What determines the budget of a project
What the price of custom software depends on
Custom software, customizable software, an automation, or an artificial intelligence project are quoted case by case. These are the factors that make one project cost more than another.
What the tool has to do
Functional scope
The effort of a development is driven by the logic behind each screen.
Concepts involved
Every concept the tool manages (products, batches, warehouses, suppliers, price lists, contracts) brings its own relationships with the rest, its validations, and its screens. Fifteen related concepts take considerably more work than two.
Edge cases
Exceptions account for much of the effort: the client billed under different terms, the order split into two shipments, the return that follows its own path. They are worth surfacing while defining requirements, before quoting.
Users, roles, and permissions
Granularity is what drives the cost. Deciding who gets into each section is straightforward. Deciding which specific field each role can see or edit, and in which state of each record, multiplies the checks across the whole application.
Concurrent access
When several people can edit the same record at once, the behavior has to be defined. In a warehouse or a seat booking system, a mismatch caused by concurrent access has consequences, and preventing it takes specific work.
Calculation algorithms
Route planning, shift assignment with constraints, or a cost breakdown with many variables are designed, tested, and tuned. They take an analysis stage before development starts.
Languages
Preparing the tool for several languages in the first phase costs less than adding it later. It affects interface texts, date and number formats, and the content managed per language.
How it is built and how it looks
Architecture and design
Two decisions taken at the start of the project that drive the cost of everything that follows.
Monolithic application or separate API and frontend
A monolithic application keeps the logic and the interface in one piece, the programming is lighter, and the result is perfectly valid. Separating the API from the frontend lets you scale each part on its own, connect other tools against the same API, change the look without touching the logic, and write automated tests. It is what we recommend for new projects. The full comparison is on the custom web applications page.
Level of design
The design can arrive defined by the company, with its brand guidelines and its approvals, which is common in large organizations. Gilsys can build it to measure, with specific requirements, and then it is design work in its own right. Or it can be a proposal of ours with no specific requirements, aimed at making the tool clear and comfortable to use: that is the most economical option, and the usual one for internal management tools.
Which systems it connects to and what data it handles
Integrations and data
Most projects live alongside data and software the company already uses.
Integrations with other software
Connecting to an ERP, a CRM, an online store, or a vendor's tool depends above all on whether that system has a documented API. When it does not, how it works has to be figured out before it can be integrated. That is the migration and systems integration work.
New project or new version
Rebuilding a tool that is already in use means understanding how it works today, deciding which behaviors to keep, and planning the switch from one to the other without interrupting the company.
Data migration
The cost of importing data from the previous system depends on the shape it is in. Duplicated or incomplete data, or data spread across files and spreadsheets, needs a prior review.
Volume and type of data
Images, documents, and data arriving continuously shape the database, the storage, and the infrastructure. A text record and a media file are neither stored nor queried the same way.
Automated steps
Automating a step with fixed rules has a bounded cost. Having an AI read a document or apply a criterion means preparing the model, testing it with real cases, and measuring how often it gets it right.
Private or commercial AI
When the data cannot leave the company, the model runs on private servers. That private AI means sizing and maintaining the server, and costs more than a commercial pay-per-use model. The decision depends on the data being processed.
What a failure costs
Criticality and fault tolerance
A platform used in critical scenarios takes more prevention, verification, and recovery work than an internal tool.
Consequences of an outage
A brochure page that stops loading has no direct cost. Payments in an online store, or goods-in records in a warehouse, lose money from the first minute of downtime, and the quote reflects that.
Internal use or product for third parties
In a SaaS used by outside clients, security also means isolation between accounts, access control per organization, and protecting each user's data from the rest.
Automatic response to failures
Having the system detect a problem, recover on its own, and lose no data along the way is an extra layer of work, needed in some projects and unnecessary in others.
Automatic alerts
Finding out about a failure as it happens, rather than when a user reports it, means defining which situations are watched, with which threshold, and who gets notified.
Monitoring and performance
Measuring how the tool responds, under how much load, and where it degrades allows you to act early. It takes its own tools and configuration.
Licenses, payments, and subscriptions
Charging inside the tool adds payment gateways, subscription states, renewals, failed payments, and invoicing. It is a development of its own within the project.
Where it runs and how it is released
Infrastructure and deployment
Type of infrastructure
A server rented from a hosting company, the infrastructure the company already runs, a container setup with Docker, or a cloud such as Amazon AWS have different setup and maintenance costs.
Number of platforms
Launching several interconnected platforms at once adds coordination, joint testing, and a release process beyond what each platform adds on its own.
Automated deployment
Automating each release avoids the errors of a manual update and allows improvements to ship more often. You can see how we do it under deployment and continuous integration.
External security audit
On every project we run our own security review. If the project also calls for an audit by a specialized firm, both that audit and the adjustments it recommends have to be accounted for.
Who takes part on each side
Team and points of contact
The budget also depends on how many people are needed and on how decisions get made.
Profiles involved
Some projects call for a mix of profiles and others can be handled by a single professional from the team. In both cases there is a project manager who oversees the work and is the person the company talks to.
Extra people for the schedule
A tight delivery date means adding more people with the same profile to work in parallel, even when the complexity of the project does not call for it.
Points of contact at the company
Gathering and reconciling requirements from production, administration, and sales takes more sessions and more sign-offs than working with a single person who decides.
Progress meetings
Some companies need to meet often to follow the project comfortably, even when the development itself does not require it. That is a legitimate need and it is taken into account when pricing the project.
On-site sessions
If the project will require on-site meetings, that is factored into the estimate from the start.
Speed of decisions
A project can sit idle waiting for a decision or a piece of data. The more the decision is spread out, the more it pays to agree up front who answers, and within what timeframe.
Advice during the project
When the company arrives with its requirements defined, the work is development. Helping decide what to build and in which order is consulting, and it is priced separately.
Documentation and training
Delivering the tool documented and training the people who will use it are line items of their own. For training, the cost comes down to the number of sessions and people.
How far the project is meant to go
Ambition and horizon
Minimum viable or finished from the start
Solving the essentials and improving them in later phases costs less up front. Aiming for a finished result in the first delivery costs more. Both decisions are reasonable, and they are settled when planning the project in phases.
One-off need or long run
A project meant to grow over years calls for foundational decisions in the data model and the architecture. They cost more at the beginning and save rebuilding the system later.
Reach inside the company
Automating how support tickets are routed affects one process. Rolling out the management tool the whole company works with affects every department, and adds support, testing, and a more careful go-live.
When the cost cannot be determined
Research points
Some requests cannot be scoped precisely by any company without starting the work or running a prior study.
Feasibility unknown from the start
A real example: assessing whether commercial activity trackers could be integrated with the tool being built. Until their APIs were reviewed, there was no way to know whether it was possible.
Development without precedent
When what is asked for is genuinely new, there are no earlier projects or industry references to base an estimate on.
New platforms or devices
Integrating a system or a device we have not worked with before may be perfectly feasible, but the work cannot be priced until its API, its documentation, and its limitations have been reviewed.
Continuing another company's work
Extending an older platform built by a third party means first understanding the existing code, the decisions that were made, and the state of its documentation. That review shapes any later estimate.
Only in these cases do we propose a prior study with its cost agreed in advance, usually as part of a discovery. On every other project the quote is free. And if during that study we find the project is not viable, or that its cost is not worth it, we say so at that point: we do not bill hours that lead nowhere.
How all of this turns into a figure
The quote
The factors on this page explain the price differences between one project and another.
The quote is free
A first conversation is enough to give an order of magnitude, including on small projects. Preparing the proposal carries no cost and no commitment.
Price closed before starting
On medium and large projects, the scope and the price of each phase are closed before it starts. If the requirements change during the project, they are revisited at that point, following the planning procedure.
A prior study, only when there is no other way
When a project cannot be scoped without investigating first, we propose a prior study with its cost agreed in advance. It is the exception, not the usual procedure.
Want to know the order of magnitude of your project?
Tell us what you need. The first conversation is free and, if you ask, we can sign a non-disclosure agreement (NDA) beforehand.