How I saved $1,500 (so far) with an Airtable DB and Replit backend

It costs almost nothing to do this and once things take off you can migrate to a proper setup. For a solo bootstrapper, having my largest expense almost disappear was a huge weight off my shoulders. You never know how long it’s going to take to A) find the right idea and B) get traction.

The thinking that led me here

When I used to boot up startup projects, I’d often go straight to designing the database schema. I would take my idea, break it into pieces, and build the database before doing anything else.

Fast-forward through many years of in-the-wild learning and I can tell you that this isn’t great. You get bogged down in building out all the tech before you even get a feel for how your app looks, smells, and feels. It’s easy to get married to your initial idea and you end up making all sorts of revisions to what you thought was sealed functionality as you get closer to actually having a purchasable product.

Mostly though, it creates a tax on changing your mind, which is a crucial ability in the early days of building any product.

This time around, I decided to see if I could successfully build an app that uses Airtable as its database and Replit as its servers — at least in the beginning.

(if you aren’t aware, Airtable is basically Google Sheets but with more of a database flavor to it, and Replit is a browser-based coding environment that also runs your code)

The pivots that proved this technique

9 months ago I started building an app called New Teammate, which aimed to help companies create spaces to welcome new hires with intro videos of their new work family. After lots of conversations with potential customers and just as I started creating tables in Airtable, I began feeling like this idea wouldn’t be urgent enough for buyers.

So I pivoted to a tangential idea, mostly because it was a much simpler product to launch and get feedback on. I built Snaptro as a link-in-bio, page creator where you could introduce yourself with a 60-second video.

Airtable made it easy to manage the data and to monitor and debug production. I could quickly create automations like "send me an email when someone signs up." And crucially, I knew this was a temporary setup so I had less angst about getting the schema just right or slapping a new table together. No worrying about proper indexes or optimizing data relationships either.

Weeks went by and I wasn’t getting the kind of validation I hoped for. Based on feedback, I pivoted again to another related idea: creating a hiring platform focused on the candidate experience. I didn’t even bother naming the product this time while I hunted for validation. More tables and fields were birthed as my efforts to gain a foothold went on.

Months since and a product called Dropboard emerged from all the experimentation. I feel more bullish that I’m on the right track, but I don’t have enough validation to know for sure so Airtable and Replit are STILL powering everything.

Let's do the math

Now I think back to what it would have been like to have gone straight to some kind of production-ready system from the beginning.

If I was going to build out real infrastructure, I’d want to get it right from the get-go. Security and backups need to be well thought out, and I’d want the ability to scale quickly if things go horribly right. I realize there are cheaper options, but I like to go with the Postgres flavor of AWS Aurora, particularly the serverless option. I’ve been in situations where database capacity is a moving target. When your app starts to grow it can be maddening to deal with configuration changes when all you want to do is capitalize on the growth that’s finally happening.

• Amazon Aurora is about $55 a month for a low-end implementation. The “serverless” option bumps that up to $89 to get the benefits of auto-scaling. I’m ignoring other potential costs for things we might need like RDS Proxy for improved connection management.

To create production-ready servers I'd probably have used Amazon Fargate or Lambda, because again, I want something that could easily scale if things went well. Depending on a few factors, this may have necessitated a load balancer, a Docker container registry, and/or API gateway.

• Let’s keep it simple and say that Fargate costs about $43 a month for a basic implementation and another $17 for a load balancer. In terms of security, a Virtual Private Cloud with a NAT Gateway costs about $32.

That means when I started my journey about 9 months ago, I would have paid at least $796 by now on a production database that would have had almost 0 workload on it. For servers, I would have paid around $830.

By contrast, my only infrastructure cost so far has been the $7 Hacker plan which makes sure my Replits are “always on.” Multiply this by 9 months and you get $63 versus $1,626.

And because I haven't reached traction yet, the savings are still flowing.

Final thoughts

I realize that setting up production-ready infrastructure is a spectrum. You don’t have to get it all right the first time and there would have been ways to start smaller and work my way up.

But with Replit and Airtable, that decision is taken out of my hands. It creates a sense of freedom in my product discovery and even in my daily philosophy and emotional state, reminding me like an affirmation, “I can still change.”

If you guys liked this, I could write another post about how I made Airtable work…there were a few curious challenges to overcome… I could also share what it’s been like to run a suite of apps off of Replit, so I’ll gauge interest in the comments!

I post weekly YouTube videos of my progress towards building a $1M startup. If you’re interested you can check me out here!