A Salesforce® integration project goes from "should work in 4 weeks" to "8 months and still broken" almost always for the same reason. The consultant picked at SOW signing did not have the capability the project required. The CFO learns this in month 5 when the integration cannot ship and the renewal conversation gets uncomfortable.
This post is the 2026 vendor checklist Sapota's Salesforce team gives founders evaluating integration consultants. 7 items that sort generalists from specialists, plus the red flags that filter out 80 percent of bad fits in the first conversation.
What a Salesforce integration consultant actually does
The role covers the work between Salesforce and every other system the business runs. Common scopes:
- CRM data sync. Salesforce talks to NetSuite, SAP, Oracle Financials. Customer master, opportunity sync, invoice posting.
- Marketing data flow. Salesforce talks to Marketing Cloud, HubSpot, or Marketo. Subscriber sync, campaign attribution, lead scoring.
- Operational integrations. Salesforce talks to ERP, OMS, fulfilment, support tooling.
- Custom data warehouse pipelines. Salesforce data feeds Snowflake, BigQuery, Redshift for analytics.
- External application integration. Salesforce embedded in a custom app, mobile app, or partner portal.
Each scope has different technical patterns. A consultant who shipped 3 CRM-to-ERP integrations may struggle with marketing data flows. A MuleSoft architect may over-engineer a point-to-point Apex callout. Capability fit matters more than brand fit.
The 3 capability profiles
Consultants cluster into 3 profiles. Each fits different engagement shapes.
Integration architect
Holds platform certifications across MuleSoft, Salesforce Functions, custom Apex callouts. Designs the integration topology, picks the right tool per integration, oversees implementation.
Fit: enterprise programs with 5+ source systems, MuleSoft licence already in play or under evaluation.
Cost: USD 200 to USD 500 per hour for senior architects. Annual MuleSoft licence USD 80K to USD 500K depending on tier.
Middleware specialist
Boomi, Workato, Jitterbit, Mulesoft Composer. Builds visual integrations on top of low-code middleware platforms.
Fit: programs with 3 to 8 source systems where citizen-developer maintenance matters. Faster to ship than custom-code; lower licence cost than MuleSoft enterprise.
Cost: USD 100 to USD 250 per hour. Middleware licence USD 10K to USD 80K per year.
Apex developer
Custom-code-only integrations via Salesforce Apex callouts, Named Credentials, and Platform Events. No middleware platform.
Fit: point-to-point integrations under 5 systems. Engineering team can support custom code long-term.
Cost: USD 80 to USD 200 per hour for senior Apex developers. No platform licence beyond Salesforce itself.
The 7-item vendor checklist
These 7 items sort consultants in a 30-minute call. Real practitioners answer with specifics; pitch-deck consultants give surface-level descriptions.
Item 1: 3 most recent sandbox-to-prod deployments
The consultant should name 3 specific recent integrations they shipped. Specific source systems, specific data flows, specific outcome metrics.
What a real answer sounds like: "Shipped NetSuite to Salesforce customer sync last month. 50K customer records, daily delta, OAuth 2.0 with refresh token, dead-letter queue for failed updates, average end-to-end latency 8 minutes." That answer demonstrates capability.
What a fake answer sounds like: "We have extensive integration experience with major ERPs." That is marketing speak.
Item 2: OAuth 2.0 setup pattern
OAuth 2.0 is the modern auth standard. The consultant should explain their specific pattern: client credentials vs authorization code flow, refresh token rotation, secret management, scope handling.
What to listen for: token storage strategy (Named Credentials in Salesforce vs external secret manager), refresh failure detection (silent vs noisy), per-environment configuration (sandbox vs prod credentials separated). Consultants that handwave OAuth produce integrations that work for 6 months then break silently when a token expires.
Item 3: Error handling discipline
Every integration has failures. The discipline is in how the consultant handles them.
What to listen for: defined error categories (transient vs permanent), retry policy (exponential backoff with jitter), dead-letter queue for permanent failures, alerting on failure thresholds, replay mechanism for fixed failures. Consultants without explicit error handling produce integrations that look successful but lose data.
Item 4: Retry logic with idempotency
Retries are dangerous without idempotency. A retry on a non-idempotent operation can double-charge a customer or create duplicate records.
What to listen for: idempotency key per request (UUID or business key), receiver-side dedup logic, defined retry count limit, escalation path on retry exhaustion. The Apex callouts pattern covers the specifics for Salesforce-side implementation.
Item 5: Bulk vs streaming decision
Integration timing matters. Some flows need real-time (order placed in Shopify, must reach OMS in seconds). Some flows are batch (nightly customer master sync to data warehouse).
What to listen for: explicit decision per integration, awareness of trade-offs (latency vs cost vs complexity), specific patterns for each. Platform Events for streaming. Bulk API for high-volume batch. Change Data Capture for moderate-volume real-time. The SOQL selectivity reference covers the data-volume-dependent patterns.
Item 6: Monitoring stack
Integration is not done at go-live. The monitoring stack catches drift, failures, and silent data loss.
What to listen for: metrics layer (request count, latency p50/p99, error rate), alerting layer (defined thresholds, escalation paths), log retention (days to weeks for debugging), trace correlation across systems. The integration is operationally healthy only when the monitoring stack proves it.
Item 7: Post-launch support model
Most integrations need 6 to 12 weeks of post-launch tuning before they stabilise. The consultant who disappears at go-live leaves 80 percent of the value on the table.
What to listen for: defined scope for first 90 days, named engineer for continuity, response SLA, escalation path. The "we hand off and you can come back if you need anything" answer is a red flag.
MuleSoft vs Salesforce Functions vs custom Apex
The tool choice depends on scope and budget.
When MuleSoft fits
5+ source systems with overlapping integration patterns. Enterprise budget that justifies USD 80K to USD 500K annual licence. Need for citizen-developer accessibility on integration changes. Long-term hub-and-spoke architecture commitment.
MuleSoft shines on the "build it once, reuse it everywhere" pattern. Connectors to standard systems exist; the API-Led Connectivity pattern keeps integrations reusable across consumers.
When Salesforce Functions fits
Need to run Dataverse-native logic that exceeds the Apex 10-second sync limit, requires Node.js or Java runtime, or integrates with external systems that need long-running operations.
Functions sit inside the Salesforce platform but run on Heroku infrastructure. The benefit is platform-native deploy and security; the cost is Functions-specific licence and complexity.
When custom Apex callouts fit
Point-to-point integrations under 5 systems. Engineering team has Apex expertise. Latency requirements that match Apex callout limits (120-second max, 100 callouts per transaction). Budget that does not justify middleware licence.
Custom Apex is the lightest-weight option. Named Credentials handle OAuth, Platform Events handle event-driven flows, Apex callouts handle synchronous integrations. The Apex callouts reference covers the patterns.
The right consultant explains which tool fits which integration. Consultants pitching MuleSoft for every project are over-engineering; consultants pitching custom Apex for every project are under-engineering.
Red flags that filter consultants fast
The 3 patterns that filter 80 percent of bad fits:
Red flag 1: Vague capability claims
"We have done many Salesforce integrations" beats nothing but loses to specific recent shipping examples. Real consultants volunteer specifics; pitch-deck consultants stay at the brand level.
Red flag 2: Universal MuleSoft pitch
Consultants pitching MuleSoft for every integration are working on commission or over-engineering. MuleSoft fits roughly 20 percent of integration programs. For the other 80 percent, the cost outweighs the benefit.
The right consultant says "MuleSoft makes sense for this program because A, B, C; for the other program, we would use Apex callouts because D, E, F".
Red flag 3: Fixed-bid SOW without retry plus idempotency plus monitoring
A fixed-bid SOW that scopes "implement integration between Salesforce and ERP" without specifying retry behaviour, idempotency, and monitoring is open-ended. The consultant is incentivised to skip the operational discipline because it does not appear in the scope.
The right SOW lists retry policy, idempotency mechanism, alerting thresholds, log retention, and post-launch support window as explicit deliverables. The scope drives the consultant's incentives.
How to evaluate the answer quality
The 30-minute initial call should cover the 7 checklist items. Score each answer 0 (unable to answer), 1 (surface-level), 2 (specific and detailed).
A score of 10 or higher across the 7 items signals a capable consultant. A score below 8 signals either inexperience or capability gap. A score below 5 is a hard skip.
The 2 most diagnostic items are #1 (recent deployments) and #4 (retry with idempotency). Consultants who score well on these typically score well across the board; consultants who score poorly on these usually score poorly everywhere.
Engagement model patterns
3 engagement structures dominate integration consulting.
Pattern 1: Fixed-bid project
Defined scope, defined timeline, defined price. Typical Salesforce integration project: USD 40K to USD 200K depending on complexity. Works when scope is genuinely fixed; breaks under scope creep.
Pattern 2: Time and materials
Hourly billing. Typical senior rate USD 150 to USD 300 per hour for Salesforce integration specialists. Works when the client has internal project management and trusts billing honesty.
Pattern 3: Monthly retainer with paid trial
Sapota's standard. 2-week paid trial scoped to one integration deliverable. USD 1,800 to USD 2,400 per engineer per month rolling monthly after conversion. No lock-in.
The trial structure works for integration projects because the first 2 weeks reveal capability gaps that pitch decks hide. A consultant can pitch retry-with-idempotency cleanly; only the actual code reveals whether they ship it cleanly.
Specific reading that helps vet consultants
Before signing the SOW, read the integration architecture patterns the consultant should know:
A consultant that can engage at depth on these 4 has the integration capability the SOW needs. A consultant that handwaves them does not.
Need a Salesforce integration consultant
Sapota's Salesforce team holds 5 Salesforce certifications including Administrator, Platform Developer I and II, and Marketing Cloud Consultant. We have shipped integrations across NetSuite, SAP, Oracle Financials, Shopify, Snowflake, and BigQuery.
Visit the Salesforce service page for the team capability list and engagement model. The complete SFMC implementation guide covers 70+ patterns across the full Marketing Cloud and integration surface.
The 2-week paid trial scoped to one specific deliverable runs USD 1,800 to USD 2,400 per engineer per month after conversion. Walk-away terms documented up front. No lock-in.
Integration consulting is mostly capability matching
The technology decisions matter but the capability fit matters more. The right consultant for an enterprise hub-and-spoke MuleSoft program differs from the right consultant for a point-to-point Apex callout integration. The 7-item checklist exists precisely because pitch decks hide capability gaps.
The founders who get this right vet consultants deliberately and walk away from bad fits fast. The founders who get this wrong pick on price or brand and pay the cost in months 4 through 8 of the engagement.