28.8 C
New York
Friday, August 15, 2025

Constructing a SaaS that makes cash in underneath 24 hours | by Uri Merhav | The Startup | Nov, 2024


Not a get-rich-quick information. However a “attempt stuff fast” information

A very long time in the past, I used to be a salaried worker for LinkedIn, main the usual company life, as a Machine Studying Engineer.

Life was good, life sucked. You get a great wage, you play loads of ping pong, and you’ve got good work-life stability. However you spend loads of time doing stuff you understand to be ineffective or unimportant. You feed the company paperwork. You’ve got repeating assembly that ought to have stopped 3 months in the past however nobody will pull the plug on it. You relaxation, you vest.

A reasonably trustworthy illustration of company life in Silicon Valley

Then I met a intelligent particular person in San Francisco. He informed me how he made a residing — he constructed his personal SaaS. No traders, no weeklies, no JIRA tickets. He simply constructed an answer that’s tremendous helpful, found out the right way to develop, and lives off its income. The tech equal of attempting to find meals. I didn’t know one may try this.

So this submit is meant for you. Leaving your soft job at Google to pursue a bootstrapped startup could also be daunting. Heck, by no means thoughts daunting. It’s most likely a mistake. Don’t do it. But.

I wish to present you that dipping your toe is straightforward. You possibly can hack collectively an internet site that gives actual worth to your customers, and cost them for it efficiently, in underneath 24 hours.

Possibly it will present a little bit of promise. Possibly you’ll be tempted daily to make it higher and never watch Netflix. Possibly someday you may make a reasoned calculation in your head the place it’s not a foolish gamble to go away the whole lot and guess on it, as a result of you understand what you’re stepping into.

I’m conscious of the click-baity nature of an article that reveals you the right way to construct something in any respect in a day. So right here’s the caveat. My declare is distinct from saying I discovered the infinite cash hack. This isn’t a “DO THESE 3 THINGS TO BE WEALTHY FOREVER” article.

What you are able to do, is give you a shortly testable concept, piece collectively a modicum of utility to another person on the planet, and get them to drag out their bank card for it. In a day. No weekly standup or JIRA tickets required.

And I believe that’s fairly superior, and a ton of enjoyable to do. So I’ll present you ways, and perhaps it will encourage you to attempt it your self.

In case you’re simply beginning out, I believe you must ask your self simply three questions, and proceed even when your confidence within the solutions is just medium:

  1. Am I fixing one thing that folks must get solved?
  2. Is there an affordable expectation that folks will give me cash for it?
  3. Can I launch it inside days, not years?

You do NOT must persuade your self that no one else on the planet has acknowledged this want (if that’s the case, most definitely your concept sucks or extraordinarily onerous). You do NOT must have a transparent estimate of the market measurement or a narrative of how this will likely be a rocketship to the moon which will get to a billion greenback valuation.

The echo chamber round entrepreneurship is skewed, and focuses on the blokes with the massive pockets. VCs are professionals who make investments massive quantities of cash and are correspondingly solely interested by multibillion greenback exits to justify a big deployment of capital. They wish to discover startups whose execution requires inordinate capital funding, with proportionally huge prizes. In case you’re making an attempt to construct the subsequent social community, you’ll want a whole bunch of thousands and thousands of funding till you hit profitability. In case you’re simply constructing a self-owned bootstrap, “this will get to a billion greenback” ceases to be a prerequisite, and virtually any concept could make you personally rich.

I’ve seen aspiring founders drown on this countless market analysis timesuck, after which hand over. Don’t spend your time asking what’s the entire addressable market, how huge are you able to develop, what’s the anticipated buyer acquisition prices, and so on.

Most definitely you’ll quantity to zero resulting from evaluation paralysis and get again to taking part in ping pong as a result of it’s extraordinarily unfun to google for Gartner Stories concerning the complete market projections for Broccoli farming apps in 2025.

You’ll achieve extra by making an attempt.

I like this meme as a result of it’s true for therefore many issues

As for what your concept is — I don’t must inform you the right way to discover concepts. In case you reside on this world, you’ve encountered an itch that isn’t correctly scratched. Scratch it.

In case you haven’t encountered something you are feeling is damaged, a hack I like is looking out current transactional markets like Fiverr or Upwork, and see what jobs are getting posted repeatedly and will conceivably be automated. However that’s simply the machine studying engineer in me, gravitating to my pure benefits. It’s best to most likely discover your personal concept supply.

Feeling the itch

I felt this itch a bunch of instances, as I’m demoing my startup’s doc extraction instrument (docupanda.io). We work with many healthcare and monetary firms. They wish to check us with a doc, however must first eliminate any personally identifiable info, which is a drag.

If solely there was a instrument to robotically detect delicate info in a doc and overlay a giant previous black rectangle on it!

Validating the itch

Simply google “PDF redaction”.

Oh no, there are three different firms that already supply precisely the service we wish! I suppose meaning I considered a nasty concept and want a extra distinctive concept that no else is fixing.

Not.

The existence of competitors that’s fixing this precise want, and shopping for up sponsored clicks from google, validates that it is a area of interest that truly makes cash.

Clicking by the primary three outcomes, I see the competitors has weaknesses:

  1. Considered one of them is a desktop app: many desire an online resolution
  2. The opposite requires a bank card earlier than you even get began: actually this could possibly be optimum pricing for them — however it means all of the individuals who refuse to start out with a bank card will maintain wanting and go down the record to search out me.
  3. I spend a half hour making an attempt them out. A couple of of them introduce loads of complexity. Most of them over-redact, e.g. eradicating my financial institution’s telephone quantity and each single transaction in my financial institution assertion, as an alternative of solely eradicating the figuring out info.
  4. Many options are simply help instruments to allow you to draw rectangles on a PDF. I wish to determine and draw them robotically, saving the person time. So it’s a really differentiated product, which I believe is far more helpful.

Many individuals counsel constructing an internet site that solely guarantees some performance and doesn’t do something, to judge how many individuals join. First see if individuals need this, and solely later implement the performance.

Possibly it’s sound recommendation, I don’t know. For positive it’s remarkably unfun and unsatisfying to construct. So I can’t in good conscience suggest it. I recomment that you simply construct an online web page that truly does the factor you’ve promised, even when it’s a really primary iteration of that factor.

Spinning up a frontend and internet hosting it

This part actually will depend on your skillset and background.

For me, the best approach to spin up a brand new SaaS is to git clone my current codebase for a SaaS I cofounded, DocuPanda.io. Then trash 99% of the code, and rewrite the present onboarding web page from scratch, in order that now it redacts PDFs. I’ll reuse my current privateness web page, pricing web page, touchdown web page, and so on, and immediate GPT to alter up the language to replicate what our instrument does as an alternative of DocuPanda.

You possibly can say that I’m dishonest — I’m beginning out with a template for a working SaaS to indicate how simple it’s to make a brand new working SaaS. Doesn’t appear truthful.

Guess what: you’re additionally beginning with a template of a working SaaS. You possibly can purchase a template like Divjoy, which is great however considerably outdated. Or perhaps you wish to use Saasfly.io template. Or you possibly can one click on deploy a easy web site from Vercel, which is a bit naked bones however you may make modifications from there. I’d suggest internet hosting your resolution on Vercel no matter which template you utilize, by the best way (not affiliated, only a glad buyer).

There’s a great likelihood you’re studying this text and you understand how to code typically however not the right way to construct an precise web site — simply flick through the templates, comply with alongside the docs, then make modifications. GPT is your good friend and can assist you redesign whole pages and work out the language. You’re not aiming for perfection right here, you’re aiming to be launched.

In case you don’t know the right way to code in any respect, you could possibly use bubble.io and different disruptors that allow you to get away with not realizing the right way to code and nonetheless constructing an internet site. I don’t have expertise on that entrance, however I firmly consider that any resourceful particular person with GPT and Google will work out the right way to host their very own, fairly performant two-pager web site that may do one thing advanced that’s helpful to its customers.

My tech stack

My tech stack, should you’re curious, is simply:

  1. Vercel.com internet hosting for the web site
  2. React, Supplies UI frontend
  3. Google Firebase for authentication
  4. Stripe for billing
  5. Docupanda.io for understanding paperwork and redacting them.

You’re most likely not aware of the final one. I’d already talked about this, however to maintain price having a full disclosure, I’m its Cofounder. DocuPanda is a service which is usually designed to grasp paperwork. So for instance it’d have a look at rental lease PDF and provides the lease, the fee schedule, transfer out date, and whether or not pet alligators are allowed.

DocuPanda may floor its understanding in precise web page and bounding containers, exhibiting the place within the doc a end result got here from (e.g. highlighting the web page and part the place the lease quantity is talked about). So, I believe “The place’s the personally delicate info on this doc”, is one thing that I ought to have the ability to construct with DocuPanda’s public API. I believe it’s ok to construct a minimal viable product for my PDF redaction instrument. So long as I give my customers a approach to repair errors and evaluation outcomes, they will most likely discover worth from our this service, regardless of its limitations.

Identify: The product redacts paperwork. The area redact-doc.com is offered. No different enterprise appears to make use of this identify. Accomplished.

Brand: Immediate Flux to generate a easy icon that reveals a redacted doc.

Not unhealthy.

Model Identification: no matter you inherited within the template you’re utilizing. You possibly can change it later. I modified the colour from inexperienced to purple. Nobody cares.

Accomplished.

That is the majority of our work on this venture. Whether or not this will realistically take hours and never days will depend on your concept. In case you really feel that you simply’ll most likely want many months to launch, I strongly counsel you strip away out of your concept. In case you actually can’t, I counsel you progress on to the subsequent concept. It’s higher to launch quick and fail a few instances earlier than you commit months or years. You’ll be taught quicker.

So let’s motive by what’s the naked minimal we’d like our internet web page to do:

  1. We wish to determine and suggest rectangles that include delicate info within the PDF the person has uploaded. We’ll try this utilizing DocuPanda.io.
  2. We wish to construct a frontend that lets individuals add a doc, and see the rectangles overlaid over their doc. They need to have the ability to add extra redactions, and once they click on save, it ought to generate a brand new PDF with black rectangles wherever the person selected.

Let’s first construct the frontend. Think about we have already got a bunch of rectangles round delicate info that we’ve recognized. How would we construct internet web page that overlays rectangles on high of the PDF, lets customers drag them round, and lets them add extra of them? Sounds daunting, I’m not a frontend whiz.

Immediate GPT O1 to do it, and don’t take no for a solution. Right here’s a video of me constructing the fundamentals of this element in half-hour.

Within the video I attempt to keep away from utilizing my prior information of the right way to construct frontend code. I attempt to persist with simply prompting GPT like a monkey repeatedly. Do that sufficient instances and also you’ll be taught precise frontend. However once more, the main focus right here is on getting stuff achieved and never getting slowed down.

It’s not going to win awards for finest design ever, however I’d name that servicable

This half is exclusive to my concept, and it wasn’t chosen on accident. I do know I already personal this instrument — DocuPanda.io, so I ought to be making an attempt to make use of it. DocuPanda finds the data you want in a doc persistently, and it will probably additionally spotlight the web page and part the place that info was discovered.

Constructing a Schema

In DocuPanda you possibly can outline a schema that claims what you wish to extract from paperwork. Most individuals use it for stuff like extract lease from a lease, particular clauses from a contract, line objects from an bill or utility invoice, and so on.

However it will probably assist some fairly area of interest use circumstances, that embrace fiinding out particular contract clauses, answering advanced scientific questions given an instructional paper — and likewise, I believe it ought to be able to find all occurrences of Personally Identifiable Data (PII) in a doc.

So how can we proceed?

  1. Googled round to search out what widespread types of PII there are. Pasted a docs web page from AWS, which clear up a similiar downside solely on plain textual content and never PDFs. Then I requested GPT to broaden that record and make it extra generic.
  2. Uploaded some paperwork that include my PII into docupanda, and within the schema creation I wrote “You could discover personally identifiable info, right here’s an inventory of sorts”. Paste that record, and hit run.
Making a schema to search out PII

I made a video tutorial of your entire course of, so you possibly can comply with alongside should you’re curious.

Getting areas and redacting them

Getting the bounding containers from DocuPanda is straightforward. You simply make a name to its API. DocuPanda can localize the web page and part that include any info it finds.

Right here’s what that appears like on DocuPanda’s web site. On the left you possibly can see it extracts precise account quantity values, names, and so on. On the proper you possibly can see how the highlighted worth maps to a selected part within the doc.

You possibly can most likely already discover a limitation. Proper now the localization is finished at whole part ranges, so we’re not highlighting particular phrases. We will enhance on that later, proper now we’re making an attempt to get to the end line and launch.

So all we have to do is write code that uploads a doc to DocuPanda, standardizes with our PII detection schema that we simply made, and pulls down the rectangles. After which we have to apply the redaction on to the PDF, making the rectangles into black ink on high of the PDF. I made one other video tutorial that reveals how we try this. That is fairly technical, and once more boils right down to the identical level I’m making an attempt to make, which is you may get loads of mileage with not realizing a lot and prompting GPT O1 like a monkey on acid. You possibly can watch this course of beneath.

Finally, you wind up with quite simple code that’s nonetheless a bit too lengthy to cite in a weblog submit, so you possibly can take a look at the gist right here.

Check out the ultimate end result: pdf-redact.com.

In case you have a look at Docupanda’s web site, the plagarism is clear. All I’ve achieved to transform from my template to this web site is to take my current web page, clarify to GPT what Pdf-Redact is about, paste within the touchdown web page content material, and ask it to present me an up to date web page with language that displays what up to date language to replicate what Pdf-Redact does.

I grabbed the pictures from undraw. If I had extra time I’d immediate Flux to present me some extra related footage.

To make certain, the result’s an internet site that appears generic and it’s removed from polished. It received me to the end line, and it’s at all times higher to launch first, enhance later.

To check out if web site is helpful, I’ve used Google Advertisements capability to generate a marketing campaign robotically primarily based in your web site content material. I eliminated a few irrelevant key phrases they got here up with, however in any other case their first guess was typically wonderful, focusing on key phrases like “Redact PDF” or “take away PII from PDF”.

Nice, lets give Google $30 a day for a bit and see if that is ok to transform anybody in any respect. I’m not anticipating this adverts marketing campaign to be worthwhile on this preliminary kind. I do count on to get a primary sign if that is helpful in any respect.

To this point I’ve made a whooping $19. Give me an opportunity although, it’s solely been a day since my google adverts marketing campaign went reside.

I’ll maintain this service operating, and who is aware of — perhaps it finally ends up rising. It’s on our growth roadmap to let DocuPanda localize particular person phrases as an alternative of whole sections, and at that time my product will change into fairly highly effective.

I hope you’re studying this and really feel like constructing one thing of your personal. Please attempt. DocuPanda is my years-long journey, which is what you hope a SaaS will develop into after these first 24 hours.

If your online business may additionally use DocuPanda’s capability to search out info persistently from PDF paperwork and scans, I hope you contemplate making an attempt it out.

Completely happy constructing!

Related Articles

LEAVE A REPLY

Please enter your comment!
Please enter your name here

Latest Articles