent to Braze, allowing the marketing team to customize emails:
to give you tips about doing better presentations in public
to tell you about presentation-related features (such as using a second screen for your notes within Canva, for example)
And so on! Now, repeat this x1000 and you might have an idea of what Canva’s teams are doing for marketing. Love it 😍
Replicating this for Ulysse
I am currently implementing for Ulysse something quite similar to what Canva did (for now, we have a little less than 55 million users 🔜).
(Ulysse is the best flight booking website on the globe — period).
One thing we noticed: customers are often asking the support team for the best places (restaurants, things to visit…) to go before their trip.
→ We now have a list of many ‘good addresses’ to go around the globe.
I can just query the data using some SQL, and create a Customer.io email campaign to give the users some advice on what to do during their trip.
This ultra-personalized message is sent 7 days before the departure flight date.
For instance, if you are traveling to Montenegro you’ll receive this email:
This is a nice way to keep customers engaged with our brand (even after their purchase) and thus increase retention.
[Customer support use-case] How Loom is prioritizing Zendesk tickets with product data
Using a reverse-ETL is not limited to growth, sales, and marketing!
You can for example send your data to other operational tools, such as customer success tools (like Zendesk, Intercom, and Gorgias)…
That’s exactly what Loom did with Census.
The problem: a single Loom user can be on several plans
Loom wanted to prioritize support tickets based on the plan of the user.
This was difficult for them, as many users are connected to several organizations (with different plans).
So they had to query their database (using SQL) to create a list of all the plans the user is subscribed to.
Ultimately they have a list, saying something like:
We have this user, he his subscribed to:
High-priced plan with Organization A
Middle-priced plan with Organization B
Free tier plan for himself (Organization C)
Then, Loom data team sent this data to Zendesk, allowing the customer support team to prioritize tickets based on what plan the users are paying for.
The results look astonishing:
Before we were diving through a sea of tens of thousands of tickets and which, without Census, was basically impossible to handle. Census solved some foundational problems that we needed to have fixed.
Said Buddy Marshburn, data engineer at Gorgias.
That’s it for today, everyone!
Hope this gives you a better idea of how to leverage your data with a reverse ETL!
If you’re interested in setting up similar things at your company, feel free to reach out: getgaas.com.
We’re specialized in implementing such things, and we’ve already implemented “growth stacks” with reverse ETLs at several companies, such as Qonto, Karbon Card, and Ulysse.
Let’s chat!
Victor