Glossary
IT impact analysis
An IT impact analysis establishes what stops or degrades when a component of the infrastructure fails, changes or is removed. It follows the dependency chain from the machine to the applications, the business activities and the people who need to be told.
Updated
What is an IT impact analysis?
An IT impact analysis answers a question that is easy to ask and hard to answer reliably: if this component stops or changes, what stops or changes with it? You start from a server, a database, a network link or a firewall rule and work your way up to the applications, the business activities and the people concerned. In plain terms, it is the answer to "what depends on this server?"
The term covers several neighboring requests. A change impact analysis comes before a planned intervention. An outage impact analysis happens during the incident, under pressure. Dependency mapping is the raw material for both: without a list of what depends on what, impact analysis comes down to the memory of whoever has been around longest.
Two other exercises carry almost the same name and deal with something else: the business impact analysis (BIA) used in business continuity, and the data protection impact assessment (DPIA) required by the GDPR.
When do you run an impact analysis?
Before a change or planned maintenance. You want to patch a virtualization host on Saturday night: which virtual machines does it carry, which applications run on them, which have no second node, who needs a heads-up? The analysis feeds the change request and the rollback plan.
During an incident. A storage array stops responding. You do not get to pick the moment, and within minutes you have to say what is affected, what will be if the outage lasts, and what to restore first. An analysis that requires opening five files is no help at that point.
Before a migration or a decommissioning. You want to switch off a server nobody claims, or move to another hypervisor. Here the analysis looks for what is written nowhere: the monthly job that drops a file on that server, the firewall rule that names its address, the subsidiary's application that still queries its database. A server with no known owner often still has users.
How do you trace dependencies from a machine to the people to notify?
The chain has four links, and each usually lives in a different tool.
The machine and what carries it: host, cluster, storage, site, cloud account. That comes from the hypervisor, the cloud provider or the IT asset inventory. The application: which components run on the machine, and which other machines they form a service with. That comes from the CMDB when it is maintained, from an application spreadsheet otherwise. The flows: who calls the application and what it calls, as recorded in the network flow matrix and the firewall rules. The business capability and the people: which activity stops (payroll, order entry, invoicing), who answers for it, who has to be told.
An analysis nearly always breaks at the joint between two links. The hypervisor knows the machine by one name, the CMDB by another, the flow matrix by an IP address. The application is attached to a server replaced two years ago. The result looks complete and is wrong, which is worse than a result that looks incomplete: maintenance is announced as risk-free, and on Monday an application nobody had tied to that host fails to start.
So an impact analysis is worth what the date and source of each dependency are worth. "This application depends on this server, according to last month's hypervisor export" can be checked. The same sentence with no source and no date cannot.
What is the difference between an impact analysis, a BIA and a DPIA?
A technical impact analysis starts from a component and looks for what depends on it. In ITIL 4 vocabulary it supports the practice called change enablement, which many teams still call change management: assessing the risks of a change before authorizing and scheduling it.
A business impact analysis starts from the other end. ISO 22301:2019, the standard for business continuity management systems, defines it as the process of analyzing the impact over time of a disruption on the organization, and notes that its outcome is a statement and justification of business continuity requirements. The NIST glossary gives a close definition: analyzing operational functions and the effect a disruption might have on them. The BIA says which activities matter most. The technical impact analysis says which machines and applications carry them.
A DPIA is a different thing altogether. The data protection impact assessment comes from Article 35 of the GDPR: where a type of processing of personal data is likely to result in a high risk to the rights and freedoms of individuals, the controller carries out the assessment before the processing starts. Article 35 lists what it contains at a minimum: a description of the processing and its purposes, an assessment of its necessity and proportionality, an assessment of the risks to individuals, and the measures planned to address them. Its subject is people's privacy.
How do you run an impact analysis, step by step?
Start by naming the component without ambiguity: its name, its address, its identifier in the hypervisor or the cloud. Many wrong analyses begin with two machines that have similar names.
List what the component carries directly (the virtual machines on a host, the databases on a server, the volumes on an array), then the applications those belong to. For each application, note whether there is redundancy: a second node, replication, a recovery site.
Add the dependencies that run through flows, in both directions: the applications that call the one going down, and the services it calls itself (directory, DNS, backup, job scheduler). Shared services are the classic omission, because they appear in no application's own documentation.
Move up to the business activities and rank them, using the BIA where one exists. Draw up the list of people: the owner of each application, the business lead, the support team, any outside provider.
The last step is to write down what you do not know. An application with no known owner, a machine missing from the CMDB, a flow that was observed but never documented: these lines go into the analysis as undetermined, and someone is assigned to clear them before the intervention. Presented to a change board, the analysis fits on one page: the component, what stops, what degrades, who is told, what remains to be checked, and the source of each line.
Which tools help with impact analysis and dependency mapping?
A spreadsheet or a Visio diagram is enough for a few dozen servers run by one team. It does not follow changes, and it depends on the person who keeps it.
A CMDB is designed for this: it models relationships between configuration items, and service management tools hang their change requests on it. It gives good results as long as relationships are kept current, and misleading ones when they are not (what a CMDB is and what it contains).
An enterprise architecture tool describes the top of the chain well: business capabilities, applications, exchanges between applications, often in ArchiMate. It seldom goes down to the machine and the firewall rule.
Monitoring and observability tools see actual dependencies between running services, through the calls they measure. They are valuable during an incident. They know nothing of the business capability, or of the powered-off server you would like to decommission.
A map derived from the sources brings together what each tool already knows (hypervisor, cloud, CMDB, flow matrix, rules) and keeps a source and a date for every link. It does not create information that is missing everywhere. It shows where it is missing (what IT system mapping is).
What Uncia brings, and what it does not do
Uncia reads your sources read-only (connectors to clouds, hypervisors, CMDB, firewalls and scanners; dropped-in files such as an RVTools export, a flow matrix or an architecture document) and reconciles them into a knowledge base. For a given machine it returns the impact of a shutdown: the applications and flows that go down. The thread runs from business capability to firewall rule in both directions, and each element's record shows what the base knows, where it got it and where sources disagree.
Uncia monitors nothing, and it replaces neither the change management tool nor the CMDB, which it reads. It only knows the dependencies present in the sources it has read, and says "undetermined" when they are missing. Its knowledge updates with each collection. It produces neither a BIA nor a DPIA, and choosing who to notify, like the decision to go ahead, stays with the team.
Sources
- ISO 22301:2019, Security and resilience, Business continuity management systems, Requirements: public preview of the standard (definition 3.5), accessed
- NIST CSRC glossary, entry "business impact analysis", accessed
- Regulation (EU) 2016/679 of April 27, 2016 (GDPR), Article 35, EUR-Lex, accessed
- AWS Well-Architected, Change enablement in ITIL 4, citing the PeopleCert ITIL 4 practice guide, accessed
Frequently asked questions
What is the difference between an impact analysis and a BIA?
A business impact analysis starts from the organization's activities: ISO 22301:2019 defines it as analyzing the impact over time of a disruption on the organization. A technical impact analysis starts from one infrastructure component and looks for what depends on it. The two complement each other.
Is an IT impact analysis the same as a GDPR DPIA?
No. The data protection impact assessment comes from Article 35 of the GDPR and applies to processing of personal data likely to result in a high risk to the rights and freedoms of individuals. It is about privacy, not technical dependencies.
Do you need ITIL to run an impact analysis?
No. ITIL is a framework of practices that nobody is required to adopt. What an impact analysis needs most is a reliable list of dependencies, however the team organizes its changes.
How do you find out what depends on a server?
By crossing several sources: what the server carries according to the hypervisor or the cloud provider, the applications attached to it in the CMDB or the inventory, and the flows that name its address in the flow matrix and the firewall rules. Each dependency you keep should carry its source and date.
What do you do when a server has no known owner?
Record it as such in the analysis, then look for traces: flows that name it, accounts that log on to it, the application declared in the inventory. Before decommissioning, a period of flow observation and a reversible shutdown lower the risk.