Show HN: Polytomic Connect – API for two-way ETL and data syncs with customers | Hacker News Reader

Show HN: Polytomic Connect – API for two-way ETL and data syncs with customers

65 pointsgms33 comments

Hello HN, we’re Ghalib and Nathan from Polytomic. Today we’re launching our second product, Polytomic Connect: https://www.polytomic.com/connect (documentation here: https://apidocs.polytomic.com).

Connect is an API you can use in your own products to either pull your customers’ data into your own systems, or push data from your own systems to your customers’, or both. We have first-class support for data warehouses too. And there's a self-hosted deployment option.

You can set up automatic data syncs in any direction between data warehouses, databases (including CDC streaming), cloud applications (e.g. Salesforce, Zendesk, etc), spreadsheets, cloud storage (S3 etc), and arbitrary APIs and webhooks. We take care of authentication, automatically pushing updates, type conversions, rate limits, scaling to handle large volumes, monitoring, and alerting.

You can also sync updates from custom queries powered by data models you define as well join data across disparate systems (e.g. HubSpot + Stripe + Airtable) to sync fields from them in one payload to any destination.

You can see demo videos and code examples for a small subset of use cases here: https://apidocs.polytomic.com/guides/code-examples/overview.

We’d love to take your comments and feedback! Happy to answer questions too.

Data team manager here, qq: how do you handle type conversions between different systems? Nice work overall!

(Polytomic co-founder here)

We automatically do the right thing depending on the situation. The type engine is highly-precise when we generate our own schema. For example, JSON objects and arrays get converted to native struct and array types in data warehouses (e.g. Databricks, BigQuery) rather than strings.

But we're forgiving in other situations where we don't control schemas. For example, we'll automatically convert a source string type to a date when mapping to a Salesforce or Zendesk (or other cloud app) date field.

What's the ELI5 here?

If I'm pushing and pulling customer's data into my own system (e.g. postgres), what sort of customer data are we talking about? Can it sync with data on their own filesystem? That would be neat. Or it just for e.g. syncing a customer's google drive with my postgres type of thing? Looks interesting and like something I may want to use but not sure I entirely grok it yet.

We're talking about system-to-system pulls and pushes.

For example, maybe your product needs to process data from your customers' sales and customer support systems. In that case you can have automatic updates show up in Postgres from your customers' Salesforce and Zendesk instances. Does that sense?

Alternatively, many products want to the ability to push data to their customers' systems. For example, suppose you're a billing software vendor showing data in your own portal that customers log in to. Your customers may also want you to automatically push this data to their accounting systems or data warehouses so that they can do their own analysis.

Does that make sense?

I work at a venture firm. Polytomic is pretty impressive in its functionality. We use it to integrate data providers (Harmonic, pitchbook, etc) with our CRM (Affinity).

(Polytomic co-founder)

Thank you, whoever you are! :D Hopefully we can continue serving you.

Do you want to share more about collision detection? Last write wins? I assume it's harder when things are being synced in the background on a schedule.

(Polytomic co-founder here)

Each sync in Polytomic is one-way so we're not forced to deal with collisions all the time.

But you can, on a per-field basis within a sync config, declare that field not to be synced if the destination system already has a value in the corresponding field.

This setting is a proxy for deciding where your source of truth is for each field if you are indeed setting up two-way syncs.

Most customers are pulling data into their data warehouse, then syncing from queries that generate other values back into other systems. This issue doesn't come up there. But customers of ours doing two-way syncs between, say, HubSpot and Airtable or such do need to decide where the source of truth is for each field.

> Each sync in Polytomic is one-way

> But customers of ours doing two-way syncs

Could you clarify where two way sync fits in to the picture?

Each integration is supported as both a source and a destination. So you can have two one-way syncs set up between systems A and B, thus moving data in two directions.

> we're not forced to deal with collisions all the time.

I don't really follow how this approach avoids having to deal with collisions all the time.

Sorry, my fault for being unclear.

Meaning, sometimes people set up one-way syncs only. Our syncs don't have to be two-way.

I work at Retool and we're a Polytomic customer with users across multiple teams - it's a major part of our data infrastructure. The product can handle a lot.

(Polytomic co-founder)

Thanks for the kind words! Pleasure serving you.

Hey Ghalib and Nathan,

Amazing work! Two questions:

1. What is the best avenue for potential customers to express interest in the addition of a new data source or sink?

2. I assume this runs on Polytomic owned compute? Is there a way to bring your own compute, or is it on the roadmap?

Thanks!

Hello! Polytomic co-founder here.

1. All integrations are built on request (see instruction at top of this page to submit one: https://www.polytomic.com/integrations). We're able to turn them around very quickly (our record is two hours from request submission time).

2. Yes, though you can bring your own compute too. We have a self-hosted distribution that you can deploy to your own VPC using Docker and our Terraform module (see details here https://docs.polytomic.com/docs/on-premise-setup). Customers that are public/large enterprises often go for this.

If you want to chat over email shoot me a message at ghalib@polytomic.com.

Interesting. I am curious how you cover the case where one of the system might not be used for short period of times (e.g. between 2-4am) or capacity limitations.

That's no problem for us. We'll just resume where we left off when it comes back on.

I'm having a hard time finding docs on the ETL itself. I want to see how flexible the language is.

What are you looking for exactly? Happy to answer questions or point you to the right place.

The language. I want to see it. I want to get an idea of how capable it is.

Hopefully the links below are helpful!

Our clients come in multiple languages. You can see the list here: https://apidocs.polytomic.com/guides/native-clients.

You can see here for sample code: https://apidocs.polytomic.com/guides/code-examples/overview.

The full API reference is here: https://apidocs.polytomic.com/api-reference.

It’s nice to see a rETL provider not try to confuse the market by positioning itself as a CDP.

(Polytomic co-founder)

Haha! I’ve been in this market for over ten years now. The term ‘CDP’ is notoriously ill-defined.

More generally, we move data in all directions, rETL is just one part amongst many others.

We used Polytomic in my last job to sync data between postgres, google sheets and quickbooks. Ghalib and Nathan were helpful in figuring out how to make sense of the messy QB data and put a meaningful structure around it. They worked with us on building out special reports that pulled specific data our team required. The support was always outstanding and the product never failed, always worked smoothly. Excited to learn about the new launch and looking forward to exploring it more!

(Polytomic co-founder)

Thank you for the kind words Ani!

QuickBooks Online or QuickBooks Desktop? Can’t tell from the docs.

(Polytomic co-founder here)

QuickBooks Online. All the applications we support are cloud versions.

We started using Polytomic for only reverse ETL, but are now working to move our ETL processes to it as well.

They've been really great to work with and they've been able to handle our data as we've scaled at Vercel.

Congrats on the launch!

Thank you! We're happy to serve.

I used Polytomic at the last place as a reverse ETL. The team is amazing and the product lets you move your data around without bugging your developers.

Congrats on the launch!

Wow thank you Noa! Hope you are well!

Just giving credit where it is due :-)

Reply on news.ycombinator.com