Zeffy Member Data Migration Guide

1. Project Overview

This guide documents a real-world project to migrate 2,094 membership records from a legacy Excel spreadsheet into Zeffy, a free nonprofit CRM and payment platform. The entire project was completed using Claude AI through Anthropic's Cowork mode, which provides a desktop environment where Claude can read, write, and manipulate files directly on your computer.

The project involved cleaning messy data, resolving quality issues with the organization's input, generating properly formatted CSV import files, and building a comprehensive workbook with documentation tabs for audit trail and future reference.

What Was Accomplished

2. What You Need Before Starting

Your Source Data

You need your membership data in an Excel file (.xlsx). Common fields include: member names, email addresses, mailing addresses, join dates, payment status, member class/type, and newsletter preferences. The data does not need to be clean — that's what this process handles.

A Zeffy Account

Create a free account at zeffy.com if you don't already have one. Zeffy is a nonprofit-focused platform that offers free CRM, payment processing, and email tools. You'll need access to the admin dashboard for the import steps.

Claude Desktop App with Cowork Mode

Claude Cowork is a feature of the Claude desktop application that gives Claude the ability to work directly with files on your computer. It runs in a secure sandbox and can read, write, and edit files in a folder you choose.

• Download the Claude desktop app from claude.ai
• Open a Cowork session and select the folder containing your data
• Claude will be able to read your Excel file and create all output files in that folder

Understanding of Your Data

Before starting, know the answers to these questions, as Claude will ask:

• What do your member class codes mean? (e.g., A, B, CA, CS)
• What does the Newsletter field indicate? (Yes/No for opt-in?)
• Are there shared email addresses between spouses or family members?
• What is the standard membership dues amount?
• Are there any minors in the data who need separate handling?

3. Setting Up Claude Cowork & The Prompts That Matter

This is the most important section of the guide. Claude is powerful but it works best when you give it clear, specific instructions upfront. In our project, many of the most useful features — the Data Quality Audit tab, version-controlled workbook tabs, the old folder structure — only happened because we explicitly asked for them. Claude won't assume you want these things.

The Opening Prompt

Your first message to Claude sets the tone for the entire project. Here is a consolidated prompt based on what actually worked in our project. You can copy and adapt this:

Prompts We Used Throughout the Project

Beyond the opening prompt, here are the specific prompts that drove key outcomes. Each one addresses something Claude would not have done on its own.

Getting the Data Quality Audit

Claude will fix obvious issues like email typos, but it won't create a formal tracking system unless you ask:

This prompt triggered Claude to do a deep investigation of every issue, present findings with record counts, and ask targeted questions so we could make informed decisions together.

Getting Workbook Tabs for CSVs

By default, Claude generates CSV files but doesn't add them as tabs in the workbook. We had to ask:

This gave us tabs like "Zeffy Pay Class A," "Contact A," etc. inside the workbook, which made review much easier than opening individual CSV files.

Getting the Old Folder Structure

Claude will version files if you ask, but won't organize old versions unless told:

After this, Claude automatically moved superseded files to old/ every time it saved a new version.

Getting the Import Plan

Discovering Zeffy Features via Screenshots

Some of the most important changes came from exploring Zeffy's interface and sharing screenshots. Claude immediately understood the screenshots, researched how the features worked, and updated all the import files accordingly. This back-and-forth between you exploring the tool and Claude adapting the data is a natural part of the workflow.

4. What Claude Won't Do Unless You Ask

Claude is highly capable but conservative by default. It focuses on what you explicitly request and avoids making assumptions about project management preferences. Here is a checklist of things you need to explicitly ask for:

5. Phase 1: Data Assessment & Cleanup

This is the most time-intensive phase. Claude reads your Excel file, analyzes every column, and identifies issues that would cause problems during Zeffy import.

Initial Analysis

Ask Claude to examine the original file and report what it finds. Claude will read every sheet, count rows, check for blanks, and identify patterns. In our project, Claude found:

• 187 email fields corrupted with "mailto:" prefixes (an Excel artifact)
• Multiple email typos (missing @ signs, .com misspelled as ,com, gamil.com instead of gmail.com)
• Names with extra whitespace, commas in last names (Jr., Sr. suffixes)
• Mixed case in STATE fields (pa, Pa, PA)
• 55 records with empty payment dates
• 18 minors (under 18) who needed separate handling
• Shared email addresses between spouses
• Phone number fields containing dates of birth

Data Quality Audit Tab

Once Claude creates the Data Quality Audit tab (remember, you must ask for this), it tracks every issue found, its severity, how many records are affected, and the resolution. Each item starts as OPEN and gets marked RESOLVED as you work through them together. This is your checklist — you're not done until everything shows RESOLVED.

Working Through Issues Together

For each issue, Claude presents what it found and recommends a fix. You make the call. This collaborative approach is essential — Claude has the technical skills but you have the institutional knowledge. Examples from our project:

Automated Fixes

Some fixes are straightforward and Claude applies them automatically (with documentation in the Change Log):

• Clearing mailto: corruption from all email fields
• Correcting known email typos (pattern-matched)
• Normalizing STATE to uppercase
• Removing commas from last names (CSV safety)
• Trimming whitespace and normalizing double spaces in names
• Filling empty dates with the current date as a fallback

6. Phase 2: Building Import Files

Once the data is clean, Claude generates the CSV files Zeffy needs. There are two types of import: Payment Import (which also creates contacts) and Contact Import (which adds list assignments and newsletter preferences).

Payment Import CSVs

Zeffy's payment import creates both a payment record and a contact profile for each row. We created one CSV per member class to keep imports manageable and to use the class name as the formTitle (which becomes a form label in Zeffy).

Contact Import CSVs

After payments are imported (which creates contacts), a second contact import updates those contacts with list assignments, subscription preferences, and additional data. We created one CSV per class. These CSVs cover ALL members in each class, not just the ones who opted out of newsletters.

Lists Strategy

We designed six lists that get created automatically when the contact CSV is imported:

• Class A, Class B, Class CA, Class CS — one list per membership class
• All Members — every imported contact belongs to this list
• No Newsletter — members who opted out of newsletters (also marked as "unsubscribed")

Tags vs. Lists

7. Phase 3: Importing into Zeffy

The import follows a specific order. Payment imports must come first because they create the contact records, then contact imports update those records with list assignments.

8. Data Quality Issues We Found & Fixed

This section documents every data quality issue discovered during the project. If you're working with similar membership data, expect to encounter many of these same issues.

9. Version Control Strategy

Every change to the workbook or CSV files was versioned. This is essential for accountability and for being able to roll back if something goes wrong. This entire system only exists because we explicitly asked Claude to implement it.

How It Works

• The original file is never modified. It stays in the project folder as the source of truth.
• Each time Claude makes changes, the workbook gets a new version number (v1, v2, ... v9).
• CSV files also get version numbers that increment independently.
• When a new version is created, the old version is moved to an "old" subfolder.
• The Change Log tab inside the workbook documents exactly what changed in each version.

Version History

10. Final File Inventory

Active Files

Workbook Tabs (v9)

11. Lessons Learned & Tips

Working with Claude Cowork

• Front-load your requirements. The consolidated prompt in Section 3 represents everything we learned over 9 versions. Give Claude all your project management requirements upfront and you'll save significant back-and-forth.
• Claude is conservative by default. It won't create documentation tabs, set up folder structures, or implement version control unless you ask. See Section 4 for the full list.
• Review Claude's work at each step. Claude shows you what it's doing and asks for decisions on ambiguous items. Take the time to review — this is where your institutional knowledge matters.
• Use screenshots to share what you find in Zeffy. We discovered Subscription Status and Lists features by exploring Zeffy's import interface. Sharing screenshots with Claude let it immediately adapt the files.
• The conversation context has limits. For very long projects, Claude's context window may fill up and the session may need to continue in a new conversation. When this happens, Claude can read a summary of the prior conversation and continue.
• Check the Minors tab (or any separated data). When Claude creates subset tabs, verify they have all the columns from the original. Our Minors tab initially only had 8 columns instead of 26.

Data Cleaning Tips

• mailto: corruption is common. If your Excel has been used to send emails, email cells often get corrupted with mailto: prefixes. Always check for this.
• Commas in names break CSVs. Names like "Smith, Jr." will split incorrectly when imported as CSV. Remove commas but keep the suffix.
• Empty dates need a strategy. Decide upfront what to use as a fallback date. We used the current date, but you might prefer the member's join date or a specific default.
• Shared email addresses are a real problem. CRM systems use email as the unique identifier. If spouses share an email, you need a deduplication strategy before import.
• Don't delete rows — clear them. We learned the hard way that deleting rows in Excel programmatically can corrupt data. Clearing cell values is safer.

Zeffy-Specific Tips

• Payment import creates contacts. You don't need a separate contact import to get people into Zeffy. The payment import does both. Use the contact import only for adding Lists and newsletter preferences.
• Email is the primary key. Everything in Zeffy revolves around the email address. Duplicate emails will be merged. Make sure emails are unique and correct before importing.
• Tags are being replaced by Lists. As of early 2026, Zeffy's Tags feature is transitioning to Lists. Use Lists for segmentation.
• Lists use semicolons in the CSV. When assigning contacts to multiple lists, separate list names with semicolons (e.g., "Class A;All Members").
• Subscription Status controls newsletter opt-out. Setting it to "unsubscribed" opts the contact out of newsletters while still allowing transactional emails.
• Import one file at a time. Even if your files are small, importing one class at a time makes troubleshooting easier.

12. Appendix: Zeffy Import Field Reference

Payment Import Template

Contact Import Template