Compliance first: We never take HIPAA/PHI data or personal customer data. Everything is de-identified before it leaves your control.
Home / Guides / Pillar guide · How-to
Pillar guide · How-to

How to Make a Header Sample: Column Names Only, With Examples for Every System

How to Make a Header Sample: Column Names Only, With Examples for Every System

A header sample is a list of the column names in your data export, with no rows and no records. This how-to shows how to make one for email, chat, CRM, tickets, accounting and phone systems, what to leave out, and how to send it so buyers can bid.

If you only do one thing before requesting a bid, make a header sample. It takes minutes, it exposes nothing sensitive, and it answers the first question every data buyer asks: what fields are actually in there?

What a header sample is

A header sample is the first line of a data export: the names of the columns, and nothing else. No customer names, no message text, no amounts, no dates filled in. Just the labels.

In a spreadsheet or CSV file, that first line is called the header row. The CSV format specification (RFC 4180) describes it as an optional first line with the same layout as the records, containing names for each field. A header sample is that line on its own.

For example, this is a header sample:

  • invoice_id, invoice_date, due_date, status, line_count, subtotal, tax, total, terms, paid_date, payment_method, customer_type

This is not a header sample, because it includes a record:

  • invoice_id, invoice_date, ... followed by a second line containing an actual invoice

If any real values appear, delete them before sending.

Why buyers ask for it

Buyers who train machine learning models need to know three things before they can price data: what fields exist, how granular the records are, and how much de-identification the dataset needs. Column names answer all three without anyone seeing a single record.

  • Fields: a ticket export with first_response_minutes and resolution_minutes supports very different models from one with only subject and status.
  • Granularity: line items versus invoice totals, per-message versus per-thread, per-call versus daily summary.
  • Sensitivity: columns like requester_email or phone_number tell everyone up front what must be removed.

According to founder Jeff Cline, bids come back within 72 hours when the data is described correctly and a header sample is provided. Results vary; not a guarantee. Without one, buyers have to guess, and guesses get discounted.

How to export a header sample, step by step

The method is nearly identical across systems:

  1. Run the normal export. Use the system's built-in export (often under Reports, Admin, Settings > Data, or a "Download CSV" button). A small date range is fine; you only need the columns.
  2. Open the file in a spreadsheet application or a plain text editor.
  3. Keep the first row only. Delete every row below the headers. In a spreadsheet, select row 2 through the end and delete.
  4. Check the headers themselves. Occasionally a custom field is named after a person or client (for example, "Notes for Acme"). Rename it to something generic like custom_field_1.
  5. Save as a new file with a clear name, such as zendesk-tickets-headers.csv.
  6. Delete the full export from your desktop if you do not need it. Keep the original data inside its system.

If the system exports JSON or an archive

Some tools export JSON, mailbox archives or ZIP folders. In that case, list the field names (the keys) from one object, then delete the values. For mailbox archives, the useful fields are usually the message headers and metadata, described in the email section below.

Example headers: email

Email exports are usually metadata-first. A buyer wants to know whether you can provide threading, timing and folder structure. An illustrative header sample:

  • message_id, thread_id, mailbox_role, folder, sent_at, received_at, direction, from_domain_type, to_count, cc_count, has_attachment, attachment_types, subject_length, body_length, reply_latency_minutes, labels

Notice that from_domain_type (internal or external) stands in for actual addresses. The bodies themselves are only discussed after a bid, and only in de-identified form. See selling Microsoft 365 data and the email category.

Example headers: team chat

Chat data is valuable because it shows how teams coordinate in real time. An illustrative header sample:

  • message_id, channel_id, channel_type, thread_ts, posted_at, author_role, is_reply, reply_count, reaction_count, has_file, file_type, edited, message_length

Channel names can be revealing (a channel named after a client, for example), so plan to replace them with generic labels. Related pages: Slack and Microsoft Teams.

Example headers: CRM

CRM data shows how deals actually move. Buyers care about stages, timing and outcomes far more than about who the customers are. An illustrative header sample:

  • deal_id, pipeline, stage, stage_entered_at, created_at, closed_at, outcome, amount_band, lead_source, owner_role, activity_count, last_activity_type, days_in_stage, lost_reason, industry, company_size_band

Contact name, email, phone and street address columns are typical in CRM exports. List them if they are there, but mark them for removal; they never travel with the data. See HubSpot and Salesforce.

Example headers: support tickets

Help desk data is among the most requested categories, because it is the raw material for service automation. An illustrative header sample:

  • ticket_id, created_at, first_response_at, solved_at, status, priority, channel, ticket_type, group, assignee_role, tags, reopen_count, reply_count, satisfaction_rating, sla_breached, subject_length

Requester name, requester email and organization name are common columns; flag them for removal. Ticket text can be valuable after de-identification, but it does not belong in the header sample. See Zendesk and Freshdesk.

Example headers: accounting

Accounting exports show the financial rhythm of operations: billing cycles, payment delays, expense categories. An illustrative header sample from a transaction list:

  • txn_id, txn_date, txn_type, account, account_type, class, location, amount, open_balance, due_date, paid_date, terms, memo_length, created_by_role, is_recurring

Vendor and customer names, bank account numbers and tax IDs are excluded. See QuickBooks and Xero, and the accounting category.

Example headers: phones and calls

Call logs show volume, timing, routing and outcomes. An illustrative header sample:

  • call_id, start_time, end_time, duration_seconds, direction, queue, ring_time_seconds, hold_time_seconds, transfer_count, result, agent_role, voicemail, recording_available, after_hours

Caller numbers and names are removed. If recordings or transcripts exist, note that in the request, but never include them in the sample. See RingCentral and Dialpad.

What never goes in a header sample

The rule is simple: column names only. Specifically, do not include:

  • any rows of data, even one "example" row;
  • real names in column headers, such as custom fields named after clients or staff;
  • screenshots of the system with records visible;
  • patient data of any kind, including headers from clinical modules (for practices, only admin-side scheduling and billing exports are relevant);
  • passwords, API keys or login links to your systems;
  • hashed values as a workaround: the FTC has said hashing an identifier does not make data anonymous, so leave values out entirely.

We never take HIPAA/PHI data or personal customer data. Everything is de-identified before it leaves your control. If a header like ssn, date_of_birth or diagnosis_code shows up, that table is out of scope. Read more in de-identified data and PII.

Common mistakes and quick fixes

Most header samples are fine on the first try. When they are not, it is usually one of these:

  • Sending a report instead of an export. A summary report (tickets by month, revenue by quarter) shows totals, not fields. Go back to the raw export so buyers see record-level columns.
  • Leaving one row "for context." It feels helpful, but a single record can contain a name or an email. Delete it; describe the context in a sentence instead.
  • Merging several tables into one file. Tickets, users and organizations are separate tables in most help desks. Send a separate header line for each, labeled clearly.
  • Forgetting custom fields. Many systems hide custom fields from default exports. If your team relies on custom fields (job type, region, product line), make sure they appear, renamed generically if needed.
  • No counts alongside the headers. A header sample says what fields exist; your counts say how much there is. Buyers need both to price the data, so include years, seats and volume with every sample.

A good rule: if you would be comfortable posting the file on a public noticeboard, it is a proper header sample.

Send it and get a bid

With your header samples saved, the rest is quick:

  1. Go to request a bid and list each tool with its years, seats and approximate volume.
  2. Attach or paste the header sample for each system.
  3. Buyers review the description and submit offers; you see the competing bids side by side.
  4. When you accept one, the data is de-identified, licensed, delivered, and you are paid.

In Jeff Cline's experience, small companies have received more than $100,000 for their data, and deals typically close in about 90 days. Results vary; not a guarantee. If you have not inventoried your systems yet, start with the data readiness checklist.

How to create a header sample for a data bid

  1. Run the system's normal export. Use the built-in CSV, spreadsheet or JSON export for one system. A short date range is enough because you only need the column names.
  2. Open the file. Open the export in a spreadsheet application or plain text editor.
  3. Delete every data row. Keep only the first row of column names and delete all rows below it, so no records remain.
  4. Genericize revealing headers. Rename any custom column named after a client, employee or project to a neutral label such as custom_field_1.
  5. Save a headers-only file. Save as a new file named after the system and table, for example zendesk-tickets-headers.csv, and delete the full export from your computer.
  6. Submit with your bid request. Attach the header sample on the request-a-bid form along with the tool name, years of history, seats and volume.
Our compliance promise: We never take HIPAA/PHI data or personal customer data. Everything is de-identified before it leaves your control.

Frequently asked questions

What is a header sample?

A header sample is the list of column names from a data export, with no rows of actual data. It shows buyers what fields exist so they can price the data without seeing any records.

Is it safe to send a header sample?

Yes, as long as it contains only column names. Remove every data row and rename any custom column that includes a client or employee name before sending.

Do I need a header sample for every system?

Ideally yes, one per system or per major table, such as tickets, contacts and deals. Each sample helps buyers price that dataset accurately.

What if my system only exports PDFs or reports?

Look for a CSV, Excel or API export in the admin settings, since PDF reports rarely show the underlying fields. If no export exists, describe the fields you can see on screen and note that in your bid request.

Should personal fields like email or phone appear in my header sample?

List them if they exist so buyers know what will be removed, but never include their values. We never take HIPAA/PHI data or personal customer data. Everything is de-identified before it leaves your control.

How fast will I hear back after sending a header sample?

According to founder Jeff Cline, bids typically come back within 72 hours when the data is described correctly and a header sample is included. Results vary; not a guarantee.

What is the difference between a header sample and a data sample?

A header sample has column names only; a data sample includes actual records. Only the header sample is needed to receive bids, and any later data is de-identified before it leaves your control.

Sources

  1. IETF RFC 4180: Common Format and MIME Type for CSV Files (header line)
  2. FTC Technology Blog: No, hashing still doesn't make your data anonymous (July 2024)

Updated November 19, 2026.

OpenDataPolicies.org connects businesses with buyers of de-identified operational data. We are not a law firm, broker-dealer or investment adviser, and nothing on this site is legal, tax or financial advice. Statements about amounts, timing and outcomes are Jeff Cline's statements based on past transactions. Results vary; nothing here is a guarantee of any bid, price, buyer or closing. Every deal depends on the data, its quality, its rights and buyer demand at the time.