Ask HN: What is the best tech stack to develop a SAAS Service? Hi Guys, I want to know what languages and frameworks should i learn in order to develop a vertical SAAS Service. This includes making an iOS, Android and a Web App. Feature that i want to develop are - - Inventory Management - Dealer Management - Accounting - Payroll - Analytics (Important) |
What language are you good at OP?
You can quickly mock up majority of your use case using PHP, ruby or even Node.
Create a monolithic app and stay away from microservices until you have clearly drawn service boundaries. I'd say (6 to 12 months) in production. You dont need the headache of the distributed overhead.
Just keep in mind that you should "write code that can easily be disposded or replaced, not extensible"
For generating client sdk's i suggest using swagger. Grpc is there but i have trouble grasping protobuf sharing at the moment
For user clients, i would go for react native.
You're better off using a 3rd party for analytics when you are starting up: if you have the money, I'd suggest using segment, new relic or even google. Else, do it with prometheus or kibana.
My 2 cents
Ruby on Rails provides huge bang for the buck with their integrated database ORM, as does Phoenix (what I used). Your scope involves a lot of referential data, stick to traditional robust SQL database system (Postgres for example). If you use a NoSQL db, you may be quicker out the gate (as it'll take any JSON you give it), but once you get to the analytics with lots of joins, it won't be as much fun (personal experience). Also, I would suggest using the server frameworks' rendering engine to get your MVP up and running, as you can then fast iterate between data model changes and visual representation (i.e. HTML output), and avoid front-end builds and all that encompasses. Phoenix is very snappy, and with something like Turbolinks can potentially be as snappy as a dedicated front-end app.
Personally, I needed very interactive front-end, so had to build a separate front-end using Ember. I found Ember with Ember Data provides a robust toolset that manages complex data. I also have experience with React and Vue, but keep in mind with their state management (e.g. Redux / Vuex) that you'll need to manage complex data relationships yourself. As far as mobile, don't think native, as you'll need to learn specific technologies to implement that as well (more time...). Grab a bootstrap template that is already responsive, and you can run the same web-site on an iPad (forget the phone initially). Good luck!
source: just finished a 2 year build of an investment-related accounting and analytics webapp
For the mobile apps - Xamarin. You will still write C#, just use a different way of thinking.
Having the same technology stack in web and mobile will allow you to reuse skills one day when you build a team.
Then i want to process and display all this data (related to Manufacturing, Sales, Marketing etc) on web, iOS and Android.
So essentially there will be apps for employee's working in the plant (for data entry) as well as for the managers and owners (to monitor everything and make decisions)
I want the web dashboard to look somewhat similar to - https://demo.baremetrics.com
And since i have no experience in building mobile apps i want to know what languages/frameworks would be best suitable to develop good/native looking apps for iOS and Android using web technologies.
Thanks.
Don't worry about keeping the vertical secret for competitive reasons. You either have the domain expertise to make it great or you don't. And anyone with the domain knowledge knows what they want, but likely don't have the tech skills to make it happen. If you have both, start with the tools you know and get going.
I went the web/bootstrap path for most of it. I'll do mobile apps this year.
Get customers first. I had customers before the project had a name LR a signup form or much of anything.
Got fast.
I'm quite confident a number of highly knowledgeable people have looked at your question but have been unable to answer for lack of context. Narrowing things down a bit would likely get you more focused answers.
With that said, the best tech stack is the one you already have in your pocket. You could argue that this is why having a good all-around knowledge of all of the major stacks is a good idea.
You learn what is evolving, and you grab onto it, and evolve right with it. This is easy territory for flame wars and "whatever fits you"/"there is no best stack", but that's not the full picture.
Your survival depends on what you know, life is full of surprises, death in family, cancers, car accidents, addiction, breakups, and depression.
Imagine a warrior, ask him, "what is the best weapon?" and can you really imagine him saying "There is no best weapon", no, s/he would never say that.
You go on the most popular code hosting website, you sort by number of stars and begin at the top.
To give you a specific answer "iOS, Android and a Web App" in context of "Inventory Management - Dealer Management - Accounting - Payroll - Analytics" is Apache Cordova, (iOS, Android, and others), you use npm/node this way you will have one code base, you won't have to write the same code in multiple languages (just imagine scaling a live data sync requirement in one app and then standard REST in another -- keep it all in one language, or you will fail.)
"Inventory Management - Dealer Management - Accounting - Payroll" is all pretty simple (but, please explore ideas like http://vorpal.js.org/ just to develop your API before getting to G-UI, finish your programs in the terminal first.
Don't pretend programs are written GUI first, they are not. GUI is a leaky abstraction to say the least.) Don't fall for redux, relax and use https://github.com/mobxjs/mobx for actual GUI, (by the time you get to this part https://vuejs.org/ will rule)
And for your Analytics front-end D3, it is a departure from standard JavaScript but it has a lot of love inside.
Finally, remember develop commands/API first, get your apps working in the CLI and you will finish your program before starting it. Think about using the power of Unix (https://www.npmjs.com/package/commander), for the love of all that is mighty think about IRC, NOBODY! has time to learn to use your apps/GUI and NOBODY! ever needed to learn how to use Siri, so create a Command Line Interface version of your program and before you jump to React/Vue meditate over https://hubot.github.com/ and programs like Siri.
Finally, if I was facing your requirements, I would build a wiki/forum where your users would interact with bots. Over time I'd rip put some of the robotic guts and boldly go where humanity needs to be headed, I'd replace them with what I call Artificial-Artificial Intelligence, networks of Human Mechanical Turks where anybody could participate.
Neural networks are easy to train, but they are also easy to trick. Nothing will beat a random sample of Humans from across the world reaching a consensus on a tough decision.
In closing, People, Real Hearts, Human Minds, powering your network would share in your profits, and you'd do your part in battling what ails us all, World's Poverty.
If a mother in Rwanda and a Boy in India participating in your Mechanical Turk can find their way out of poverty, by helping your to detect fraud in your circuit. Then, you did your part, to help Humanity grow.
In almost all cases, the tech stack is essentially meaningless. This runs counter to what we frequently want to believe as engineers because there certainly are meaningful technical differences between languages and frameworks. These differences just don't generally translate to meaningful business results.
For example, I am aware of a SaaS company with fewer than five employees total that generates over $3M in revenue annually. The profit margins are mouth watering - total non-payroll expenses are high five figures/low six figures each year. Everything is developed with PHP and MySQL - it would be vilified here, but it's more successful than virtually all the SaaS companies mentioned on HN.
I cannot tell you how many YC companies I have worked with that use Python, Ruby, JavaScript, Go, etc. At this point when I'm on a scoping call with a founder, I can roughly map out when the company was founded +/- a year simply by listening to their technology choices. I have not seen any meaningful difference in whether or not the company ultimately went on to great success based on the tech stack. Instead, I have seen companies go on to great success based on proper identification of real pain points, product-market fit, marketing and ability to execute well while wearing many hats. (But do keep in mind that this is anecdata and subject to usual biases of statistical rigor, or lack thereof).
My advice to you is this: develop your software in whatever language allows you to think more about business decisions and less about technical decisions. Your aim should be to build fast, iterate fast and fail fast. Many engineers who start companies struggle with this because technical decisions are their comfort zone, but your company will succeed or fail based on how well you identify market opportunities and figure out what your users want. If you just happen to be most comfortable in an "academic" language or a really trendy one, go ahead and use it. However, if you're most comfortable in Python, but you fetishize functional programming or React + Node.js because of the Medium posts that pop up on HN, go with Python.
Practically speaking for you specifically - I would first revisit whether or not you actually need a web app and two mobile apps. Could you get away with a minimal web application and choosing only one mobile application to start? Maybe develop in iOS or Android first, observe how your potential customers use the app, address grievances and product a mobile app for the other platform in the next iteration.
As for the actual technical choices - if you know several ways of developing iOS/Android/web applications, go with what is most familiar to you unless you have clear and incontrovertible evidence that one technical choice is clearly superior. If you have to learn a new tech stack to develop one of these, then 1) don't focus on it now, focus on market validation and 2) choose tech stack with the best combination of stability and ease of use to get up to speed with that platform.
Good luck!
So what i am planning to build is SAAS service that provides all the feature's listed above and more, customised for an (unspecified for competitive purpose) industry. The web client is used for entry of data related to sales, stock, payroll. I am also making apps for various operator's inside the plant to upload data using their android phone's and tablets.
Then i want to process and display all this data (related to Manufacturing, Sales, Marketing etc) on web, iOS and Android.
So essentially there will be apps for employee's working in the plant(for data entry) as well as for the managers and owners(to monitor everything and make decisions)
I want the web dashboard to look somewhat similar to - https://demo.baremetrics.com
And since i have no experience in building mobile apps i want to know what languages/frameworks would be best suitable to develop good/native looking apps for iOS and Android using web technologies.
Thanks.
On the surface you're describing something that's been done many many times (kinda obvious). On the one hand this means you could find someone to work with to help you on this fairly easily, on the other hand you're looking at 5 figures easily... which is likely why you're trying to figure this out yourself, now I think about it :)
I'm hesitant to suggest looking at existing off-the-shelf solutions; I get the impression that finding systems that haven't rusted together in the areas you need flexibility is touch-and-go and hard to figure out in advance. It may go incredibly well, or you may end up maintaining a massive headache that's incredibly hard to fit-out for your needs. One possible idea that may work to some extent is finding open-source projects that implement some or all of the functionality you want; they may be useful for parts/ideas or real-world answers to "how did they solve this bit," or even better you may find they actually work for you (yay free stuff!).
While you may come up with extra defining/deciding context/requirements, the main decision point I'm thinking of is whether you consider it a must (or a huge-value add) for things to be partially functional if there's no consistent network connection. This applies to sales/management (on a train, or at exactly the right vortex point between a bunch of buildings, and cellular just dropped out?) as well as plant operators (hypothetical contrived example: a non-essential machine just went bezerk and is spouting so much RF the Wi-Fi just died). That said, generally supporting transient network loss is... hard; you need a database design that, at every point in read/write, can tolerate "another client may have committed an update but that data might not be in the query we just did yet," which the application design must be built with awareness of (it's so fundamental it can't be abstracted).
If "the plant Wi-Fi just died" disrupting operations is okay (for example, you have somone on-site who can rapidly diagnose a wide range of issues, you have spare APs and related kit available, etc), then you'll save yourself a lot of time by just developing this as a webapp. Others may argue for native apps, but if you do a webapp, you can target desktop (Win/macOS/Linux(?)) + mobile (Android/iOS) just by using media queries. Suddenly several thousand lines of Swift and Java disappear :)
With enterprise setups like this, you also get to call the shots on what you contractually agree to support Android/iOS-wise; you get to specify the minimum versions, and you can also check the user agent and throw warnings or blocking error messages for versions you don't like. This isn't a public Web site, so you don't have to work for everybody. And so you get to guarantee that certain Web APIs are available :)
In all fairness there is the AppCache and ServiceWorkers+IndexedDB can cache web apps for offline use, but I wouldn't rely on them because the storage areas they use can be cleared.
Now I think about it, you could build minimal Android/iOS shells that load your HTML/CSS/JS into a WebView with some simple caching tacked on to guarantee code availability in offline situations. That could work quite nicely, and the native codebase would be tiny. With this Wi-Fi dying isn't a problem but now you need to go back to mentally modelling eventually-consistent database schemas.
On the backend side, what sort of traffic volume do you intend to handle? Do you intend to have the backend be on-site? (You could do a managed on-site setup where you have (monitored) SSH by contract but the system runs on client servers.) This would distribute processing load across the sites. If you want to centralize everything then you need to think about how many clients you want - then double or triple that figure (and then maybe multiply it by itself) - and you should have a good idea of the sort of volume you need to be able to process. There's a reason all the "how we process 1.21 jigabytes of messages every second" posts on here are so popular! You may need to look into high-throughput, easily-shardable datastores.
Regarding those graphs... well... note that you're looking at the revenue the guy who built those graphs is getting! :D I can see there's a fair bit of intelligence in the graph displays, and fine-tuning a similar system with a quarter of the functionality is something that would take a bit. Certainly possible, but it would take a bit.
Unfortunately I have very little advice to offer in the area of actual Web development; I haven't explored the scene much yet. I know that there are a dizzying number of developer offerings available. Your best bet is to forge ahead trying them and see which one you like.
I'm not a developer myself (in that I don't have a software engineering position - nor the mindset for one, just yet); I just spend all my time on HN :)
I wonder how much more specific you could be (for the benefit of everyone) without giving anything away.
By the way, thanks for mentioning the Baremetrics demo - that's really cool that they're open like that.
One more point: It's important to understand what parts of your app are capitalizable, and which parts are aren't. The really capitalizable parts are the ones you think bring real competitive value to the market you want to attack - the code that addresses your valuable SME and original ideas. So something to ponder: if you want to also develop a presentation layer, how valuable is it really to develop that layer internally when in the long run that work is not much of an asset, and is a time hole for development? No matter how skilled your team is with the latest js framework I can guarantee you that there's some team of kids out there who are much better with the same framework and can develop your UI at half the price. So create your mockups, define your inputs and outputs, and farm that shit out. Unload the UI headache while in startup mode.
You are totally right in that i don't have any real web/application development. In fact i just started university but i have been programming for a few years (mostly problems on sites like code chef in C, C++ no products) but recently got this idea when i visited a plant that my father owns. There are about 600 such plants in the same area and they use fairly old software or not at all. So i talked to him about my idea and he thinks it will be very useful and also commercially successful since there is virtually no competition. That's why i decided to learn web development. But now i am realising it's tougher than it looks. I asked this question in order to get an idea what all the things i need to learn in order to even attempt to build something like this. Now that i look at my question i think i should have asked "How to be a full stack web developer?". So if anyone is still reading i think this would be the question that you should answer or maybe i should have just googled it and started working
The path I would recommend to learn full stack development.
Go through the books Head First HTML and CSS (get an inexpensive hosting account you can FTP files to and setup some domains, point to your hosting), next go through Head First PHP and mySQL. Next is Headfirst Javascript then Headfirst jQuery.
After those you should sign up for teamtreehouse.com.
You could play around with developing this in PHP without a framework, setting up a login in system you would learn a lot. Sometimes frameworks are tough to understand what all the magic is happening behind the scenes if you haven't creating a simple app with login, etc in vanilla PHP on your own. For your customer version though you will want to use a framework, more secure, better organized code, etc.
I would recommend developing this with Laravel(PHP) it has a great community so you'll be able to find great learning resources to get started. (Rails is good too, but I think Laravel will be easier to get started/deploy your app)
Also I recommend developing on a mac. Windows always seems to have obscure issues that you will waste hours/days on.
laracasts.com will get you started with Laravel.
There are great deployment tools forge and envoyer. Hosting on Digital Ocean is perfect for most apps, AWS and linode too)
Use stripe for billing.
There is a SaaS add-on to the framework call Spark. It's ok but, I find I'd rather roll my own stripe/SaaS integration.
Also Spark does use Vue.js though so since you're just learning I would stick with developing the app using jquery for ajax interactions and think about moving up to vue in the future.
I would create the MVP as a web application for all the interactions. Then once you have that established you can look at adding true android and iOS apps. Those might be better to hire out the development of. Or a future learning project.
You should be able to set everything up as a mobile web application, use the bootstrap framework for the look and feel of your app. There are some ready made admin templates that can get you something similar to baremetrics.
Having your Dad's plant as an initial user for your app will be great to refine what they need get their feedback.
Good luck. Follow up and let us know how it's going.
C and C++ aren't the easiest languages to learn given they require manual memory management, so you have it fairly easy learning something else, unless you actually want to use C++ for the backend (I would hesitate a lot to recommend C). The main advantage of using something like PHP/Python/Ruby/Node vs C++ is the nonexistent compile times, along with the fact that they're all much higher level and let you focus on your app's requirements. That said, no language is perfect; it's about picking the one you can live with the easiest. Another potential consideration is to choose a language that a lot of programmers in your area are familiar with, so that you have a competitive market to take advantage of if you need to bring in extra help.
I'd say that it would be significantly easier to not spin this a general-purpose SaaS stack, but instead build this as a bespoke system specific to these plants' requirements. It will make developing the system significantly easier, especially given that this would be your first major project. Practically every enterprise/environment is full of custom applications; these are not unique at all.
(That said, from an educational standpoint, thinking about how you might generalise what you build would be a good idea, and serve you in building similar systems later on.)
Assuming you haven't already done this, what you definitely need is information about exactly what everybody needs to be able to do. You have an absolutely excellent opening with the fact that your Dad owns one of these plants: you should be able to easily get access to wander around your Dad's factory and ask questions of everyone when they have moments of free time. I would strongly recommend having long conversations with the workers themselves and completely hearing out how they would solve the issues they have on the floor. Potentially be prepared for the application to suddenly become very different than what you planned :) - you may end up electronically enabling/adding brand-new functionality to these plants' capacity/capabilities alongside the analytics and number crunching features you initially started out with.
It might be a good idea to do several rounds of questions, as you can discuss things in greater depth as you internalize what you learn and come up with unknowns as you work out how to turn problems into helpful code. You may or may not find asking questions helpful at the point you're at right now. Maybe an idea to discuss with your Dad.
As for an immediate long-term goal to figure out how to reach, I'd suggest taking the architectural requirements you have thus far and going ahead and figuring out how to put together a system that provides those things. This might be an existing system you buy (although I would strongly recommend grabbing some freelance consultant help for an independent go/no-go if you think about this route) or an open-source software stack that you can start with, or a stack you build yourself from scratch.
You may need to go through several iterations of designing, spending long days and nights on building, and then realizing you have to throw away half the system and start again. This is not a bad thing; in fact right now I'm building a crawler for a specific website that I'll use once and throw away. It's gotten to a couple thousand lines of code right now (it's quite simple, but PHP is quite a verbose language for async network stuff, and I'm unfortunately yet to learn a better language like Go or Python), and I'm basically putting it together for this once-off use-case. I've mentally saved a LOT of valuable information about Web crawler design that will go into my next major project, a more general-purpose crawler (note the things about generalization I mentioned earlier). I couldn't get this info from reading tutorials or even academic papers or theory.
Oh, another thing - databases are weird and annoying, but I would highly recommend you learn them in depth. There are lots of choices; MySQL/PostgreSQL are the de-facto relational options (and for MySQL there's phpMyAdmin, which is awesome for mental modeling and simple setup); Cassandra scales out to a staggering level; SQLite3 is great for learning on and prototyping with; MongoDB comes with utterly insecure defaults and doesn't scale well but does kinda work (not dissing it, just being honest). There are many others. Query plan optimization is something that should be absolutely logical but is full of obscure edge cases, and writing the same query with different syntax sometimes gives inexplicable speed boosts. Lots of confusing fun - my point is figure it out to the point where you're comfortable with databases and can easily reason through them.
You probably discovered http://stackoverflow.com/ long ago; if you don't already have an account with a few hundred reputation I would recommend going on there and answering a few questions so your account has enough reputation that you can do things like use chat and such. Some of the other Stack Exchange sites like http://codereview.stackexchange.com/ may also be of help.
Good luck, and I kinda want to hear how this goes :)