What tech stack should a startup use in 2027?
By Bonaventure Ogeto, Founder and CEO, Mctaba Labs
Published
Use a common, well-documented stack that many developers near you already know. A sensible default going into 2027 is TypeScript with Next.js for the web app, a Postgres database such as Supabase, managed hosting, and M-Pesa through Daraja or Paystack. Approve something unusual only when your developer can give a specific reason your product needs it.
Your developer sends a proposal. Under "Technology" there is a line of names: a language, two frameworks, a database, a hosting company. You recognise none of them. You are being asked to approve a decision you cannot evaluate, about a product you are paying for.
You do not need to learn to code to handle this well. You need four tests, a sense of what "normal" looks like going into 2027, and a short list of questions for the day someone proposes something that is not normal.
Why should a non-technical founder care about the stack?
Because the stack decides three things you will live with long after this developer has moved on: who can work on the product, what it costs to run each month, and how hard it would be to change course.
A tech stack is the set of tools a product is built from. It has a handful of parts:
- The front end. What the customer sees and taps.
- The back end. The rules of the business: who can do what, what happens when a payment arrives.
- The database. Where customer records, orders and payments are stored.
- Hosting. The computers the product runs on.
- Integrations. Services you connect to, such as M-Pesa, SMS and WhatsApp.
Customers will never know which tools you chose. The next developer you hire will know within an hour, and their reaction will tell you whether you chose well.
What four things matter when choosing?
Put any proposal through these four tests.
1. Can you hire for it in Nairobi? This is the test that matters most. Developers leave, get ill and take other jobs. If your product is written in something only a few people in the country know, your choice of replacement is narrow and their price is high. Ask any two developers you know, separately, whether they could pick up the proposed stack. If both hesitate, think again.
2. What does it cost to run when you are small? A new product has little traffic. It should cost little to host, and the bill should grow with use. Many hosting and database services have a free or low starting tier that is enough for a pilot. Be wary of designs that need several servers running all day before the first customer has signed up. Ask for the expected monthly cost at ten customers and at a thousand, in writing.
3. Does it work with what Kenyan customers use? Your product will need to take M-Pesa, send an SMS and probably send a WhatsApp message. Safaricom's Daraja API, payment providers such as Paystack and Flutterwave, Africa's Talking for SMS and USSD, and the WhatsApp Business Platform all work over standard web requests, so any mainstream language can use them. The difference is in how many working examples and libraries exist, and how many local developers have done it before. A stack where your developer has already shipped an M-Pesa integration is worth more than a better stack where they have not.
4. Could you leave? Every service you depend on is a supplier. Ask what happens if a supplier raises prices or closes. Good signs: your data sits in a standard database you can export, the code is in a repository you own, and the main tools are open source. Bad signs: business logic written inside a proprietary tool that has no export, or a platform whose format nothing else can read.
What is a sensible default stack going into 2027?
Here is a common, well-supported set of choices. It is not the only good answer. It is the answer that needs no special justification.
| Layer | A sensible default | Why |
|---|---|---|
| Language | TypeScript | One language for the front end and the back end, widely taught and widely known |
| Web framework | Next.js, built on React | Handles pages and server logic in one project, with a very large community and thorough documentation |
| Database | PostgreSQL, often through Supabase | A standard, open-source relational database. Supabase adds sign-in, file storage and an interface on top of it |
| Hosting | A managed platform such as Vercel for the app, with the database run by its provider | No servers to look after, and cost that tracks usage |
| Payments | M-Pesa through Daraja directly, or through Paystack or Flutterwave | Daraja gives you a direct M-Pesa connection. A provider adds cards and other methods through one integration |
| Messaging | Africa's Talking for SMS and USSD, the WhatsApp Business Platform for WhatsApp | Established services with local coverage |
| Mobile app, if needed | Flutter or React Native | One codebase for Android and iPhone |
| Code repository | GitHub or GitLab, in the company's account | Standard, and every developer knows them |
Two notes on this table.
First, the relational database is the choice to hold firmest. Businesses that take payments have customers, orders and transactions that relate to one another, and that is exactly what a database like Postgres is built for. Because Postgres is open source and offered by many hosting companies, you can move it between providers. Supabase's own documentation describes its cloud product as compatible with a version you can host yourself, which is the kind of exit you want.
Second, equally good defaults exist. Laravel with PHP and Django with Python are mature and well documented, and you will find developers in Nairobi who know them. Check with the two-developer test above. If your developer is strong in one of these, that is a better choice than a stack from the table that they would be learning as they go.
When is an unusual stack the right call?
When the product has a specific need that the default handles badly, and the developer can name it.
Real examples of good reasons:
- The product does heavy data analysis or machine learning, so the back end is in Python, where the tools for that work are strongest.
- The product must work offline on cheap Android phones all day, so the team builds a native app first.
- The team has five years of experience in one stack and a library of code they have already tested in production.
- The product must process a very high volume of events in real time, and the team has done this before.
Weak reasons sound different. "It is the future." "It is faster", with no explanation of what is slow today. "Big technology companies use it." Those companies have thousands of engineers and problems you do not have. "I want to try it" is honest, but your budget is not a training fund.
What should I ask a developer who proposes something unusual?
Ask these eight questions and write down the answers. A good developer will enjoy them. A poor one will become irritated, which is also an answer.
- What does our product need that the common stack cannot do?
- What have you built and launched with this stack before? Can I use it?
- If you were unavailable tomorrow, how many developers in Nairobi could take over, and how would I find them?
- Have you integrated M-Pesa in this stack? Can you show me a test payment going through?
- What will hosting cost each month at ten customers and at a thousand?
- Where does our data live, and how would we export all of it?
- Which parts depend on a single company's service, and what is the plan if that service changes its terms?
- What would you choose if you had to hand this project to a junior developer in six months?
Question eight is the most revealing. Developers often pick differently when they imagine someone else maintaining the work.
What do I need to own, whatever the stack?
The accounts and the keys. This matters more than which tools are inside them.
- The code repository belongs to the company, and you are an owner of it.
- The hosting and database accounts are opened with a company email address.
- The domain name is registered to the company, with renewal on a card you control.
- The Daraja, Paystack or Flutterwave account is in the business's name, because that is where the money lands.
- App store developer accounts, if any, belong to the company.
- Passwords and API keys are stored in a password manager you can open.
- A short document in the repository explains how to run the product and how to deploy it.
With these in place, a mediocre stack can be fixed. Without them, a perfect stack can be lost in an afternoon.
Does the stack decide whether the startup succeeds?
Rarely. Products fail because nobody wanted them, because the money ran out, or because the team fell apart. They almost never fail because of the choice between two mainstream frameworks.
That should make the decision lighter. Pick something common, check that your developer has shipped with it before, own the accounts, and keep your data in a standard database. Then put your attention back on customers. If the product works, you will have revenue to pay for a rewrite of whatever turns out to be wrong. If it does not, no stack would have saved it.
Key takeaways
- The best stack is the one you can hire for in Nairobi next year, not the newest one.
- Judge any proposal on four tests. Hiring, running cost, local integrations and how easily you could leave.
- A common default is TypeScript, Next.js, Postgres and managed hosting, with Flutter or React Native if you need a mobile app.
- Ask to see M-Pesa and messaging working in the proposed stack before you approve it.
- Own every account and make sure your data can be exported in a standard format.
- Stack choice rarely kills a startup. Building the wrong product does.
Frequently asked questions
- Does the tech stack matter to customers?
- Customers never see it. They see whether the product is fast on their phone, whether payment works first time and whether it is there when they need it. Any mainstream stack can deliver all three. The stack matters to you because it decides who you can hire and what it costs to change direction.
- Is it a problem if my developer wants to use PHP, Python or another established language?
- No. Laravel with PHP and Django with Python are mature and well documented, and you will find developers in Nairobi who know them. An established stack your developer knows deeply will usually beat a fashionable one they are learning on your budget. Apply the same four tests and you will often approve it.
- Should I use no-code tools instead of a coded stack?
- For testing demand, often yes. A form, a spreadsheet, a payment link and WhatsApp can run a pilot for weeks. Move to code when you hit a limit that costs you customers, such as a payment flow the tool cannot support. Check first how you would get your data out of the tool.
- Do I need microservices, Kubernetes or a blockchain?
- Almost certainly not for a first version. These solve problems of scale or trust that a new product does not have yet, and each adds running cost and specialist skills. A single application with one database is easier to build, cheaper to host and simpler for the next developer to understand.
- Who should own the hosting and database accounts?
- The company. Open each account with a company email address, add your own payment card, and invite the developer as a member. If the product runs inside a developer's personal account, you depend on that person's goodwill for access to your own business.
Sources
- Next.js Docs, Next.js documentation (Vercel). Checked on 19 November 2026.
- Architecture, Supabase Docs. Checked on 19 November 2026.
- Payment Channels, Paystack Developer Documentation. Checked on 19 November 2026.
- Safaricom Developers Portal (Daraja), APIs, Safaricom. Checked on 19 November 2026.
