Cost to build a custom Salesforce integration
Cost to build a custom Salesforce integration
Building a custom Salesforce integration in 2026 typically costs $5,000 to $250,000, depending on direction, latency, and whether you ship to AppExchange. A one-way nightly sync runs $5k to $15k. A two-way real-time sync runs $15k to $40k. A Lightning component or AppExchange app runs $30k to $100k. A fully managed package with security review runs $80k to $250k.
The real cost driver is not Salesforce itself. It is the architecture choice you make in week one: REST vs Bulk, polling vs Platform Events, OAuth user flow vs JWT bearer, Apex vs Flow vs external orchestrator. Get those right and a senior ships a clean two-way sync in six weeks. Get them wrong and you spend three months untangling governor-limit errors at 3 a.m.
What actually drives the cost
Five things, in order of impact:
- Direction. A one-way pull from your DB into Salesforce is roughly half the work of a two-way sync. Two-way means conflict resolution, change tracking on both sides, and a tie-breaker rule for simultaneous edits.
- Latency target. Nightly batch is cheap. Sub-minute real-time means Platform Events or Change Data Capture, which means streaming infrastructure on your side too.
- Auth model. OAuth 2.0 with the user flow is the wrong choice for almost every server-to-server integration in 2026. JWT bearer with a Connected App and a dedicated Integration User license is the right choice. Picking the wrong one adds two engineer-weeks of rework.
- Where business logic lives. Apex triggers and Flow are powerful but bind you to Salesforce release cycles. External orchestration (in your own service) is more portable but means more API calls.
- Distribution model. A single-tenant integration with one Salesforce org is one job. A managed package distributed via AppExchange is another job entirely, and the security review alone takes 4 to 12 weeks in queue.
Salesforce Enterprise gives you 100,000 API calls per rolling 24 hours plus 1,000 per user license. A 50-seat org has 150,000 calls per day. A poorly designed iPaaS workflow burns 5 to 7 calls per record. Sync 30,000 records and you have eaten the daily budget.
The four real cost tiers
This is where most cost guides hand-wave. Here is the actual engineer-week math, using Cadence pricing tiers as the unit of account.
Tier 1: One-way sync, $5k to $15k
What you are building: a nightly or hourly job that pulls records from your application database and upserts them into Salesforce. Maybe contacts, maybe orders, maybe a custom object.
Architecture:
- REST API with the upsert endpoint and an external ID
- Bulk API 2.0 if you cross 2,000 records per run
- JWT bearer auth with a Connected App
- A simple cron job (GitHub Actions, AWS EventBridge, your existing scheduler)
Engineer profile: a junior or mid engineer with good Salesforce REST docs in front of them. At Cadence pricing, that is $500 to $1,000 per week for 2 to 6 weeks of work. Total: $1,000 to $6,000 in engineer time, plus $4,000 to $9,000 for testing, edge cases, and a small admin UI to monitor sync health.
This is the tier where iPaaS often wins. If you only need to push contacts daily and want zero infra, Zapier at $240/year is the right call.
Tier 2: Two-way real-time sync, $15k to $40k
What you are building: changes in either system replicate to the other within seconds.
Architecture:
- Outbound: Platform Events or Change Data Capture for real-time push from Salesforce
- Inbound: REST upserts from your service, idempotent
- Conflict resolution rule (last-write-wins, or field-level merge, or "Salesforce wins for ownership fields, app wins for engagement fields")
- A reconciliation job that runs nightly and catches drift
- Dead-letter queue and retry policy for both directions
Engineer profile: a mid or senior engineer for 4 to 10 weeks. At $1,000 to $1,500 per week, that is $4,000 to $15,000 in engineer time, plus $10,000 to $25,000 for the conflict-resolution work, the reconciliation job, and the observability layer that keeps you from finding out about a sync failure from a screaming customer.
This tier is where most teams underestimate. The naive plan ("we'll just listen to Platform Events on our side") becomes a real distributed-systems problem the moment you hit your first network partition.
Tier 3: Custom Lightning component or AppExchange app, $30k to $100k
What you are building: a UI that lives inside Salesforce. A custom record page, a related-list with bespoke logic, a side panel that shows data from your service inline on the contact view.
Architecture:
- Lightning Web Components for the UI
- Apex controller for server-side logic and external callouts
- Named Credentials for outbound auth
- Apex test classes (Salesforce mandates 75% code coverage on production deploys)
- Possibly an AppExchange listing if distributed
Engineer profile: senior, often with named Salesforce certifications. Six to fifteen weeks at $1,500 per week, so $9,000 to $22,500 in engineer time, with the rest in design, certification-flavored testing, and admin documentation.
For this tier, the Apex 75% coverage rule is real money. It adds 30 to 40 percent to dev time. AI test-stub generation (Cursor or Claude Code) cuts that meaningfully if your engineer is fluent with it, but you cannot skip it.
Tier 4: Managed package with security review, $80k to $250k
What you are building: a packaged integration distributed to multiple Salesforce orgs via AppExchange.
Architecture:
- Everything in Tier 3
- Namespace prefix on every metadata object
- Multi-tenant data isolation
- Trial and provisioning flow
- License Management App for billing
- AppExchange security review (the big one)
The security review is the cost driver nobody quotes. The queue runs 4 to 12 weeks before Salesforce even looks at your submission. Plan for two to three review rounds, each adding 2 to 4 weeks. The review is free in dollars but expensive in lead-engineer time and delayed revenue.
Engineer profile: lead-level for 12 to 25 weeks. At $2,000/week, that is $24k to $50k in engineer time, plus $50k to $200k in security remediation, listing assets, partner fees, and the inevitable mid-project pivot.
Our breakdown of the cost to build a custom WordPress plugin walks through the same managed-distribution math (namespace, review queue, remediation cycle).
REST vs SOAP vs Bulk vs Streaming
This decision happens in week one and locks you in. Pick wrong and the rebuild is six weeks.
| API | When to use | Throughput | Real-time? |
|---|---|---|---|
| REST | Default for any new build, sub-2,000 records per run | 25 records per request | Synchronous |
| Bulk API 2.0 | Anything over 2,000 records, ETL jobs, migrations | 150M records per day | Asynchronous |
| SOAP | Only if integrating with legacy middleware that requires WSDL | Similar to REST | Synchronous |
| Streaming (Platform Events) | Push-based, sub-second latency, decoupled architectures | 50k events/hour standard | Yes (push) |
| Change Data Capture | Real-time replication of Salesforce data changes | Limited by daily event allocation | Yes (push) |
Polling adds 5 to 15 minutes of latency and burns API calls. Platform Events and CDC give you sub-second push at the cost of running a streaming consumer on your side (Kafka, Pub/Sub, or a long-lived server with the EMP connector).
If your sync needs are similar to what Stripe or Twilio webhooks already handle in your stack, model the Salesforce side the same way. Outbound webhooks from Flow are simpler than Platform Events for most teams.
Auth: JWT bearer beats OAuth user flow for server-to-server
Pick the wrong auth model and every other engineering choice gets harder.
The right default for a backend integration is the OAuth 2.0 JWT Bearer flow with a Connected App and a dedicated Integration User license. The Integration User license is roughly half the price of a full Salesforce user license and is purpose-built for this. You generate a key pair, upload the public key to your Connected App, and your service signs JWTs to mint short-lived access tokens. No refresh-token storage, no user redirect, no token expiry surprises in production.
Use OAuth 2.0 with the standard user flow only when each end-user of your application has their own Salesforce account and you are acting on their behalf inside the app (think: a Chrome extension that reads the logged-in user's records).
Username-password OAuth flow is deprecated for new builds. Do not use it.
Multi-tenant adds a layer. If your app connects to many customer Salesforce orgs, you need per-tenant token storage, key rotation, and revocation handling. Roughly an extra week of dev plus a week of testing. Our breakdown on the cost to integrate Stripe payments into your app covers the same multi-tenant credential pattern.
Build vs buy: when iPaaS beats custom (and when it does not)
iPaaS (MuleSoft, Workato, Boomi, Zapier, Tray, Integrate.io) is real software with real strengths. Be honest about where it wins.
| Tool | Median annual cost | Best for |
|---|---|---|
| Zapier | $240 | Under 100k tasks/month, standard objects, single workflow |
| Workato | $65,000 | Mid-market, multi-system orchestration, ops teams |
| MuleSoft | $55,000 to $250,000+ | Enterprise, complex API governance, existing Mule shop |
| Boomi | $50,000 to $190,000 | Mid to enterprise, hybrid cloud-on-prem |
| Skyvia / Hevo | $1,000 to $3,000 | Lightweight ETL, dashboards, batch syncs |
Where iPaaS wins:
- Time to first sync is days, not weeks
- Standard objects (Contact, Account, Opportunity) with no custom logic
- Non-engineering teams own the workflow long-term
- You only need 5 to 15 minute latency
Where custom beats iPaaS:
- Cross 30,000 record syncs per day (the API call tax becomes painful)
- Bespoke business logic that does not fit a no-code mapper
- Multi-tenant SaaS distributing one integration to many Salesforce orgs
- An 18-month total-cost-of-ownership window (recurring license vs one-time build)
Run the math. A two-way sync that an iPaaS does for $65,000 a year costs you $20,000 once if a senior engineer builds it custom. After year one you save $65,000 every twelve months and you keep the IP.
The catch: you also keep the maintenance. Salesforce ships three major releases a year (Spring, Summer, Winter), and each has a non-zero chance of breaking something subtle in your Apex or your Bulk job pagination.
Hidden costs nobody quotes
Five things that do not show up in the headline number.
Sandbox seats. Full Sandbox runs $25k/year and up. Partial Copy ~$5k. Developer is free but caps at 200 MB. Two-way sync work needs Partial Copy minimum.
AppExchange security review. Free in dollars, expensive in time. 4 to 12 week initial queue, plus remediation cycles. Listing fees are separate.
Apex test coverage. The 75% rule is enforced on production deploys. Add 30 to 40% to dev time. AI tooling (Cursor, Claude, Copilot) cuts this when your engineer is fluent, which every Cadence engineer is by default.
Platform releases. Three major Salesforce releases a year, each with deprecations. Budget two engineer-days per release.
Maintenance and on-call. A two-way sync is a real distributed system. Someone has to wake up when it breaks. Either build on-call in house or book a recurring engineer-week to cover it.
Our notes on the cost to migrate from Heroku to AWS cover the same "what nobody quoted you" math on the on-call piece.
Cost-by-approach comparison
| Approach | Cost | Timeline | Pros | Cons |
|---|---|---|---|---|
| US full-time Salesforce engineer | $160k-$220k base + 30% benefits | 8-14 weeks to hire | Owns the system long-term | Slow to staff, hard to right-size for one project |
| Salesforce consultancy (US/EU) | $200-$350/hr blended | 2-4 weeks to start | Predictable delivery, certified team | Premium rate, change orders, hand-off risk |
| Upwork freelancer | $25-$120/hr | 1-2 weeks to start | Cheapest sticker price | 40% rework risk, no AI fluency baseline |
| Toptal | $80-$200/hr | 1-3 weeks to vet | Vetted bench | Limited Salesforce-specific availability |
| MuleSoft / Workato (iPaaS) | $55k-$190k/yr license | Days to weeks | Fast for standard objects | API tax, recurring license, generic |
| Cadence | $500-$2,000/wk | 48-hour trial, then ship | Every engineer is AI-native by default, weekly billing, replace any week | Less suited to enterprise procurement |
A worked example: a two-way real-time sync with conflict resolution (tier 2). A consultancy quotes $40k to $80k. A senior on Cadence at $1,500/week ships in 6 weeks for $9,000 plus PM time. You keep the source and you can replace any week if the fit is wrong. To scope yours, you can book a senior engineer on Cadence and use the 48-hour free trial first.
How to cut the cost in half without cutting corners
Five things that compound.
- Pick the lightest auth model that works. JWT bearer with an Integration User license. Avoid full user licenses for service accounts. Avoid the username-password flow entirely.
- Use Salesforce Flow before you reach for Apex. Flow is declarative, faster to ship, easier to maintain, and does not need 75% test coverage. Drop into Apex only for logic Flow cannot express.
- Outbound webhooks from Flow before Platform Events. A Flow that fires an HTTP callout on record change is one click. Platform Events are powerful but add a streaming consumer on your side. Start with the webhook and graduate when you actually need the durability.
- Defer the AppExchange listing until you have customers. A managed package without a security review can still be installed via metadata API in customer orgs. Ship the integration first, list it once you have proof of pull.
- Book engineer-weeks instead of hiring FTE for one-shot work. A custom Salesforce integration is a 4 to 12 week project. Staffing it with an FTE means you carry the engineer for 50 weeks. Booking weekly means you stop paying the day the project ships.
The best path from "we need this integration" to "it is in production" in 2026 is usually three steps:
- Sketch the integration on a single page: direction, latency target, objects, auth model
- Decide build vs buy using the 18-month TCO test (custom under $40k almost always beats iPaaS recurring)
- If building, book a senior engineer on Cadence for the 6-week sprint and pay weekly
Cadence has 12,800 vetted engineers in the pool, every one of them AI-native by default and tested on Cursor / Claude / Copilot fluency before they unlock bookings. Trial-to-active conversion runs at 67%, which means most founders who try a 48-hour trial keep the engineer through delivery.
If you are scoping a Salesforce integration right now, see what it costs on Cadence. Get matched in 2 minutes, run a 48-hour free trial, then ship the integration weekly with a senior engineer who is AI-native by default.
FAQ
How long does a custom Salesforce integration take?
Two to twenty-five weeks. A one-way nightly sync ships in 2 to 6 weeks. A two-way real-time sync takes 4 to 10 weeks. A Lightning component or AppExchange app takes 6 to 15 weeks. A managed package with security review takes 12 to 25 weeks because of the AppExchange queue.
REST API or SOAP API: which should I use?
REST for everything new. SOAP only if you are integrating with legacy enterprise middleware that already speaks WSDL. Use Bulk API 2.0 for any sync over 2,000 records per run. Use Streaming or Change Data Capture for sub-minute real-time push.
Do I really need an AppExchange listing?
Only if you want to distribute your integration to other Salesforce orgs you do not own. For a single-tenant integration with your own customers, skip it. The 4 to 12 week security review queue alone can cost $30,000 or more in delayed revenue, and you can install via metadata API or unmanaged package in the meantime.
Can I just use Zapier or Workato instead of building?
For under 30,000 record syncs per day on standard objects with no custom business logic, yes. Above that volume, the API call tax (3 to 7 calls per record) and the per-record license fee make a custom build cheaper inside 18 months.
What is the cheapest way to staff a one-off Salesforce integration?
Book engineer-weeks instead of hiring full-time or paying agency rates. A senior engineer at $1,500 per week shipping a two-way sync in 6 weeks costs $9,000 plus your project manager time. The same scope at a Salesforce consultancy runs $40,000 to $80,000 and locks you into a statement of work with change-order surprises.