Amazon S3 - How to format and send data to MadKudu via a bucket

You'd like to use product usage or CRM data from a source MadKudu does not currently have an integration with? No worries, we can easily set up a transfer using Amazon S3 from Amazon Redshift or flat files (JSON or CSV). MadKudu's preferred way is to pull data from your S3 bucket where the data is formatted as described below, and from which MadKudu has access through an IAM role.

• Please refer to this documentation to give MadKudu access to your bucket.

• For transfer from Redshift, please refer to this documentation.

Note:

Depending on the volume of data to transfer, it may take from a few hours to a few weeks (>500M records) given the transfer rate limit between Amazon S3 to MadKudu rate limit. We recommend only sending the events necessary to configure MadKudu otherwise your implementation will be delayed until we get the full history of data. Please refer to this documentation or consult our implementation team at success@madkudu.com to understand what are relevant versus irrelevant events to configure your scoring.

Pre-requisites

• You have access to an AWS account to create/manage an S3 bucket

How to format your data

MadKudu works with 3 types of objects:

• Event: what are users doing?

• Contact: who is the user? (coming soon)

• Account: what accounts my users belong to? (coming soon)

Person level events

To send behavioral data (product usage, web activity, marketing activity...), create a file named event with the following attributes (with headers included):

Example in JSON format

{"event_key": "abcd1234", "event_text":"signed up", "event_timestamp":1234567890, "contact_key":"paul@madkudu.com"}
{"event_key": "abcd2345", "event_text":"visit web page", "event_timestamp":1234567890, "contact_key":"paul@madkudu.com", "event_url":"http://www.domain.com/pricing"}

If you send events, please note that MadKudu needs to receive individual events, not aggregations.

Meaning MadKudu needs to receive this:

Instead of this:

Account level events

If you are sending account level events (de-anonymized website visits, 3rd part intent, etc) the same events format applies. To attach events to the respective account, MadKudu uses the domain. The events file needs to contain a ‘fake’ email address anonymous@domain.com as contact_key. See details here.

Example in JSON format

{"event_key": "abcd1234", "event_text":"signed up", "event_timestamp":1234567890, "contact_key":"anonymous@madkudu.com"}
{"event_key": "abcd2345", "event_text":"visit web page", "event_timestamp":1234567890, "contact_key":"anonymous@madkudu.com", "event_url":"http://www.domain.com/pricing"}

Points of attention

How to format the files

MadKudu currently supports two file formats:

• Newline-delimited JSON (preferred)

• CSV

Newline-delimited JSON

Our preferred format for upload is newline-delimited JSON, which is more standardized and less error-prone than CSV.

In this format, the different records are separated by the newline \n character. Each line is a valid JSON object:

{"event_text":"signed up", "event_timestamp":1234567890, "contact_key":"paul@madkudu.com"}
{"event_text":"added a friend", "event_timestamp":1234567890, "contact_key":"paul@madkudu.com", "some_other_event_field":"some_value"}

{"event_text":"signed up", "event_timestamp":1234567890, "contact_key":"abc1234", "key": "val"ue"}

{"event_text":"signed up", "event_timestamp":1234567890, "contact_key":"abc1234", "key": "val\"ue"}

CSV

We also support the .csv format, with the recommended format:

• column names (header) in the first line

• separator: ~ → separate the value with ~ (ex: abc~def~) Please do not use ,  or -as it easily creates parsing issues

• delimiter: "  → this adds quotes around the values (abc -> "abc")

• line separator: line-break \n

abc,cde,ef

"abc","cde","efg"

"abc","cd"e","efg"

"abc","cd""e","efg"

Data validation

JSON line and CSV are relatively easy to corrupt (for example with " or , characters in the data).

We will validate the data on our side and warn you of any corruption issues, but it helps a lot if you follow the format requested above.

Compression

To speed up the data upload part, we highly recommend that you compress your file with GZIP before uploading them to S3.

You can call your file whatever you want it (we recommend event, contact and account). However, please make sure to add the correct extension depending on your file format:

• .json.gz for compressed JSON (recommended)

• .json for uncompressed JSON

• .csv.gz for compressed CSV

• .csv for uncompressed CSV

How to store your file

We recommend that the files you want to share with MadKudu are in a dedicated folder and that you create an IAM policy and role for MadKudu to access these files.

You will also need to set up a recurring push of your data to this folder for MadKudu to score fresh data. This is done by creating distinct files, as described below.

File naming

In the S3 bucket, please upload data into separate folders by date and by objects

{object}/{year}/{month}/{day} where the objects are

• event

• contact

• account

• opportunity

MadKudu will pull the files on the date from the folder name. Files in a folder containing /2020/11/20/ will be pulled on November the 20th, 2020.

If you use the S3 API, simply “prefix” your destination file name. For example, uploading to "contact/2020/11/20/name_of_file.csv" will add a file name name_of_file.csv to the contact folder.

Please use this recommended file naming and storing system in the bucket for MadKudu to be able to automatically pull any new file.

s3://bucket_name/object/year/month/day/name_of_file.csv

Compression

To speed up file transfer, you can compress files locally before transferring them to Amazon S3. If you want to compress your files, please use the GZIP compression method and use .gz or .gzip as your file extension (we currently don’t support other methods or other extensions).

Frequency: setting up a recurring push of data to MadKudu

We pull from your S3 once a day at 00:01am (midnight) UTC. Therefore we recommend you load new files before, like an hour before at 11pm UTC.

When uploading new files please use the recommended naming convention described here File naming.