Migration: CSV import, field mapping, Smart Merge
Move customer data in with CSV import, field mapping and Smart Merge.
Read the lesson
Goal: Move the client's customer list into ServiceAgent cleanly, with the right fields, no new duplicates and no consent mistakes.
You need: The client's export file. Custom fields built first (CRM course, lesson 2). A spreadsheet tool.
Why this matters
The customer list is the client's most valuable data. It holds every past job, every address and every phone number. Win-back campaigns, review requests and maintenance reminders all start from it.
A bad import is hard to undo. There is no contact merge on the contact menu. So duplicates made by an import stay until someone fixes them one by one. Wrong field mapping can put phone numbers in the email column for 2,000 people.
The fix is simple. Clean the file first, then test a small file, then import the rest.
The limits you work within
The import takes CSV, TSV or TXT files. Each file can be up to 10MB and 10,000 rows. Commas, semicolons, tabs and pipes all work as separators.
Larger lists go in more than one file. Split the file by row count, and keep the header row on each part.
The CSV import and its limits
Upload a CSV file with contact information
You pick one of two duplicate modes. Simple Import creates new contacts only and skips rows that match. Use it on an empty account or a new list. Smart Merge merges obvious duplicates and creates review tasks for unclear matches. Use it when contacts already exist from calls, forms or an earlier import.
No undo option was found for an import. Treat every full import as final.
Simple Import vs Smart Merge
- Creates new contacts only
- Skips rows that match existing contacts
- Use it on an empty account or a new list
- Merges obvious duplicates
- Creates review tasks for unclear matches
- Use it when contacts already exist
What to bring over, and what to leave
Bring over what the business will use in the next year. That is usually name, phones, email, service address, tags for customer type, last service date and equipment facts. Each one should map to a field that someone will read.
Leave behind what nobody uses. Old internal codes, blank columns and notes like "see paper file" add noise. If the owner wants a full archive, keep the original export file in a safe folder. It does not need to live in the CRM.
Job history is a special case. ServiceAgent stores contacts, not a past job log from another tool. Keep the last service date and the main equipment facts as fields. Those drive reminders and win-back. The full job history stays in the old system's export.
Size changes the work. An owner-operator may have 300 customers in a phone and a notebook. You may type those into a sheet by hand. A fleet may have 20,000 rows across several exports. A multi-location business may need one file per location. Tag each row with its location, so each shop sees its own customers.
Clean the file before you touch the app
Open the export in a spreadsheet. Work on a copy. Keep the original untouched.
- One row per customer. Old systems often export one row per job. Collapse those into one row per customer. Keep the last job date in its own column.
- One format per column. Phones as 10 digits. Emails in lower case. Dates in one format.
- Split names. First name and last name in their own columns.
- Clear the junk. Remove test records, "Cash customer" rows and staff entries.
- Find duplicates in the sheet. Sort by phone, then by email. Merge the pairs by hand. Fixing them here is far faster than fixing them in the CRM.
- Map to fields that exist. List each column and the ServiceAgent field it goes to. Create any missing custom fields first.
- Mark consent honestly. If the old system has no consent record, do not invent one. Add a tag such as
imported-no-consent.
There is no property record in ServiceAgent. Put the service address in the contact's Address. For a landlord with many properties, see CRM lesson 1.
Get the export
Ask the client for the export early, on the day the contract is signed. Exports from old systems often take days. Some tools only export a few columns unless an admin changes a setting. Some need a support ticket.
Ask for every customer column, not a report. A report may cut off rows or merge fields. Ask for the file as CSV if possible. If the client sends a spreadsheet file, save a copy as CSV before you start.
Check the file the day it arrives. Count the rows. Look at the first and last 20 rows. Check that phone numbers did not turn into scientific notation. Spreadsheet tools sometimes do that to long numbers. If they did, ask for a new export, or fix the column format before you save.
Build it in ServiceAgent
- Build any custom fields the file needs:
Settings→Contacts & Pipeline→Custom fields. - Make a test file of about 20 rows. Pick hard rows on purpose: a long name, a missing email, two phones, an odd address.
- Go to
Contacts→More options→Import from CSV. - Upload the test file.
- Map each column to a field. Check every custom field mapping.
- Pick
Simple ImportorSmart Merge. - Run the import. Open five of the new contacts. Check every field.
- If anything is wrong, fix the file, not the records. Delete the 20 test contacts, and test again.
- When the test is clean, import the full file, or each part of it.
- Check it works. Compare the contact count to the row count. Open 10 random contacts and check them.
- If you used
Smart Merge, work through the review tasks it created.
The safe import order
- 1Build custom fields
- 2Clean the fileSeven steps, on a copy
- 3Import 20 hard test rows
- 4Check five contactsEvery field
- 5Fix the file and retestDelete the test contacts first
- 6Import the full list
- 7Check counts and 10 contacts
After the import, tell the office what changed. Show them where a customer's address, tags and equipment fields now live. Ask them to search for three customers they know well. If anything looks wrong to them, check the file row for that customer.
Consent comes with the data, or it does not
An imported contact with no consent record counts as not opted in to marketing texts. A past job is not a yes to promotions. An existing customer relationship alone is not an opt-in.
So tag those contacts on the way in. Keep them out of marketing texts until they agree. Lesson 8 shows how to build an opt-in path. Appointment texts to a customer who books a new job are a separate case. Check the client's consent setup with their advisor.
Consent comes with the data, or it does not
- Each imported contact
- Does the old system hold a consent record?YesMap it to a field and keep itNoTag imported-no-consent
- No marketing texts until they agreeAppointment texts are a separate case: check with the client's advisor
- Build the opt-in path (lesson 8)
Worked example: Polar Air Services
Polar Air exports 14,600 rows from its old software. The export has one row per job.
Step 1, collapse to customers. The 14,600 job rows belong to 6,350 customers.
Step 2, find duplicates in the sheet. Sorting by phone finds 381 customers listed twice with different spellings. After merging: 6,350 minus 381 = 5,969 customers.
Polar Air: export rows to clean contacts
Step 3, check size. 5,969 rows is under the 10,000-row limit. One file is enough, if it stays under 10MB.
Step 4, test file. You import 20 hard rows. The System Install Year column landed in a text field, not the Number field. You fix the mapping and test again. The second test is clean.
Step 5, full import with Smart Merge, because 212 contacts already came in from calls and forms. Smart Merge joins the obvious matches and creates review tasks for the unclear ones.
Step 6, check. The CRM shows the expected count after merges. Ten random contacts look right. All 5,969 imported rows carry the tag imported-no-consent.
Common mistakes
- Importing job rows as customers. Each customer appears many times. Fix: collapse to one row per customer first.
- Skipping the test file. A mapping error hits thousands of records. Fix: test 20 hard rows first.
- Fixing records instead of the file. It takes days. Fix: correct the file and reimport the test.
- Assuming consent. Fix: tag imported contacts without a consent record, and keep them out of marketing texts.
- Building custom fields after the import. The data has nowhere to land. Fix: build fields first.
Do this now
Take any customer export you can get, or make a 50-row practice file. Clean it with the seven steps. Import a 20-row test into your practice account. Record each mapping error and how you fixed it.
Practice the conversation
- Operator I'll move your customer list over this week.
- Owner Just upload the whole thing. It's all in there.
- Operator I'll test 20 rows first. If something maps wrong, I fix it once.
- Owner How long does that take?
- Operator About an hour. A bad full import can take days to clean.
- Owner Can we text all of them a spring offer after?
- Operator Not the ones with no consent record. We'll build a way for them to opt in.
- Owner They're my customers, though.
- Operator Yes. But a past job is not a yes to promotions. We keep you safe on that.
Quick check
- A cleaning client exports 2,400 customers. Custom fields already exist. What do you do first in the import?
- Import all 2,400, then fix mistakes in the CRM
- Pick `Smart Merge` and import all rows
- Upload a test file of about 20 rows and check the mapping
- Import all rows and check the count
- A client's list has 23,500 clean customer rows. How many import files do you need at least?
- 1
- 2
- 4
- 3
- An old system has no consent records for 900 imported customers. How do you treat them?
- As opted in, because they are past customers
- As not opted in to marketing, tagged on import
- As opted in if they booked in the last year
- As opted in until someone replies STOP
ServiceAgent UniversityLearn to run a service business on autopilot. Free courses, Blueprints and certification for Service Operators.