Why software integration projects get complex, and how to plan for it
“We just need the two systems to talk to each other.” It is one of the most common requests in business software, and one of the most underestimated. Connecting two applications is usually the easy part. The hard part is agreeing what the information means, what should happen when it does not match, and who looks after the connection once it is live.
This article explains where the complexity comes from, when a ready-made connector is enough and how to plan an integration that keeps working.
Connecting two applications is rarely the whole job
Most modern systems offer an API, and many integration platforms provide ready-made connectors. Moving a record from one system to another can take an afternoon. But a dependable integration also has to answer business questions: which system is right when they disagree, what happens to a record that fails, how duplicates are handled and how the people affected will know something has gone wrong.
What makes integrations complex
The same information means different things
A “customer” in your CRM might be a company. In your accounting system, it might be a billing account. In your support tool, it might be an individual. Statuses, product codes and dates can differ in the same way. Before data can flow, someone has to decide how these map to each other.
There is no single source of truth
If an address can be changed in three systems, which one wins? Without a clear owner for each piece of information, integrations copy errors around faster than people ever could. Duplicate records make this worse: the same customer or patient can exist several times with small differences. Matching and merging those records, which is what our product Transform does, is often a project in itself.
Business processes have conditions
Real processes branch. An order may need approval above a certain value, a booking may depend on availability, and a lab result may need to reach a specific doctor. The integration has to follow these rules, not just copy fields.
APIs and vendors have limits
APIs can be incomplete, slow or limited in how many requests they accept. Some vendors charge for API access or change it with little notice. Older systems may offer only file exports or direct database access. These limits shape what is possible and how much it costs.
Errors need recovery, not just reporting
Networks fail, systems go offline for maintenance and data arrives in the wrong format. A good integration retries sensibly, avoids creating duplicates when it does, records what happened and gives someone a clear way to fix problems. Reconciliation reports that compare both sides are invaluable.
Security and access must be designed in
Each connection is a route into your data. Integrations need their own credentials with only the access they require, encryption in transit, audit logs and a plan for rotating secrets. Where single sign-on and role-based access are in place, the integration should respect them rather than bypass them.
When a ready-made connector is enough
For common, low-risk flows between popular products, such as sending new web enquiries to a CRM, a connector or an integration platform is often the right answer. It is quick to set up, and the vendor maintains it. If the data is simple, the rules are straightforward and occasional manual fixes are acceptable, start there.
When custom integration is justified
Custom work makes sense when the information is sensitive or business-critical, the rules are complex, volumes are high, one of the systems is old or unusual, or errors are expensive. In healthcare, for example, HEAL integrates with hospital information systems and brings laboratory orders and results into the patient record. In India, HEAL is also integrated with the Ayushman Bharat Digital Mission for ABHA creation and verification and consent-based sharing of health records. Flows like these need careful mapping, strict access control and dependable error handling. You can read more on our HEAL integrations page.
Plan for the day after go-live
An integration is a living part of your operations. Before you start, agree:
- Who owns it, both technically and in the business
- What is monitored, and who receives the alerts
- How failed records are reviewed and corrected
- How changes to either system will be tested before they reach production
- Where the documentation lives
Start with the outcome
The most useful first question is not “can these systems connect?” but “what should be different for the business when they do?” Fewer re-keyed orders, faster invoicing, or results reaching a doctor without a phone call are outcomes that guide every design decision that follows.
Questions to ask before you start
- What information needs to move, in which direction and how often?
- Which system owns each piece of information?
- What should happen when records do not match or a transfer fails?
- Do both systems have APIs, and are there costs or limits?
- Who will monitor the integration and fix problems?
If you are adding AI capabilities to systems you already run, see AI integration into existing software. If the integration is part of replacing an older system, our guide to replacing legacy business software covers the wider picture.
Planning an integration?
Tell us which systems are involved and what needs to move between them. We will help you work out whether a connector will do or something more is needed.
