Post
Stitch is sunsetting the free plan
Like many other companies, we and several of our clients have relied on the free version of Stich in certain use cases for transferring data from one platform into another (i.e. Facebook Ads to BigQuery).
The minimum plan includes 5M rows and costs $100/mo. While 5M rows is quite a lot, in some cases, we'd only need >500K rows and then it can be hard to justify paying for 5M.
What are some of the other tools to consider for this purpose and does any of them still have a free plan?
Really hope their open-source contributions aren't dying with the free plan.
https://www.stitchdata.com/blog/free-plan-sunset/
Share
There are a few "free" options that instead cost you in engineering time and infrastructure, such as Meltano (built by the amazing Gitlab data team) and Pipelinewise (by Transferwise). Both of those tools allow you to host your own Singer taps, which should provide easy compatibility with Stitch sources and destinations, and Transferwise have even improved some of the taps with their own versions if I understand correctly.
Personally, I'd advise just paying the $100/mo after December, which is still very reasonable in the domain (about the same as three EC2 t3.medium instances). When I've compared products in the data loader domain Stitch has always come out on top for our use case even at the $100/mo level. Other options I investigated include Fivetran, Alooma, Blendo, Segment Sources (for few sources), and Matillion (AWS marketplace), all far more expensive last time I checked (for our use case) and often more complex to set up.
For what it's worth, we currently use Stitch's paid plan and are happy customers - we're currently a couple of tiers up but have moved up over the years from the free tier. I'd make the same decision again if we were paying the $100/mo from the start. That said, I empathise with the low-budget startup position, having been there myself for a few years - just try to look out for your future self (rationalise that cost to the client / CFO) and not over-commit to maintaining infrastructure!
AWS Glue might be the cheapest if it’d fit your needs—its first million rows are free, though the pricing after that is confusing as its billed per second a job is running. Skyvia has a free plan, though that only covers 5k records a month—500k rows would run $99/month (or $79 with an annual discount).
One more general option here I’ve used in part in the past is Supermetrics (which I’d used to pull data into Google Data Studio and Google Sheets)—and they have a tool to load Facebook Ads into BigQuery. But most of their connections start at $99/month, while their BigQuery offering requires contacting support for pricing, so that likely won’t work out better for you.
Then, in other similar seeming tools, Blendo starts at $150/mo; Hevo and Panoply start at double that price. Fivetran looks like it’d cost $1 per thousand “monthly active rows”, so far higher at first but possibly cheaper if all 500k of your rows aren’t updated each month. Xplenty doesn’t list pricing, but shared pricing stories place it at starting around $100/month.
Perhaps that’s why Stitch dropped their free plan; seems they were unique in having one!
Good list of tools you mentioned there @maguay!
To be honest, we find that most of them are actually quite overpriced - depends on the use case, of course, but we tend to use Stitch and others when we need to prototype something or just add a single integration.
For anything more advanced (that our clients usually require) we build a custom data pipeline on GCP or AWS using tools like Cloud Compute Engine, Dataflow etc. and this enables us to transfer billions of rows for a fraction of the cost of any of these tools.
I guess it just means it's time for us to move on to setting some Singer instances for quick prototyping and smaller single-purpose data pipelines. Singer has worked quite well actually, just a bit more work when compared to Stitch.
I'm not too surprised the pricing feels high—if anything, I was surprised that no other products seem to have a reasonable free option (though hopefully someone in the community with a bit more experience here will know of something we're missing!).
Gottcha, makes sense. So that per-second pricing on AWS works out far better than per-row pricing?
How much additional work is it for you to use Singer self-hosted vs. something like Stitch?
Exactly, per-second pricing is a lot cheaper compared to per-row pricing, in most cases at least.
It's definitely more work to boot up a self-hosted Singer instance but it isn't something too difficult either. I'd say the first time it can take 2-5 hours and then it's ~1 hour if you've done It before.
Thank you for asking this question. We have a very low usage, but it has been instrumental in getting us setup with some regular analytics for Harvest (which has a Singer tap). If we can add a bunch more integrations with Stitch (Salesforce, Quickbooks, etc.), then it might add up, but $100/month is still a lot for us for something like this.
How difficult is it to use Singer instead of Stitch?
Singer is definitely a more technical tool and a developer with Python skills should be involved. The documentation is quite good and the initial setup should not take more than a day.
Looking for a better way to plan remote meetings across time zones, and keep up with events. What software is doing that best today?
@dan-hunt Thanks for sharing this incredibly detailed reply, and it’s great to hear how your team approached the purchase decision and found Stitch to still be worth its pricing. Those hidden maintenance costs can absolutely add up with self-hosted infrastructure.