10

Python Source

Add our Python library to your project to send server-side events representing your audience’s activities. This library saves you the trouble of writing your own code to send events into Customer.io and other cloud-based destinations.

How it works

Our python library helps you record source events from your node-side code. Requests from your python app go to our servers, and we route your data to your destinations.

This library uses an internal queue so that your identify and track calls are non-blocking and fast. It also batches requests and flushes asynchronously to Customer.io’s servers.

Like our other libraries, you can log anonymous activity—track and page events—with an anonymousId. When you identify a person, you can pass the anonymousId and we’ll associate the anonymous activity with the identified person.

Getting Started

1. Go to the tab and click Sources.

2. Click Add Source and pick Python.

3. Give the source a Name and click Complete Setup. The name is simply a friendly name to help you find and recognize your source in Customer.io.

4. Install the python library. If you use a system to manage dependencies, you should pin the library to 1.X to avoid breaking changes when we make updates.

   pip install customerio-cdp-analytics
   

5. Import the library in your app and set your write_key before making any analytics. If you’re in our EU data center, you can also set the endpoint parameter to https://cdp-eu.customer.io.

    from customerio import analytics
   
    analytics.write_key = 'YOUR_WRITE_KEY'
    # If you're in our EU data center
    # analytics.endpoint = 'https://cdp-eu.customer.io'
   

Now you’re ready to make calls to Customer.io!

The default initialization settings are production-ready and will queue individual analytics calls. A separate background thread is responsible for making the requests to Customer.io, so calls to the library won’t block your program’s execution.

If you’re in our EU data center

You’ll need to set the endpoint parameter to our EU URL (https://cdp-eu.customer.io). Note that our EU regional endpoints account for the location of your data in Customer.io; they don’t account for the locations of your sources and destinations.

from customerio import analytics

analytics.write_key = 'YOUR_WRITE_KEY'
analytics.endpoint = 'https://cdp-eu.customer.io'

Development settings

By default, the python library is set to queue and send requests directly to Customer.io. But, while you’re integrating this library, you should enable some settings to help you troubleshoot problems.

• Use analytics.debug to log debugging information to the python logger
• Set an on_error handler to print the response you receive from our API.

def on_error(error, items):
    print("An error occurred:", error)


analytics.debug = True
analytics.on_error = on_error

You can also prevent the library from sending data to Customer.io during testing. This can save you the trouble of cleaning out bogus data later.

analytics.send = False

Identify

The identify method tells Data Pipelines who the current website visitor is, and lets you assign unique traitsA key-value pair that you associate with a person or an object—like a person’s name, the date they were created in your workspace, or a company’s billing date etc. Use attributes to target people and personalize messages. Attributes are analogous to traits in Data Pipelines. to a person.

You should call identify when a user creates an account, logs in, etc. You can also call it again whenever a person’s traits change. We’ve shown a typical call with a traits object, but we’ve listed all the fields available in an identify call below.

You can send an identify call with an anonymousId and/or userId.

• anonymousId only: This assign traits to a person before you know who they are.
• userId only: Identifies a user and sets traits.
• both userId and anonymousId: Associates the data sent in previous anonymous page, track, and identify calls with the person you identify by userId.

analytics.identify('f4ca124298', {
    'email': 'cool.person@example.com',
    'first_name': 'cool',
    'last_name': 'person'
})

In most cases, you’ll simply provide userId, anonymousId, and traits, but we’ve provided the full list of writable-fields below. You’ll find the complete source payload in our API documentation.

Track

The track method tells us about actions people take—the events people perform—on your site. Every track call represents an event. When you send an event with a track call, you can provide associated properties about the event. For more information, see the track method reference.

You should track events that are indicators on your site, or the kinds of things that you want to know that your audience does on your website, like Video Viewed, Item Purchased or Article Bookmarked. In some products—like Customer.io Journeys—you can use events to trigger messages and other downstream actions.

You can send events with an anonymousId or a userId. Calls that you make with an anonymousId are associated with a userId when you identify someone by their userId.

Track calls typically contain an event name—so you know what a user did—and a series of properties, which contain additional information about the event. The examples below show added_to_cart events, containing properties that let you know about the product a person added to their shopping cart.

analytics.track('f4ca124298', 'added_to_cart', {
    'product': "shoes",
    'revenue': 39.95,
    'qty': 1
    'size': 9
})

In most cases, you’ll simply provide an ID for the user, the event, and properties, but we’ve provided the full list of writable-fields below. You’ll find the complete source payload in our API documentation.

Page

The Page method records page views on your website, along with optional extra information about the page a person visited.

If you’re using Customer.io’s client-side JavaScript library in combination with our python library, then the client side JavaScript library already captures page calls for you by default.

But, if you have a single page app or you don’t use our JavaScript client library on your website, you’ll need to send your own page calls.

analytics.page('<user_id>', 'category', 'name', {
    'properties': 'any'
}, {
    #options
    'integrations': {
        #Enable/disable integrations
        #By default, all destinations are enabled
    }
})

analytics.page('<user_id>', 'Retail Page', 'shoes', {
    'url': 'https://example.com/products/showes'
})

In most cases, you’ll simply provide an ID for the user and the page name, but we’ve provided the full list of writable-fields below. You’ll find the complete source payload in our API documentation.

Group

The Group method associates an identified person with a group—like a company, organization, project, online class or any other collective noun you come up with for the same concept. In Customer.io Journeys, we call groups objectsNot to be confused with a JSON object, an object in Customer.io is a non-person entity that you can associate with one or more people—like a company, account, or online course. You can use objects to message people based on changes to their company, account, or course itinerary..

Group calls are useful for destinations where you maintain relationships between people and larger organizations, like in Customer.io! In Customer.io Journeys, you can store groups as objectsNot to be confused with a JSON object, an object in Customer.io is a non-person entity that you can associate with one or more people—like a company, account, or online course. You can use objects to message people based on changes to their company, account, or course itinerary., and trigger campaigns based on a person’s relationship to an object—like an account, online class, and so on.

Find more details about group, including the group payload, in our API spec.

analytics.group('user_id', 'group_id', {
  'name': 'Initech',
  'domain': 'Accounting Software'
})

In most cases, you’ll simply provide an ID for the user, the groupId, and traits, but we’ve provided the full list of writable-fields below. You’ll find the complete source payload in our API documentation.

Alias

The Alias method combines two previously unassociated user identities. Some destinations automatically reconcile profiles with different identifiers based on whether you send anonymousId, userId, or another trait that the destination expects to be unique. But for destinations that don’t, you may need to send alias requests to do this.

In general, you won’t need to use the alias call; we try to handle user identification gracefully, so that you don’t need to merge profiles. But you may need to send alias calls to manage user identities in some destinations.

For example, in Mixpanel it’s used to associate an anonymous user with an identified user once they sign up.

analytics.alias(previous_id, user_id)

Here’s how you might use the alias call. In this case, we start with an anonymous_user and switch to an email address when a person provides their userId.

# the anonymous user does actions under an anonymous ID
analytics.track('92734232-2342423423-973945', 'Anonymous Event')

# the anonymous user signs up and is aliased to their new user ID
analytics.alias('92734232-2342423423-973945', '1234')

# the user is identified
analytics.identify('1234', { 'plan': 'Free' })

# the identified user does actions
analytics.track('1234', 'Identified Action')

Configuration and Library Options

If you want to change the library’s default settings want to send data to multiple sources, you can create your own client(s). Remember that each client runs a separate background thread, so you won’t want to create new clients on every request.

from analytics import Client
Client('YOUR_WRITE_KEY', debug=True, on_error=on_error, send=True,
       max_queue_size=100000, upload_interval=5, upload_size=500, gzip=True)

Selecting Destinations

You can pass an integrations object to alias, group, identify, page and track calls that lets you turn certain destinations on or off. By default all destinations are enabled. Passing false for an integration disables the call to that destination.

You might want to do this for things like alias calls, which aren’t supported by all destinations.

In this case, Customer.io specifies the track to only go to Vero. All: false disables all destinations except the ones you explicitly specify.

analytics.track('user_id', 'Membership Upgraded', integrations={
    'All': False,
    'Mixpanel': True,
    'Google Analytics': False
})

Destination flags are case sensitive and match the destination’s name in the docs.

Backfilling historical data

You can backfill data by adding a timestamp to your calls. This can be helpful if you’ve just switched to Customer.io.

You can only do this for destinations that accept timestamped data—most analytics tools like Mixpanel and Amplitude do. The notable destination that doesn’t support timestamped data is Google Analytics.

import datetime
from dateutil.tz import tzutc

timestamp = datetime.datetime(2538, 10, 17, 0, 0, 0, 0, tzinfo=tzutc())
analytics.track('019mr8mf4r', 'started_class', {
    'class': 'How to Use CDP'
}, timestamp=timestamp)

Timezones in Python

Python’s datetime module supports two types of date and time objects: naive objects without timezone information, and aware objects that include timezones. By default, newly created datetime objects are naive. Make sure that you use timezone aware objects when you import data so that you send timezone information correctly.

We created an aware datetime object in the previous section using the tzinfo argument to the datetime constructor. If you omitted this argument, we would not pass timezone info:

>>> naive = datetime.datetime(2015, 1, 5, 0, 0, 0, 0)
>>> aware = datetime.datetime(2015, 1, 5, 0, 0, 0, 0, tzinfo=tzutc())
>>> naive.isoformat()
'2015-01-05T00:00:00'
>>> aware.isoformat()
'2015-01-05T00:00:00+00:00'

If you have an ISO format timestamp string that contains timezone information, dateutil.parser can create aware datetime objects.

>>> import dateutil.parser
>>> dateutil.parser.parse('2012-10-17T18:58:57.911Z')
datetime.datetime(2012, 10, 17, 18, 58, 57, 911000, tzinfo=tzutc())
>>> dateutil.parser.parse('2016-06-06T01:46:33.939388+00:00')
datetime.datetime(2016, 6, 6, 1, 46, 33, 939388, tzinfo=tzutc())
>>> dateutil.parser.parse('2016-06-06T01:46:33.939388+07:00')
datetime.datetime(2016, 6, 6, 1, 46, 33, 939388, tzinfo=tzoffset(None, 25200))
>>> dateutil.parser.parse('2016-06-06T01:46:33.939388-07:00')
datetime.datetime(2016, 6, 6, 1, 46, 33, 939388, tzinfo=tzoffset(None, -25200))

If you find yourself with a naive object, and know what timezone it should be in, you can also use pytz to create an aware datetime object from the naive one.

>>> import datetime
>>> import pytz
>>> naive = datetime.datetime.now()
>>> aware = pytz.timezone('US/Pacific').localize(naive)
>>> naive.isoformat()
'2016-06-05T21:52:14.499635'
>>> aware.isoformat()
'2016-06-05T21:52:14.499635-07:00'

The pytz documentation contains additional information on timezone usage, and can help you handle edge cases.

Batching

Our libraries are built to support high performance environments. It’s safe to use this library on a web server that serves hundreds of requests per second.

But every method you invoke does not result in an HTTP request. Instead, we queue requests in memory and then flush them in batches, which allows for more efficient operation.

By default, our Python source library flushes:

• every 100 messages (control with upload_size)
• if 0.5 seconds has passed since the last flush (control with upload_interval)

There is a maximum of 500KB per batch request and 32KB per call.

What happens if there are too many messages?

If our python module can’t flush calls faster than it’s receiving them, it’ll simply stop accepting requests. This means your program will never crash because of a backed up analytics queue. The default max_queue_size is 10000.

Flush on demand

You can flush your queue on demand. For example, at the end of your program, you’ll want to flush to make sure there’s nothing left in the queue. Just call the flush method.

analytics.flush()

This method blocks the calling thread until there the message queue is empty. You’ll want to use it as part of your cleanup scripts and avoid using it as part of the request lifecycle.

How do I gzip requests?

You can compress batched requests before you send them to Customer.io by setting the gzip argument when constructing your Client.

from analytics import Client
Client('YOUR_WRITE_KEY', gzip=True)

Detecting errors

You can listen to events on failed flush attempts.

def on_error(error, items):
    print('Failure', error)

analytics.on_error = on_error

Logging

Our library uses the standard python logging module. By default, logging is enabled and set at the WARNING level. If you want more verbose logs, you can set a different log_level:

import logging
logging.getLogger('customerio').setLevel('DEBUG')