Uploads

The Uploads page lets you analyse a large batch of customers in one go. Instead of calling the API once per customer, you upload a single CSV file and Trustfull runs the same enrichment on every row.

Uploads is one of the two ways to send Trustfull a file:

• Uploads — customers you want to analyse now. Use it to score a backlog, check a campaign, or try Trustfull on your own data.
• Training Sets — labelled results of lookups you already ran. Use it to measure how well the scores matched what really happened.

In short: use Uploads to get scores and enrich your data, use Training Sets to check how good past scores were.

What Your File Must Look Like

Start from the template, then edit it

In the upload modal, click Download CSV template. The template holds every column Trustfull supports, filled with one example row:

customer_id,phone_number,email,ip,first_name,last_name,domain
your-customer-id,393481122333,[email protected],127.0.0.1,Mario,Rossi,example.com

Delete the columns you do not need, then replace the example row with your own data.

Columns

Every file needs two things.

The first is the customer_id column. The second is at least one signal column.

The customer_id is the name you give to the request. A signal is the data you want Trustfull to check. A phone number is a signal. An email address is a signal.

A row with a customer_id and no signal gives Trustfull nothing to check. Trustfull skips that row.

You must also include at least one signal column. Any other column name stops the upload.

What Trustfull analyses

Trustfull works out which products to run from the columns you include:

Three rules to remember:

• You cannot upload names on their own. If you include first_name and last_name, add at least one other signal as well.
• Keep domains in their own file. A file with a domain column can only contain customer_id and domain. Any other column will be rejected.
• Two or more signals become an Onboarding. Trustfull returns a score for each product, and an Onboarding score for the whole identity.

Examples

Every file starts from the same template. What changes is which columns you keep.

Onboarding

Keep two or more signals and delete domain. Trustfull returns a score for each product it runs, plus an Onboarding score for the whole identity. Any combination works, as long as first_name and last_name travel together.

customer_id,phone_number,email,ip,first_name,last_name
cust-001,393481122333,[email protected],127.0.0.1,Mario,Rossi
cust-002,393482233444,[email protected],127.0.0.2,Anna,Bianchi

customer_id,phone_number,email
cust-001,393481122333,[email protected]
cust-002,393482233444,[email protected]

customer_id,phone_number,first_name,last_name
cust-001,393481122333,Mario,Rossi
cust-002,393482233444,Anna,Bianchi

customer_id,email,ip
cust-001,[email protected],127.0.0.1
cust-002,[email protected],127.0.0.2

Single product

Keep customer_id and one signal, then delete every other column. You get that product's score only, with no Onboarding score for the identity. Names cannot be used on their own, so there is no name-only file.

customer_id,phone_number
cust-001,393481122333
cust-002,393482233444

customer_id,email
cust-001,[email protected]
cust-002,[email protected]

customer_id,ip
cust-001,127.0.0.1
cust-002,127.0.0.2

Domain

Keep customer_id and domain, then delete every other column. Domains cannot be combined with customer signals, so they always need a separate file.

customer_id,domain
cust-001,example.com
cust-002,another-example.com

Using semicolons

Any file above also works with semicolons, as long as every line uses the same separator.

customer_id;email
cust-001;[email protected]
cust-002;[email protected]

How to Upload a File

Uploading takes three steps: fill in the modal, wait for the check, then confirm.

1. Open the upload modal

Click Upload on the CSV Upload page. Select the App key the analysis runs against, give the file a name, then drop your CSV in the upload area.

2. Wait for the check

Trustfull reads the whole file before it enriches anything, so you learn about problems before you spend credit. The upload waits in Validating while this runs.

3. Start the import

When the file passes, a Validation completed banner shows the record count and the Upload button becomes active. Click it to start the run. Rows are enriched in the background, and the counters update as they complete.

What Trustfull Checks

The check runs before anything is enriched, so you do not start an upload with data that cannot be processed. A misspelled column, a broken phone number, or a customer_id used twice would otherwise cost you credit and give you nothing back for those rows. Trustfull finds these problems first, so you can fix the file and upload it again.

Trustfull reads every row, not a sample. The check happens in four passes.

The header

Trustfull confirms you used the same separator everywhere, that every column name is one it supports, and that customer_id is there. It also confirms a domain column is not sitting next to other signals, and that names are not the only signal in the file.

The columns

Trustfull looks down each column. If a column is named in the header but every row under it is empty, Trustfull tells you, so you can remove it.

Every row

Trustfull confirms customer_id is present, is not used twice, and is not longer than 256 characters. It checks that the row contains at least one signal. It then reads each phone number, email address, IP address and domain to check it is valid, and confirms no value is longer than 200 characters. It also looks at names, but a name problem never stops the row.

Your balance

Before the run starts Trustfull works out what the file costs: the price of the products it detected, multiplied by the number of rows. If your balance does not cover it, the upload stops here.

The report groups problems by type and shows up to 100 examples, with a maximum of 5 per type. If there are more, the report says so.

Main Errors

A problem can stop the whole upload, skip a single row, or just be flagged as a warning.

Problems that stop the upload

Nothing is imported. Fix the file and upload it again.

Problems that skip one row

The rest of the file is imported as normal. Only the row with the problem is left out.

Warnings

The row is imported and analysed as normal. A warning tells you the data is poor quality, so the result for that row is less reliable.

Upload Statuses

FAQ

Related

• Training Sets — the other upload type, for measuring model accuracy.