Salesforce Data Loader - When the Web Importer Gives Up

Data Loader is what you use when Salesforce's web importer shits the bed for the third time this week. Version 64.1.0 is the current release (Summer '25) and actually works, which is more than I can say for the previous versions.

Salesforce's built-in Data Import Wizard craps out at 50,000 records. That's cute. Data Loader handles up to 150 million records through Bulk API 2.0. I've personally pushed 2 million contact records through this thing in one go, and it didn't break a sweat - which honestly surprised the hell out of me.

How This Thing Actually Works

You get two ways to use Data Loader: a GUI for clicking around when you're doing one-off imports, and a CLI for automation. Here's the kick in the teeth - CLI only works on Windows. If you're on Mac and want automation, you're screwed harder than a one-legged cat in a sandbox. It's 2025 and they still haven't figured out cross-platform scripting.

The thing talks to Salesforce through their Bulk API, which is actually decent when it's not having one of its famous afternoon naps. You can tweak batch sizes, but the default 200 records per batch works fine unless you're dealing with some weird edge case that makes you question your career choices. You need Java 17+ installed, and it works on Windows 10/11 and macOS 13-15. ARM Macs finally work, thank fuck.

The data processing is actually pretty straightforward for once - your CSV gets chunked into batches, each batch gets processed asynchronously, and you get detailed success and error logs for when shit inevitably goes sideways. Unlike most Salesforce tools, the error messages actually make sense instead of just saying "ERROR: Something bad happened."

What You Can Actually Do With It

Data Loader does five basic operations:

• Insert: Dump new records from CSV
• Update: Fix existing records (you need the Salesforce ID)
• Upsert: Insert or update based on external ID - this one's actually useful
• Delete: Nuke records (careful with this)
• Export: Pull data out with SOQL queries

Works with all Salesforce objects - standard stuff like Accounts and Contacts, plus whatever custom objects your admin dreamed up. Field mapping is drag-and-drop, which is nice. When things break (and they will), you get detailed CSV error logs that actually tell you what went wrong. Pro tip: 100K records takes about 20-30 minutes if everything goes right, 2 hours if it doesn't.

Security Stuff You Need to Know

Version 64.1.0 finally ditched security tokens for OAuth 2.0 with PKCE, which means no more copy-pasting that stupid token from your email every time you reset your password. Thank god.

Data transfer is encrypted over HTTPS, and CLI credentials get encrypted locally. But here's the gotcha that'll bite you in the ass - your exported CSV files just sit there unencrypted on your hard drive like a security team's worst nightmare. If you're dealing with sensitive data, encrypt that shit yourself or store it somewhere secure. I learned this the hard way when our compliance team had a complete meltdown over exported customer data sitting in plain text in my Downloads folder for three weeks.

Set up a dedicated 'data exports' folder with encryption enabled, or your security team will make your life miserable during the next audit - and trust me, they remember everything.

The tool also supports IP restrictions and respects your org's login hours, which is actually useful for limiting when automated jobs can run.

Here's how you actually get Data Loader working without wanting to throw your laptop out the window.

Getting This Thing Installed

Download it here and install like any other desktop app. You need Java 17+ - if you don't have it, the installer will crash faster than my motivation on Monday morning. Download Java from Oracle, install it, then try again.

The "minimum" requirements are 120MB disk space and 256MB RAM, which is complete bullshit. In reality, if you're processing anything serious, give it at least 2GB RAM or it'll be slow as hell. I learned this when trying to import 500K records on a laptop with 4GB total RAM - took 6 hours and crashed twice, nearly giving my manager an aneurysm.

Check the installation troubleshooting guide if things go sideways. The macOS installation is usually smooth, but Windows can be a pain with permission issues.

Permission Bullshit You'll Deal With

Your admin needs to enable "API Enabled" permission in your user profile, which they'll probably forget to do. You'll get "INVALID_LOGIN: Invalid username, password, security token; or user locked out" and waste an hour wondering why Data Loader won't connect before realizing you need to bug them again.

You also need read/write access to whatever objects you're importing to, plus "Bulk API" permission for the good stuff. If your import fails with some cryptic "INSUFFICIENT_ACCESS" error, this is probably why.

Check the user permission checklist to make sure you have everything needed. The permission set configuration is usually cleaner than modifying profiles directly.

Batch Size: Don't Overthink It

Default batch size is 200 records, which works fine 90% of the time. You can crank it up to 2000 if you're feeling brave, but you'll probably hit some API limit or timeout. I stick with the default unless I'm in a hurry.

Bulk API 2.0 bumped the record limit from 5 million to 150 million, which is actually impressive. I've never hit either limit, but it's nice to know I probably won't.

Batch processing is straightforward - Data Loader chunks your CSV into batches, submits them asynchronously, and polls for job status updates. Monitor API usage because each batch counts against your daily limits.

Command Line Setup (Windows Only, Because Screw Mac Users)

CLI automation only works on Windows, which is annoying if you're on Mac. You'll need to set up a bunch of config files:

• Encrypt your password with the built-in utility (don't store it in plain text, obviously)
• Create process-conf.xml files for each operation you want to automate
• Build .sdl mapping files so it knows which fields go where
• Write batch scripts to tie it all together

Most people use Windows Task Scheduler to run this crap nightly. Works fine until Windows decides to randomly not run your scheduled task for no apparent reason - usually around 2am when your manager gets the alert that the nightly sync didn't run and suddenly you're debugging on your phone from bed. Check the CLI troubleshooting guide when (not if) it breaks.

Enable task history in Task Scheduler or you'll be debugging blind when it fails - learned this during a 3am troubleshooting session that lasted until 6am. When you see "Task Scheduler service is not available", the service probably crashed - restart it and cross your fingers. Also, if you get "OutOfMemoryError: Java heap space" during large imports, increase the JVM heap size in the config - this one cost me an entire weekend once.

When Shit Breaks (And It Will)

Data Loader spits out CSV error logs that actually tell you what went wrong. This is refreshing compared to most Salesforce error messages that just say "UNKNOWN ERROR" and leave you to figure it out.

Clean your data first or you'll get thousands of "invalid email format" errors that could've been caught with basic validation. Always test in sandbox first - I once accidentally updated 50,000 account names in production because I forgot to change my target URL. That was a resume-updating conversation with my manager who asked very pointedly why the board of directors was suddenly listed as "Test Account 1" through "Test Account 50,000".

When imports inevitably fail, check the error troubleshooting guide and common error solutions. The data quality checklist could've prevented most of these disasters if I'd bothered to read it first.

API Limits: The Thing That'll Ruin Your Day

Data Loader uses SOAP API for figuring out your org's structure and Bulk API for actually moving data. It's smart enough to pick the right API version automatically, which is nice.

The real pain is API limits. If you're on Professional Edition, you get like 1,000 API calls per day, which sounds like a lot until you realize each batch uses an API call. Suddenly your 200,000 record import becomes a 1,000 batch operation and you're fucked. Monitor this in Setup → System Overview → API Usage or you'll get locked out halfway through.

Use Bulk API 2.0 when possible - it's more efficient and uses fewer API calls. The API limit calculator helps plan large imports.