The real cost of building software in Algeria

When an Algerian software shop quotes you "two million dinars for a mobile app", they are quoting one of three things and not telling you which one. Sometimes it is the cost of writing the code. Sometimes it is the cost of writing the code plus a thin design phase. Sometimes it is a deposit on a much larger project where the rest of the cost lives in change requests. The three are not the same project, and the three end at very different total costs.

A quote without a written scope is a marketing brochure, not a contract. Anyone can promise anything when nothing is written down. The first signal of a serious vendor is that they refuse to send a final number until the scope document is signed.

For a real, production-grade mobile app — iOS and Android, real backend, real authentication, real launch — count between four and twelve million dinars in Algeria today. Below four, you are buying a prototype that will need to be rebuilt. Above twelve, you are paying for either complexity (fintech, healthcare) or for an agency margin that does not reflect the engineering work.

For a custom web platform — multi-tenant, real database, admin dashboards, the things actual companies need — count between three and fifteen million dinars depending on whether it has to integrate with anything legacy. The integration is almost always more expensive than the application itself.

For a custom ERP, see our companion brief — but the short version is twenty-five to one hundred and twenty million dinars for an Algerian SME. For an applied AI engagement, between five and forty million dinars depending on data quality and integration depth. The model is always the cheapest part.

Three line items kill more projects than the headline price ever does. The first is integration — every connection to a bank, a customs system, a logistics provider, a payroll engine, or a legacy ministry portal is a project of its own. Most quotes pretend this work does not exist.

The second is data migration. Moving five years of Excel data into a new system is not a small task. It is detective work, cleanup, manual review, and emotional negotiation with the people who own the spreadsheets. Budget at minimum 15% of the build cost for migration. Most quotes budget zero.

The third is the operational handover — training, documentation, the first three months of support after go-live. This is the difference between a project that runs and a project that gets quietly abandoned. Budget another 10% to 15%, even if the vendor tells you it is included.

Three questions tell you whether a quote is honest. One: does it list assumptions explicitly? A quote with five lines of assumptions ("we assume the bank will provide a sandbox environment", "we assume legacy data is exported in CSV") is a vendor who has thought about the project. A quote without assumptions has not.

Two: does it separate fixed-scope work from time-and-materials work? Software that is well understood can be quoted as a fixed price. Anything involving integration, data, or an unknown third party should be time-and-materials. A vendor who fixed-prices the unknown is hoping you will not notice when they invoice you for change requests later.

Three: does it name the engineers? "A senior team will deliver the project" is not a commitment. The names of the people who will actually write the code, attached to a CV, is a commitment. The first is what failed projects look like before they fail.

Our quotes are usually 20% to 40% higher than the competition on the headline number. They are also usually 30% to 50% lower on the total cost of the engagement, because we price the integration, the migration, and the handover into the original quote rather than discovering them as change requests later.

It looks worse on day one and feels better on day two hundred. That is the trade we have made deliberately. We would rather lose a deal in the first meeting than win one that ends in adversarial billing six months later.