User Guide

Submitting jobs, downloading results, and managing your account on the ATP Data Services platform.

1. Getting Started

1.1 Requesting access

New customers don't get an account created for them out of the blue – they request one.

  1. Go to the Request Access page (linked from the landing page).
  2. Fill in: Full name, Company name, Email address, Confirm email address, a Username of your choice (3–30 characters: letters, numbers, dots, dashes and underscores, starting with a letter or number), Phone (optional), Message (optional).
  3. There's an invisible honeypot field on this form – a spam-bot defense, not something a real visitor will ever see or need to touch.
  4. Submit. If an account already exists for that email, a request is already pending, or the username you picked is already taken, you're told so directly instead of creating a duplicate.
  5. On success, an email goes to the platform's admin address with your details and a Review this request link – this doesn't go to you yet. You'll hear back once an admin approves it.

1.2 Getting approved

An admin reviews your request from the Approve Customer Request page (reached via the link in the admin-notification email) and either approves or rejects it. On approval:

  • An account is created under the username you chose on the request form.
  • A temporary password is generated.
  • A welcome email is sent to you at the address you gave, with the subject "Your Job Processing Platform Account" and your temporary password.
  • If that email can't be delivered (sending-domain issue), the admin sees the temp password directly on screen and will relay it to you another way.

1.3 First login

  1. Go to the Login page and sign in with the username and temporary password you were given.
  2. You'll be forced to change your password before doing anything else – the account is flagged ForcePasswordChange until you do. New password must be at least 10 characters, including a letter and a number.
  3. After that, you land on My Jobs, your home screen.
Forgot your password instead of just needing to change it on first login? Use the Forgot your password? link on the Login page – see Section 9.

1.4 What you can do from there

Every customer account can submit jobs, track them, and download results. If your account is flagged as a customer admin (see Section 8), you additionally get a Manage Users link to add more logins under your own company account – useful if more than one person at your company needs to submit or review jobs.

2. Navigation Overview

The top nav bar adapts to your current page:

  • My Jobs – your own job history.
  • Submit Job – hidden only while you're already on the Submit page itself.
  • Reports (dropdown) – Account Statement, Data Quality, and Records Processed.
  • Manage Users and Account Settings – customer admins only (see Sections 8 and 6).
  • Change Password – always available in the nav.
  • A Live refresh badge and your user chip (initials + display name) sit on the right, with Sign out at the far right.

3. Submitting a Job

Submitting a job is a 5-step wizard at Submit Job. Steps enable progressively – you can't jump to Step 3 until Step 1 and 2 are satisfied.

  • Step 1 – Job Type. Pick one of the job type tiles (hovering shows a one-line description). Two tiles behave differently: Custom and Data Profiling each kick you out of this wizard entirely into their own standalone flow (see 4.11 and 4.12) rather than continuing through Steps 2-5.
  • Step 2 – Field Type (skipped for Convert, which instead shows Select Conversion Formats here – see 3.2). Pick the type of data in your key/address field (name, email, phone, etc.) – this sets a recommended algorithm and match threshold you can still change later on Step 4.
  • Step 3 – Key Position (titled differently depending on job type – "Address Field Position" or "Address Field Name" for LibPostal, "Key Field Position" for everything else). This is where you tell the platform where the matching/parsing field lives in your file. Exactly what you enter here depends heavily on your file's format – see 3.3 below.
  • Step 4 – Matching Parameters (skipped entirely for non-fuzzy job types, which jump straight from Step 3 to Step 5). Choose:
    • Algorithm: Jaro-Winkler (recommended for names), Levenshtein (recommended for IDs/codes), or Soundex (recommended for phonetic matches, e.g. Smith/Smyth).
    • Match threshold (50-100, default 90): minimum similarity score to count as a match. Higher = stricter, fewer matches. 86+ is recommended for name data.
    • Review threshold (optional): if set, pairs scoring at or above this but below the match threshold land in a separate "review" file instead of being silently called a non-match – e.g. match threshold 90, review threshold 75 catches everything scoring 75-89 for a second look. Leave blank for the traditional strict match/non-match behavior.
    • Blocking Key (optional, its own panel): controls how records are grouped before the fuzzy comparison runs – only records sharing the same blocking key ever get compared against each other. Leave everything blank to block on the key field itself (recommended for almost every job). For two-file jobs (Match/Merge), File B gets its own blocking key fields – leave those blank to reuse File A's.
  • Step 5 – Upload & Submit. Upload your file(s) (a second file field appears automatically for two-file job types), review the summary table, and click Submit job.

3.1 File formats your file can be in

Across job types, the platform accepts four input layouts:

  • Fixed-width – every field sits at a fixed byte position/width, no delimiters. Requires a layout manifest (see 3.4) for most job types; optional for LibPostal.
  • CSV (delimited) – comma or pipe-separated, with or without a header row.
  • JSON – one object per record, or an array of objects (LibPostal and Convert only).
  • XML – one element per record (LibPostal and Convert only).

Which formats are offered depends on job type: LibPostal and Convert support all four; Dedupe and Match/Merge job types support Fixed-width and CSV only. Excel files (.xlsx) can't be uploaded directly – save them as CSV first.

3.2 Convert jobs – a different Step 2

Convert doesn't use the Field Type step at all. Instead Step 2 becomes Select Conversion Formats: pick a Source format and a Target format (they must differ), plus CSV delimiter/header options if CSV is involved on either side. Fixed-width on either end requires a layout manifest.

3.3 What "Key Position" actually asks for, by format

  • Fixed-width: Start position (byte offset, 1 = first character) and Length.
  • CSV: Column number (1 = first column) instead of Start position, and an optional Number of columns if the key spans more than one column (this means whole additional columns, not a character count – leave blank to take the rest of the line from the starting column).
  • JSON/XML with LibPostal: instead of a position, you type the exact JSON key or XML element name holding the address. If the address is split across several fields (street, city, zip as separate fields), use Detect Fields instead – upload a small sample of your file, check off which fields make up the address, and they'll be combined into one string for LibPostal to parse automatically.

Two-file job types (Match/Merge) ask for File B's key position separately, since File B can have a different layout than File A. On those jobs the boxes are labeled File A and File B with the same wording for each (start position, length, column number, number of columns), so you can see which file each value belongs to. The Field name box appears only for JSON/XML address parsing, where it names the JSON key or XML element holding the address.

3.4 Layout manifests

A manifest is a small text file describing your fixed-width file's field layout – required for Fixed-width-involved Convert jobs, optional for LibPostal (you can enter position/length manually instead). Upload it and click Parse; the platform reads back the fields it found. For every job type except Convert, you then pick which parsed field is your key/address field from a dropdown. For Convert, the whole manifest is used as-is to build the fixed-width output – no field selection needed, since the whole file (not one field) is what's being reformatted. A downloadable sample manifest is linked directly on the page.

4. Job Types – Full Reference

Job typeWhat it does
Address Parsing (LibPostal)Parses a free-form or structured address into its components (street, city, state, zip).
File Conversion (Convert)Reformats a file between fixed-width, CSV, JSON, and XML layouts without changing the underlying data.
Dedupe - ExactFinds exact duplicate records within one file and separates unique from duplicate.
Dedupe - Exact (Address)Same as above, with an address pre-pass (via LibPostal) tuning the comparison for address fields.
Dedupe - FuzzyFinds near-duplicate records within one file (typos, spelling variants) using approximate matching.
Dedupe - Fuzzy (Address)Same as above, tuned for address fields via an address pre-pass.
Match/Merge - ExactCompares two files on an exact key and splits records into matched / File A-only / File B-only.
Match/Merge - Exact (Address)Same as above, tuned for address fields via an address pre-pass.
Match/Merge - FuzzyCompares two files using approximate matching and splits records into matched / File A-only / File B-only.
Match/Merge - Fuzzy (Address)Same as above, tuned for address fields via an address pre-pass.
CustomOne-off request handled manually – describe what you need and we'll follow up with a quote.
Data Quality Review (Profile)Free pre-flight check on a sample of your file – completeness, uniqueness, and duplicate estimates before you commit to a real job.

4.1 Address Parsing (LibPostal)

Parses whatever address text you have into clean, separate components. Accepts all four file formats. For JSON/XML you name the address field(s) directly (see 3.3); for Fixed-width/CSV you give a position or column, optionally backed by a manifest. Output includes both the original record and the parsed address, matched back to it (LibPostal Match/Append) – this shows up automatically on the job's results page as a "Parsed address + your original data" download, no extra setup needed.

4.2 File Conversion (Convert)

Pure reformatting – the data itself doesn't change, only the container format. Pick Source and Target format on its own Step 2 panel; Fixed-width on either end needs a manifest.

4.3–4.10 Dedupe and Match/Merge (Exact, Fuzzy, and Address variants)

These eight job types share the same wizard shape:

  • Dedupe finds duplicates within a single file. Match/Merge compares two files against each other and splits the result into matched / File-A-only / File-B-only.
  • Match/Merge jobs also give you one combined file (merged.dat) with every unique record from both files – matched records once, plus the records found in only one file (and any set aside for review) – in addition to the separate matched / File-A-only / File-B-only files.
  • Exact requires the key field to match character-for-character. Fuzzy uses one of the three similarity algorithms (Jaro-Winkler / Levenshtein / Soundex) and a match threshold, and unlocks the Step 4 Matching Parameters panel plus the optional Blocking Key panel. Fuzzy jobs compare every record against every other record within a block, so the blocking key materially affects both runtime and match quality on large files.
  • The (Address) variants of all four base types run the same key/fuzzy logic but add a LibPostal address pre-pass first, so the actual comparison happens against a normalized address rather than the raw, messy input – meaningfully better match quality on address data, priced at a premium over the plain version to reflect that extra processing step.

4.11 Custom

  1. Go to Submit Job, select the Custom tile – this leaves the standard wizard and drops you into its own flow.
  2. Describe what you need in the What do you need done? text box (multi-line, required) – e.g. "Merge 11 files into one deduplicated list, matched on email address."
  3. Submit the request. It goes into our queue as a request for a quote, not an active job yet.
  4. Once we review it and provide a quote, you'll see an Accept quote panel to approve pricing before any processing starts.
  5. After acceptance, upload your data files (as many as you need) and, optionally, a signed SOW/contract file (or email it directly instead) – sending over an SOW or a clear description of scope is what lets an associate come back with an accurate timeline and cost.
  6. Results become available for download once the job is processed and delivered, same as any other job type.

4.12 Data Quality Review (free preflight check)

A free, no-commitment check on a sample of your file before you submit a real (billed) job.

  1. From Submit Job, select the Data Quality Review tile (also reachable directly at Submit Job → Data Profiling) – this also leaves the standard wizard.
  2. Choose your file's format. For CSV, pick the delimiter (comma or pipe) and say whether the first row is a header. For Fixed-width, upload your layout manifest and click Parse first. For JSON or XML, just choose the format – the file is turned into columns automatically (JSON may be nested; XML needs one repeating record element with simple text fields). Column names come from the first record, and a field that is missing or empty in a later record counts as missing data. If your data is in Excel, save it as CSV first (File → Save As → CSV); .xlsx files can't be uploaded directly.
  3. Upload a sample file. Up to the first 10,000 rows are analyzed; if your file is larger, the results say so.
  4. Review the results: for each field, how complete it is, how many of its values are unique, its shortest, longest and average length, and the kind of data it holds (letters, digits or mixed), plus an estimated duplicate count for the whole file. Rows with missing data or duplicates are flagged, and you can download the full summary (profile_summary.txt) and the flagged rows (flagged_rows.csv). This gives you a sense of what a real Dedupe/Match job would find before you pay for one.

This job type always bills at $0.00 – it exists purely to help you decide what real job to run and with what settings.

5. My Jobs and Job Detail

5.1 The My Jobs dashboard

My Jobs is your home screen and refreshes on its own (the Live badge in the top bar). It is organized into panels:

  • Needs Attention – jobs that failed or were cancelled and that you haven't marked as reviewed yet.
  • In Progress and Queued – jobs being processed right now, and jobs waiting their turn.
  • Completed Jobs – finished jobs with records in, records out, matches and duration. Click a row to open its Job Detail page and download results.

Job ID. Every job has a unique Job ID. The dashboard shows the first 8 characters; hover to see the full ID, or click it to copy the full ID to your clipboard. The same ID appears in your job emails, so you can paste it when you contact support and we can find your job right away.

5.2 Job Detail

Click any job to open its Job Detail page:

  • Job Summary – job type, submission date, status.
  • Job Settings – the format/key/algorithm/threshold settings the job actually ran with.
  • What happened? – shown only for failed/cancelled jobs, explaining the outcome.
  • Download Results – appears once a job succeeds, with a count of ready output files and a note on how long they will be kept (see Section 6).

6. Your Files, Retention & Free Reruns

6.1 How long we keep your files

Your uploaded files and your results are kept only as long as you choose. A customer admin sets this under Account Settings (visible to customer admins only – see Section 8). It applies to every job under the account from then on, and it does not bring back files that were already deleted. The choices are:

  • Delete immediately – your files are removed as soon as you have downloaded every result file for the job (see 6.2).
  • 7, 30 or 90 days – your files are removed that many days after the job finishes. Cleanup runs once a day, so removal happens within about a day after that date.

Only the files are removed. Your job history, job settings and invoices stay on your account. Custom jobs are always kept for at least 7 days, even if your setting is shorter. Some accounts have an agreed “keep indefinitely” arrangement instead; if yours does, the job's results page says so.

6.2 Downloading results and “Delete immediately”

The Download Results panel on Job Detail lists every result file for the job. If your account is set to Delete immediately, the job's files (your input and your results) are deleted automatically right after you have downloaded every file in that list. Until then they stay available, so a partial download never loses anything. A few things to know:

  • The extra “Parsed address + your original data” download is a convenience and does not count toward this. Only the files in the main list do.
  • If you never download everything, we still don't keep the files indefinitely: they are removed no later than 14 days after the job finished.
  • Deleted files cannot be recovered. The job page shows a “files have been deleted” message instead of download links. To get results again, resubmit the job – see 6.3.

6.3 Free reruns

If a job came out wrong because of the settings you picked – the wrong match threshold or the wrong key column, for example – you can run it again at no charge. There is nothing to request. When you submit a job, we compare your uploaded file(s) with your earlier jobs, and if the new job qualifies it is automatically not billed. It qualifies when all of the following are true:

  • It is the same job type as an earlier job of yours that finished successfully.
  • The uploaded file is identical. We compare a fingerprint of the file's exact contents, so re-saving it from Excel, adding a row, or changing a delimiter makes it a different file. For two-file jobs (Match/Merge), both files must be identical and in the same File A / File B order.
  • You submit it within 7 days after the original job finished.
  • The original was a normal billed job – not itself a free rerun – and it has not already been rerun for free. A rerun that fails or is cancelled does not use up the free rerun.

You can change any settings on the rerun – that is the point of it. A qualifying rerun is marked “Free rerun” on its Job Detail page and does not appear as a charge on your invoice. It works even if your files were deleted after the original job, because you upload the file again. Each original job gets one free rerun; a further run of the same file is billed normally.

7. Reports

Reached via the Reports dropdown in the nav:

  • Account Statement – your invoice history, with View/Download on each invoice.
  • Data Quality – results from any Data Quality Review (Profile) jobs you've run.
  • Records Processed – the total records you have run through the platform to date, broken down by job type.

7.1 Pricing and invoices

Jobs are priced per record, starting at $3.80 per 1,000 records for the simplest job types, with a $5 minimum charge per job. Larger jobs earn an automatic discount based on the number of records in the job (5% over 10,000 records, up to 25% over 1,000,000). The Data Quality Review is always free.

First job free. The first 1,000 records of your account's first successful job are not billed. It is applied automatically; if that job is larger, the records beyond the first 1,000 bill at the normal rate. The invoice line for that job is marked "first job free" or "first 1,000 records free".

Your invoices are available under Reports → Account Statement, where you can view or download each one.

8. Manage Users (Customer Self-Service)

Only visible, and only usable, if your login carries the IsCustomerAdmin claim – a flag set per user, not per company, so not everyone at your company can manage users even if someone there can.

  1. Go to Manage Users to see everyone currently on your account.
  2. To add someone: fill in Full name, Username, Email, whether they should see every job under your account or only ones they personally submit ("Sees"), and whether they themselves should be a customer admin ("Admin?").
  3. Existing users' "Sees" setting can be changed any time from the same table.

9. Change Password & Forgot Password

Changing a password you remember: reachable from the nav, or directly at /ChangePassword. Enter your current password, then a new one (minimum 10 characters, including at least one letter and one number, entered twice to confirm). On first login with a temporary password, you're forced through this page automatically before you can do anything else.

Forgot your password entirely? Click "Forgot your password?" on the Login page and enter your username or email. If we find a matching account, a password reset link is emailed to the address on file – the link expires after 15 minutes and can only be used once. Click it, choose a new password, and sign in. If you no longer have access to that email address, a multi-user account's admin can reset your password for you from Manage Users instead.

Too many wrong passwords. After five wrong passwords for the same username within 15 minutes, that username is locked for 15 minutes. Wait and try again.

10. Email Notifications Reference

The platform sends the emails below, from the sender name ATP Data Services LLC. If one doesn't arrive, check your spam folder.

TriggerSent toSubjectWhen it fires
New self-service account requestAdmin"New account request: {Name}"Immediately on Request Access submission
Account request approvedCustomer"Your Job Processing Platform Account"On admin approval – includes username + temp password
Password reset (admin-initiated)CustomerReset-specific wordingWhen an admin resets an existing customer's password
Password reset (self-service)Customer"Reset your Job Processing Platform password"When a customer uses Forgot Password on the Login page
Job completedCustomer"Job completed - {JobType}"Job reaches SUCCEEDED, only if that customer has NotifyOnJobComplete=On and an email on file
Job cancelledCustomer"Job cancelled - {JobType}"Admin cancels a job before it finishes
Job failedCustomer"Job failed - {JobType}"Job reaches a failed status
Late payment noticeCustomerInvoice/notice document (PDF or DOCX attachment)Sent when an invoice becomes overdue

Notes: job-completion emails are skipped silently (and logged) if the customer has notifications turned off, has no email on file, or can't be matched back to an AppUser – none of these block the job itself from completing successfully. If the welcome email fails to deliver (e.g. sending-domain not yet verified), the admin sees the temp password directly on the approval screen to relay manually.

11. Glossary

Key field
the field a Dedupe or Match/Merge job compares records on.
Blocking key
the field fuzzy jobs group records by before running comparisons, to avoid comparing every record against every other record in the entire file.
Match threshold
minimum similarity score (0-100) for a fuzzy comparison to count as a match.
Review threshold
an optional second, lower threshold that catches "close but not quite" pairs in a separate review file instead of calling them a non-match outright.
Manifest
a text file describing a fixed-width file's field layout (positions and lengths).
Address pre-pass
running LibPostal Address Parsing on data before the main Dedupe/Match/Merge comparison, used by the "(Address)" job type variants.
Job ID
the unique identifier of a job. Shown in short form on the My Jobs dashboard (click to copy the full ID) and included in your job emails.
Retention period
how long your uploaded files and results are kept after a job finishes (Delete immediately, 7, 30 or 90 days). Set by a customer admin under Account Settings.
Free rerun
a rerun of the identical file, same job type, within 7 days of the original finishing, that is not billed. See Section 6.3.
CustomerAdmin
a per-user flag (not per-company) granting access to Manage Users for that customer account.
No-Charge
an invoice-line override an admin can apply to a job so it doesn't bill the customer, independent of what job type or volume it was.

12. Appendix: Layout Manifest Format

A manifest is a plain-text file, one FIELD= line per field. Lines starting with # are comments. There are three modes, mixable within the same file only insofar as each field declares its own:

  • POSITION (fixed-width files) – FIELD=FieldName,StartPosition,Length (legacy shorthand) or explicitly FIELD=FieldName,POSITION,StartPosition,Length. StartPosition is 1-based (byte 1 = first character).
  • COLUMN (delimited/CSV files) – FIELD=FieldName,COLUMN,ColumnNumber (1-based column index), or FIELD=FieldName,COLUMN,ColumnNumber,Count to bound it to Count whole columns starting there.
  • HEADER (delimited/CSV files with a header row) – FIELD=FieldName,HEADER,HeaderName, resolved to a column position by matching the header row's actual text.

Real sample, fixed-width (the sample file linked from the Submit wizard's manifest upload panel):

# Sample Layout Manifest (fixed-width) # Format: FIELD=FieldName,StartPosition,Length (legacy / POSITION mode) # or: FIELD=FieldName,POSITION,StartPosition,Length (explicit POSITION mode) # Lines starting with # are comments # StartPosition is 1-based (first byte = position 1) # FIELD=AccountID,1,10 FIELD=Surname,11,20 FIELD=GivenName,31,15 FIELD=DateOfBirth,46,10 FIELD=AddressLine1,56,20 FIELD=PostalCode,76,10

Equivalent sample, CSV/delimited (COLUMN mode – illustrative):

# Sample Layout Manifest (CSV/delimited) FIELD=AccountID,COLUMN,1 FIELD=Surname,COLUMN,2 FIELD=GivenName,COLUMN,3 FIELD=DateOfBirth,COLUMN,4 FIELD=AddressLine1,COLUMN,5,3 FIELD=PostalCode,COLUMN,8

Still need help?

Check the FAQ for quick answers, or reach out directly.

Contact Us