ARC-101 Design and Build Integration Solutions Exam Topics and Questions
These Salesforce Design and Build Integration Solutions (ARC-101) exam topics are organized according to official exam domains to help candidates quickly verify coverage and focus on assessment rather than theory. Each domain is paired with topic-wise ARC-101 sample questions that reflect how objectives are tested in the actual exam. This structure enables efficient review, targeted self-assessment, and rapid identification of weak areas when preparing for the Salesforce Design and Build Integration Solutions certification exam.
Let's Practice Free Salesforce ARC-101 Questions Aligned with Official Exam Topics
This topic establishes the conceptual ground for everything that follows. You need to recognise why organisations integrate, how systems connect in practice, and what role middleware plays. The cost comes from ambiguity: terms like single view and system landscape sound self-explanatory until an item asks you to distinguish overlapping use cases or identify which layer owns a transformation. β Why integration matters and how architectures shape it Integration exists because business processes span systems. The importance of integration is measured...
This topic establishes the conceptual ground for everything that follows. You need to recognise why organisations integrate, how systems connect in practice, and what role middleware plays. The cost comes from ambiguity: terms like single view and system landscape sound self-explanatory until an item asks you to distinguish overlapping use cases or identify which layer owns a transformation.
β Why integration matters and how architectures shape it
Integration exists because business processes span systems. The importance of integration is measured by the cost of not doing it: duplicate data entry, stale records, manual reconciliation, and decisions made on incomplete information. Integration architecture defines the topology: point-to-point, hub-and-spoke, or message-bus patterns. You need to recognise when each fits and when one becomes a liability. Point-to-point scales poorly. A hub reduces connections but introduces a single point of failure. The exam expects you to match architecture to constraint, not recite a definition.
β Single view, system landscape, and middleware roles
Single view of the customer means one logical representation built from multiple physical sources, usually achieved through real-time federation or scheduled aggregation. The distinction matters when latency or data ownership is in play. System landscape is the inventory of systems, their boundaries, and the flows between them. Drawing one forces you to name what connects to what and why. Enterprise Service Bus and Extract Transform Load are middleware patterns. ESB routes, transforms, and orchestrates messages in near real time. ETL moves bulk data on a schedule, often for reporting. Confusing the two loses marks: an ESB is not a batch tool, and ETL does not handle transactional requests. Know which workload each serves and when the wrong choice creates a bottleneck.
How Integration Overview is tested
Items present a scenario with multiple systems and ask which architecture or middleware pattern fits. The trap is choosing the pattern that sounds most sophisticated rather than the one that matches the constraint. A scenario with five source systems and a nightly reporting load does not need an ESB; ETL is sufficient and cheaper. Another scenario with sub-second latency requirements and transactional consistency rules out batch. The exam also tests whether you can distinguish single view as a logical concept from the physical mechanism that delivers it. A question might describe federated queries and ask whether that constitutes a single view, or present a nightly data warehouse and ask the same. The answer turns on whether the consumer sees one coherent representation, not on how fresh the data is. Candidates lose marks by conflating architecture with implementation or by defaulting to the most complex option when a simpler one satisfies the requirement.
The practice test includes items that describe a system landscape and ask you to choose the appropriate middleware or integration style. Seeing a dozen variations on that question type before the exam helps you spot the constraint that drives the answer, rather than guessing from keywords in the stem.
The question below gives you a scenario with multiple systems and asks which pattern applies. The correct answer depends on recognising the workload and latency requirement, not the number of systems involved.
An Integration Architect has built a Salesforce application that integrates multiple systems and keeps them synchronized via Platform Events.
What is taking place if events are only being published?
Salesforce offers a dozen native mechanisms for moving data in and out, and the exam expects you to know when each one applies. The weight here is in distinguishing overlapping tools: REST versus SOAP, Streaming API versus Platform Events versus Change Data Capture, callouts versus External Services. Every capability has a sweet spot and a boundary. Cross that boundary and you hit a governor limit, a latency problem, or a security gap. β SOAP, REST, and Bulk APIs SOAP API...
Salesforce offers a dozen native mechanisms for moving data in and out, and the exam expects you to know when each one applies. The weight here is in distinguishing overlapping tools: REST versus SOAP, Streaming API versus Platform Events versus Change Data Capture, callouts versus External Services. Every capability has a sweet spot and a boundary. Cross that boundary and you hit a governor limit, a latency problem, or a security gap.
β SOAP, REST, and Bulk APIs
SOAP API is strongly typed and contract-first, suited to enterprise integrations where the schema is fixed and tooling expects WSDL. REST API is lightweight and stateless, the default for mobile and web applications. Bulk API handles large data volumes asynchronously, processing records in batches to avoid timeouts and governor limits. The exam tests whether you can match API to workload. A mobile app syncing a few records at a time does not need Bulk. A nightly load of half a million records cannot use REST without hitting limits. SOAP versus REST often turns on whether the client can parse JSON or needs XML, or whether the integration predates REST adoption. Know the request and response structure for each, and know which limits apply: daily API calls, concurrent requests, batch size, and processing time.
β Event-driven mechanisms
Streaming API, Platform Events, and Change Data Capture all deliver near-real-time notifications, but they differ in scope and control. Streaming API pushes changes to standard and custom objects via PushTopic queries or generic events. Platform Events are custom-defined, fire from code or process automation, and support high-volume pub-sub. Change Data Capture publishes every create, update, delete, and undelete on subscribed objects with minimal configuration, including header fields that standard change tracking omits. The exam asks which one fits a given requirement. If the scenario needs a custom payload or event logic, Platform Events. If it needs every field change on an object with no code, Change Data Capture. If it needs a filtered subset of changes defined by SOQL, Streaming API. Confusing them costs marks because the capabilities overlap but the constraints do not.
β Outbound messages and callouts
Outbound messages are declarative, workflow-driven, and fire when a record meets criteria. They send an XML payload to an external endpoint and retry on failure. The receiving system must handle SOAP and respond with an acknowledgment. Callouts are programmatic, invoked from Apex, and support both SOAP and REST. They run synchronously or asynchronously depending on context, and you control the request and response. The exam tests whether you recognise when declarative is sufficient and when you need code. If the requirement is to notify an external system when a case closes, an outbound message is simpler. If the requirement is to call an API, parse the response, and update multiple objects conditionally, you need a callout. The trap is choosing the programmatic option when the declarative one works, because over-engineering is penalised as much as under-engineering.
β Salesforce Connect, Heroku Connect, and External Services
Salesforce Connect federates data from external sources using OData or custom adapters, making it appear as external objects without storing it locally. Queries hit the remote system in real time, so latency and availability matter. Heroku Connect synchronises data bidirectionally between Salesforce and a Postgres database on Heroku, using triggers and polling. It is not real-time but near real-time, and it stores data in both systems. External Services wraps an external REST API as an invocable action, callable from Flow or Process Builder with no code. The exam asks which tool fits a scenario. If the data must stay external and queries are infrequent, Salesforce Connect. If the data needs to be local for reporting and the source is Heroku, Heroku Connect. If a business user needs to call an API without writing Apex, External Services. The wrong choice usually stems from ignoring latency, data residency, or the skill level of the person building the integration.
How Integration Capabilities is tested
Items describe a requirement and ask which API or mechanism to use. The stem includes constraints: data volume, latency, direction of flow, whether the integration is synchronous or asynchronous, and who builds it. The correct answer satisfies every constraint. The distractors satisfy some but violate others. A scenario with 100,000 records and a four-hour processing window rules out REST but allows Bulk. A scenario requiring sub-second response rules out Bulk and outbound messages. A scenario where a business analyst must configure the integration rules out callouts and favours External Services or declarative tools. The exam also tests whether you know the limits: Bulk API batch size, Streaming API event delivery guarantees, Platform Event daily allocations. Candidates lose marks by choosing the tool they know best rather than the tool that fits, or by missing a constraint buried in the middle of the stem. Read every word. The constraint that disqualifies an answer is rarely in the first sentence.
This topic spans a dozen sub-topics, and each one shows up in items that look similar but turn on different constraints. The practice test lets you work through enough variations that you stop defaulting to REST API for everything and start matching capability to requirement.
The question that follows presents a scenario with volume, latency, and skill constraints. The correct answer is the only one that satisfies all three, and the distractors each violate at least one.
The URL for an external service has been changed without prior notice. The service provides up to date money exchange rates that is accessed several times from Salesforce and is a business critical function for end users.
Which two solutions should an Integration Architect recommend be implemented to minimize potential downtime for users in this situation?
Choose 2 answers
Patterns are named solutions to recurring integration problems. The exam gives you a business requirement and expects you to recognise which pattern applies. Each pattern has a canonical use case, a set of constraints it satisfies, and failure modes when misapplied. The cost is in the overlap: several patterns can solve the same problem, but only one is optimal given the constraints. Missing a nuance in the stem sends you to the wrong pattern. β Request and reply, and fire...
Patterns are named solutions to recurring integration problems. The exam gives you a business requirement and expects you to recognise which pattern applies. Each pattern has a canonical use case, a set of constraints it satisfies, and failure modes when misapplied. The cost is in the overlap: several patterns can solve the same problem, but only one is optimal given the constraints. Missing a nuance in the stem sends you to the wrong pattern.
β Request and reply, and fire and forget
Request and reply is synchronous. The caller sends a request and waits for a response. Use it when the caller needs the result immediately to continue processing, such as validating an address or checking inventory before confirming an order. The constraint is latency: if the remote system is slow or unavailable, the caller blocks. Fire and forget is asynchronous. The caller sends a message and moves on without waiting. Use it when the caller does not need confirmation or when the operation can tolerate delay, such as logging an event or triggering a background process. The exam tests whether you can distinguish need from preference. A scenario that says the user must see the result on screen before proceeding requires request and reply. A scenario that says the operation happens overnight and the user checks status later allows fire and forget. Choosing fire and forget when the business process requires an immediate answer loses marks.
β Batch data synchronisation and remote call-in
Batch data synchronisation moves large volumes on a schedule, typically using ETL or Bulk API. Use it when data does not need to be real-time, when volume exceeds transactional API limits, or when the target system processes data in batches. The trade-off is latency: data is stale between runs. Remote call-in means an external system initiates the transaction by calling into Salesforce, usually via REST or SOAP API. Use it when the external system owns the business event, such as an order placed on an e-commerce site that creates a case in Salesforce. The exam asks which pattern fits a scenario. If the requirement is to synchronise a million customer records nightly, batch. If the requirement is for an external application to create a lead when a form is submitted, remote call-in. The trap is choosing batch when the scenario requires event-driven behaviour, or vice versa.
β UI update based on data changes, and data virtualisation
UI update based on data changes means the user interface refreshes automatically when data changes in Salesforce or an external system. Implement it with Streaming API, Platform Events, or Change Data Capture pushing updates to the client. Use it when users need to see changes without refreshing the page, such as a dashboard showing live case volume. Data virtualisation means querying external data in real time without storing it locally, typically using Salesforce Connect. Use it when data must stay in the source system for compliance or when the data changes too frequently to cache. The exam tests whether you recognise when virtualisation is viable. If the scenario requires complex reporting or high query volume, virtualisation introduces latency and load on the external system. If the scenario requires a single record lookup and the data is authoritative elsewhere, virtualisation fits. Candidates lose marks by choosing virtualisation when the query pattern or volume makes it impractical.
How Integration Patterns is tested
Items present a business scenario and ask which pattern to apply. The stem describes the initiating system, the direction of data flow, the latency requirement, the volume, and whether the caller needs a response. The correct answer is the pattern that satisfies every requirement. The distractors are patterns that fit part of the scenario but fail on a constraint. A scenario with real-time user interaction and a requirement to display the result immediately rules out fire and forget and batch. A scenario with a nightly load of 500,000 records rules out request and reply. A scenario where an external system triggers the flow rules out patterns where Salesforce initiates. The exam also tests whether you know the failure modes. Request and reply fails when the remote system is slow. Data virtualisation fails when query volume is high. Batch fails when the business process requires immediate consistency. Read the scenario for the constraint that eliminates three of the four options. That constraint is the key to the question.
Pattern questions are common, and each one tests whether you can map a scenario to a named solution. The PDF question bank includes stems that describe the same pattern from different angles, so you learn to recognise the underlying requirement rather than matching keywords.
The question below describes a business process and asks which pattern applies. The answer depends on who initiates the transaction and whether the caller waits for a response.
Security is where integrations fail audits and where candidates lose marks by confusing authentication with encryption or by choosing a mechanism that does not match the threat model. This topic covers transport security, credential management, and data protection at rest. The exam tests whether you know which control to apply and when applying it introduces complexity without reducing risk. β Two-way TLS and remote site settings Two-way TLS, also called mutual TLS, means both client and server present certificates to...
Security is where integrations fail audits and where candidates lose marks by confusing authentication with encryption or by choosing a mechanism that does not match the threat model. This topic covers transport security, credential management, and data protection at rest. The exam tests whether you know which control to apply and when applying it introduces complexity without reducing risk.
β Two-way TLS and remote site settings
Two-way TLS, also called mutual TLS, means both client and server present certificates to authenticate each other. Use it when the remote system requires proof of the caller's identity beyond a username and password, common in financial services and regulated industries. One-way TLS, where only the server presents a certificate, is sufficient for most integrations. The exam tests whether you recognise when mutual TLS is required and when it is overkill. Remote site settings are a Salesforce-specific control that whitelists external endpoints for outbound callouts. If the endpoint is not in the remote site settings list, the callout fails. Use it to prevent code from calling arbitrary URLs. The exam asks when to configure it: always, for any outbound HTTP or HTTPS callout from Apex or external services. The trap is thinking it handles authentication. It does not. It only permits the connection.
β Named credentials, certificates, and encryption at rest
Named credentials store authentication details for external systems, including endpoint URL, username, password, and OAuth tokens. Use them instead of hardcoding credentials in code or custom settings. They integrate with Salesforce authentication providers and support per-user or per-org credentials. The benefit is centralised management and automatic token refresh. Certificates are used for signing and encrypting messages, and for mutual TLS. Salesforce lets you upload a certificate and private key, then reference it in integrations. The exam tests whether you know when a certificate is required: mutual TLS, SAML assertions, signed requests. Salesforce Shield Platform Encryption encrypts data at rest using AES-256. Use it when compliance requires encryption of sensitive fields even when the database is compromised. The trade-off is cost and complexity: encrypted fields cannot be used in certain query filters, formula fields, or validation rules. The exam asks when encryption at rest is necessary and what functionality it limits.
How Integration Security is tested
Items describe a security requirement or compliance mandate and ask which control to apply. The stem includes clues: regulatory context, the sensitivity of the data, whether the external system requires client authentication, and whether credentials must be rotated. The correct answer matches the requirement. The distractors either over-engineer or under-protect. A scenario requiring client authentication to an external API that demands mutual TLS rules out username and password alone. A scenario where credentials must be updated without deploying code rules out hardcoding and requires named credentials. A scenario requiring encryption of social security numbers at rest when the database is compromised requires Shield Platform Encryption, not field-level security or sharing rules, which protect data in use but not on disk. Candidates lose marks by confusing controls: thinking remote site settings handle authentication, or thinking named credentials encrypt data at rest. Each control has one job. Know which job each does and when the scenario requires it.
Security questions test whether you can distinguish overlapping controls. The practice test includes scenarios that describe a threat or compliance requirement and ask which mechanism applies, so you learn to match control to risk rather than applying every control to every integration.
The question that follows describes a compliance requirement and asks which security control satisfies it. The answer turns on what the control protects against, not what it sounds like it should do.
This topic shifts from what to build to how to build it so it survives production. Performance and scale are where integrations fail under load. Regulatory requirements constrain design choices. Best practices are the accumulated scar tissue of past failures. The exam expects you to apply these lenses to a scenario and identify the design flaw or the compliance gap before it becomes an incident. β Performance and scale Performance is about latency: how fast a single transaction completes. Scale...
This topic shifts from what to build to how to build it so it survives production. Performance and scale are where integrations fail under load. Regulatory requirements constrain design choices. Best practices are the accumulated scar tissue of past failures. The exam expects you to apply these lenses to a scenario and identify the design flaw or the compliance gap before it becomes an incident.
β Performance and scale
Performance is about latency: how fast a single transaction completes. Scale is about throughput: how many transactions the system handles concurrently without degrading. Designing for performance means reducing round trips, caching where safe, and choosing synchronous or asynchronous based on user experience. A UI-driven integration that blocks the user for three seconds fails the performance test. Designing for scale means respecting governor limits, batching work, and using asynchronous patterns when volume exceeds API rate limits. An integration that works for 100 records and fails at 10,000 fails the scale test. The exam presents a scenario with volume, concurrency, and latency requirements, then asks what breaks and how to fix it. The trap is optimising for performance when the real problem is scale, or vice versa. A slow query is a performance problem. A design that hits the daily API limit at noon is a scale problem. The fixes are different.
β Regulatory requirements and best practices
Regulatory requirements vary by industry and geography: GDPR, HIPAA, PCI-DSS, SOX. Each imposes constraints on data residency, encryption, audit trails, and access control. The exam asks which requirement applies to a scenario and how it changes the design. A GDPR scenario might require data to stay in the EU, ruling out certain cloud regions or requiring data virtualisation instead of replication. A HIPAA scenario might require encryption at rest and in transit, and an audit trail of every access. Best practices are lessons learned: idempotency to handle retries safely, correlation IDs to trace requests across systems, circuit breakers to fail fast when a downstream system is unavailable, and monitoring to detect failures before users report them. The exam tests whether you recognise when a design violates a best practice and what the consequence is. An integration with no retry logic fails silently. An integration with no idempotency key processes the same order twice. The cost of ignoring best practices is invisible until production, and then it is expensive.
How Advanced Integration Discussions is tested
Items describe a production scenario: volume has grown, latency has increased, a regulatory audit has identified a gap, or a failure has occurred. The question asks what caused the problem or how to fix it. The correct answer identifies the root cause and proposes a solution that addresses it without introducing a new problem. The distractors fix a symptom or introduce complexity that does not help. A scenario where an integration hits the daily API limit at midday needs a scale fix: batching, asynchronous processing, or Bulk API. Adding more API calls does not help. A scenario where users wait five seconds for a response needs a performance fix: caching, reducing payload size, or making the call asynchronous. Adding a faster server might help but does not address the design. A scenario requiring GDPR compliance needs data residency and the right to erasure. Encrypting the data without addressing residency does not satisfy the requirement. Candidates lose marks by proposing solutions that sound good but do not match the constraint, or by missing the regulatory requirement buried in the scenario.
Advanced topics test whether you can diagnose a failure or design flaw from a description of symptoms. Working through practice items that describe production problems helps you learn to identify the root cause rather than guessing from the first keyword you recognise.
The question below describes a production issue and asks what caused it. The correct answer identifies the design flaw, not the technology involved.
The case study synthesises everything. You are given a business scenario and asked to design, build, or maintain an integration solution. This means drawing a system landscape, translating business needs into technical requirements, choosing the right pattern and capability, and planning for change. The exam tests whether you can apply the previous five topics to a realistic problem, not whether you can recall a definition. β System landscape and business needs Drawing a system landscape diagram means identifying every system...
The case study synthesises everything. You are given a business scenario and asked to design, build, or maintain an integration solution. This means drawing a system landscape, translating business needs into technical requirements, choosing the right pattern and capability, and planning for change. The exam tests whether you can apply the previous five topics to a realistic problem, not whether you can recall a definition.
β System landscape and business needs
Drawing a system landscape diagram means identifying every system involved, the data each owns, and the flows between them. The diagram forces you to name what connects to what and why. The exam might present a scenario and ask which systems must integrate, or present a diagram and ask what is missing. Evaluating business needs means reading a set of requirements and identifying what must happen, who initiates it, what the latency and volume constraints are, and what constitutes success. The exam tests whether you can distinguish a business need from a technical preference. A requirement that says users must see inventory levels before confirming an order is a business need. A requirement that says the integration must use REST API is a technical preference, and it might not be the right one.
β Requirements, design, build, and maintain
Translating needs to integration requirements means turning business language into technical constraints: synchronous or asynchronous, real-time or batch, which system owns the data, how errors are handled, and what the rollback plan is. Designing an integration solution means choosing the pattern, the API, the authentication method, and the error handling strategy. The design must satisfy every requirement and respect every constraint. Building an integration solution means configuring the chosen capability: setting up named credentials, writing the callout or the external object definition, configuring the outbound message or the Platform Event. The exam tests whether you know the steps and the order. Maintaining an integration solution means monitoring for failures, handling schema changes in external systems, and scaling as volume grows. The exam asks what breaks when the external system changes its API or when volume doubles, and how to detect the failure before it cascades.
How Case Study is tested
Case study items present a multi-paragraph scenario with several systems, several requirements, and several constraints. The question asks you to choose the right pattern, identify the missing requirement, spot the design flaw, or propose a fix. The correct answer satisfies every constraint and addresses the root problem. The distractors satisfy some constraints but violate others, or they solve a different problem. The stem is long and the constraints are scattered across paragraphs. A sentence early in the scenario might say the external system is only available at night, ruling out real-time patterns. A sentence in the middle might say the integration must be built by a business analyst, ruling out Apex. A sentence at the end might say volume will grow ten times next year, requiring a solution that scales. Missing any one of those constraints sends you to the wrong answer. The exam rewards careful reading and eliminates candidates who skim. Case study items take longer to answer than single-topic items, and they are worth the same number of marks, so time pressure is part of the test.
Case study questions are the longest and the hardest, and they test whether you can apply multiple topics to one problem. The practice test includes full-length case scenarios so you learn to extract every constraint from a long stem and eliminate answers systematically.
The question below presents a multi-system scenario with several requirements. The correct answer is the only one that satisfies all of them, and finding it requires reading every sentence.
Ready to Start Practicing?
Access all questions and start your exam preparation journey
Upgrade to Full ARC-101 Exam Questions π