SaaS Development
A product you can charge for, with the per customer cost known in advance.
What is SaaS development?
SaaS development is building software sold as a subscription, which requires customer data kept properly separate, billing that works without manual intervention, and a cost per customer low enough for the price to make sense.
Building the product is the part founders plan for. The parts that decide whether the business works are the ones underneath: keeping one customer data away from another, billing that runs itself, and knowing what serving a customer actually costs.
AI features make that last point sharper. Traditional software costs almost nothing per extra user; AI features cost real money per request, so a flat monthly price against unlimited AI usage can lose money on your best customers.
Working out the unit economics before launch is unglamorous and it is what separates a product from an expensive hobby.
Why we work out cost per customer first
Duara builds subscription products from Nairobi for founders and for established businesses productising something they already do well internally.
We work out cost per customer before build, particularly where AI features are involved. A flat price with unlimited AI usage is a pricing model that punishes you for having heavy users, and founders discover that after signing the customers they most wanted.
Who this is for
- You are building a subscription product
- AI features make your unit costs unpredictable
- You are productising something you already do internally
- An existing product cannot handle multiple customers safely
Industries we serve
- Software startups
- Fintech
- Professional services productising a method
- Logistics technology
- Education technology
- Health technology
Problems this solves
These are the structural questions that decide whether a subscription product becomes a business.
- Unknown cost per customer. Nobody has calculated what serving one customer costs, so the price is a guess and margin is unknown.
- AI usage unbounded by price. Flat pricing against per request AI costs means the heaviest users are the least profitable, sometimes severely.
- Tenant separation as an afterthought. One customer seeing another data is the failure that ends a product, and it is difficult to retrofit.
- Billing that needs a person. Subscriptions, upgrades and failed payments are handled manually, which does not survive growth.
What we deliver
A product built with proper tenant isolation, automated billing including failed payments, usage metering per customer, and pricing informed by the actual cost of serving one.
A subscription product with proper tenant isolation
Automated billing including failed payment handling
Usage metered per customer and per feature
Cost per customer calculated before launch
Onboarding that works without you present
A pricing model that survives heavy users
Features
- Unit economics before build. What serving one customer costs is calculated first, because pricing decided afterwards is a guess with a decimal point.
- Tenant isolation designed in. Separation is architectural rather than a filter somebody remembers to apply, because one leak ends a product.
- AI usage metered per customer. Model costs are attributed, so you can see which customers are unprofitable rather than discovering it in aggregate.
- Billing handles failure. Failed payments, retries and dunning are built, since that is where subscription revenue actually leaks.
- Self serve onboarding. A customer can start without you, which is the difference between a product and a service with a login.
Benefits
- You know your margin. Cost per customer turns pricing from a guess into a decision.
- Heavy users stop being a problem. Metering and appropriate pricing mean your best customers are not your least profitable ones.
- Growth does not need headcount. Automated billing and self serve onboarding are what make subscription revenue scale.
- Customer data stays separate. Architectural isolation prevents the single failure that no amount of good product recovers from.
Technologies we use
- Multi tenancy. Architectural separation of customer data with isolation enforced rather than applied by convention.
- Subscription billing. Plans, upgrades, proration and failed payment handling without manual intervention.
- Usage metering. Recording consumption per customer and feature, which AI costs make necessary rather than optional.
- Self serve onboarding. Signup, setup and first value without a person involved.
- Scaling and cost control. Infrastructure that grows with customers and spend caps that prevent a runaway feature.
AI features break the assumption that serving one more user is nearly free. Any pricing model designed before that was true needs rechecking against per request costs.
How the process works, step by step
A first SaaS release usually takes twelve to twenty weeks, and the pricing work happens before the build rather than before the launch.
-
Step 1
Free product and pricing review
We work through what you would charge and what serving a customer would cost, before designing anything.
-
Step 2
Unit economics established
Cost per customer is calculated including AI usage, since that is what decides whether the price works.
-
Step 3
Isolation designed in
Tenant separation is architectural from the start, because retrofitting it is expensive and risky.
-
Step 4
Narrow first release
A deliberately small product ships to early customers rather than a complete one shipping late.
-
Step 5
Billing and metering live
Automated billing and usage attribution go in before scale, not after the first accounting problem.
Quality assurance and how we work
The commitments attached to building a product you intend to sell.
- We calculate cost per customer before building. A flat price with unlimited AI usage punishes you for having heavy users, and founders discover that after signing the customers they wanted most.
- Tenant isolation is architectural. Separation enforced by design rather than by a filter somebody remembers, because one customer seeing another data ends a product.
- We will question the business model. Where the numbers do not work, we say so before the build rather than after launch.
- Spend caps on AI features. A runaway feature on a subscription product is a loss that scales with your success.
- You own the product entirely. Code, infrastructure and customer relationships are yours, with no dependency on us to operate it.
Frequently asked questions about SaaS development
What do founders most often get wrong?
Unit economics. Nobody calculates what serving one customer costs, so the price is a guess and the margin is unknown.
Why do AI features change the pricing question?
Because traditional software costs almost nothing per extra user, while AI costs real money per request. Flat pricing against unlimited usage loses money on heavy users.
Will you tell us if the model does not work?
Yes, before the build rather than after launch. Where the numbers do not work, saying so is more useful than taking the project.
What is tenant isolation?
Keeping each customer data architecturally separate, rather than filtering it in code somebody has to remember to apply.
Why does that need designing in?
Because retrofitting it is expensive and risky, and one customer seeing another data is the failure no good product recovers from.
How do you stop a runaway AI cost?
Spend caps per customer and in aggregate. A runaway feature on a subscription is a loss that scales with your success.
How long to a first release?
Twelve to twenty weeks, with a deliberately narrow product shipping to early customers rather than a complete one shipping late.
Should we charge per user or per usage?
It depends on your cost shape, which is why we calculate that first. Usage based pricing suits AI heavy products more often than founders expect.
Can it handle M-Pesa subscriptions?
Yes, and recurring mobile money collection has real quirks worth designing around rather than discovering.
What about international card payments?
Yes, and most Kenyan SaaS products need both, since your local and export customers pay differently.
Does billing handle failed payments?
Yes, including retries and dunning, because that is where subscription revenue quietly leaks rather than dramatically fails.
Can customers sign up without us?
That is the goal, and it is the difference between a product and a service with a login attached.
Could we run the product without you?
Yes. Code, infrastructure and customer relationships are yours, with no dependency on us to keep it operating.
Can you productise something we do internally?
Often, and it is a good starting point since the method is already proved. The work is generalising it beyond your own case.
How do we know which customers are unprofitable?
Usage metering per customer, which AI costs make necessary rather than optional.
What if we grow faster than expected?
Infrastructure scales, and spend caps keep the AI side from scaling faster than revenue.
Can we start with a single customer?
Yes, and it is often sensible, provided isolation is built in from the start rather than added when the second arrives.
Do you take equity instead of fees?
No. We are a development company rather than an investor, and mixing the two makes our advice less useful to you.
Where does customer data live?
In your own infrastructure, under your control, which also matters to your customers asking the same question.
What does a SaaS build cost?
Scope and integration count determine it, quoted after the product and pricing review.
Glossary of terms
Subscription product terms in plain language.
- Multi tenancy
- One system serving many customers with their data kept separate.
- Unit economics
- What it costs to serve one customer, against what that customer pays.
- Churn
- The rate at which subscribers cancel, which determines whether growth compounds.
- Dunning
- The process of recovering failed payments, where subscription revenue quietly leaks.
- Self serve
- A customer signing up and getting value without a person involved.
- Usage metering
- Recording what each customer consumes, made necessary by per request AI costs.
Related services in AI Development
Talk to us about SaaS development
Work out what one customer would cost you to serve for a month, including AI usage. If you cannot, that is where to start.
Start your project