Quick start with Retentioneering

Retentioneering makes product analytics very easy once you have the raw data.

Every user action and every visited page or screen from your website or app, all these interactions we call events. To understand deeply how different types of user behavior in your product affects your business metrics, you need to analyze the sequences of events for each user.

1. Load data

To start you can pick from any of two options:

Option 1. To start with our dummy online shop dataset sample.

import retentioneering

# load sample user behavior data as a pandas dataframe:
data = retentioneering.datasets.load_simple_shop()

Here data is a regular Pandas Dataframe with clickstream example:

data.head()

	user_id	event	timestamp
0	219483890	catalog	2019-11-01 17:59:13.273932
1	219483890	product1	2019-11-01 17:59:28.459271
2	219483890	cart	2019-11-01 17:59:29.502214
3	219483890	catalog	2019-11-01 17:59:32.557029
4	964964743	catalog	2019-11-01 21:38:19.283663

As you can see in this fragment of example dataset, user with id 219483890 has 4 events on the website with specific timestamps on 2019-11-01. This is all you need to try out what Retentioneering is about. You are ready to go with this dataset and proceed to step 2.

Option 2. Alternatively, you can start with your dataset.

If you have your raw data of user behavior for example in csv format simply import it as pandas dataframe:

import retentioneering
import pandas as pd

# load your own csv
data = pd.read_csv('yourowndatafile.csv')

How to get a csv file with data? Raw data in the form of {user,event,timestamp} can be streamed via Google Analytics 360 or free Google Analytics App+Web into BigQuery. From the BigQuery console you can run SQL query and export data into csv file, alternatively you can use the Python BigQuery connector to get directly into the dataframe. If you have big datasets, we suggest you take fraction of users in SQL query, filtering by the user id (just add this condition to SQL WHERE statement to get 10% of your users : “and ABS(MOD(FARM_FINGERPRINT(fullVisitorId), 10)) = 0)”.

2. Explore the data

Next step is to simply specify columns names, so that Rete will know how your own data matches the conventional dataset of user_ids, event names, timestamps. This is defined by this global config dictionary which will be used by Rete functions:

# update config to pass columns names:
retentioneering.config.update({
    'user_col': 'user_id',
    'event_col':'event',
    'event_time_col':'timestamp',
})

Now we are ready to explore the user behavior in our data. For example, you can plot graph (read more about plot_graph function here):

data.rete.plot_graph(norm_type='full',
                     weight_col='user_id',
                     thresh=0.06,
                     targets = {'payment_done':'green',
                                'lost':'red'})

Note, that graph is interactive and you can move graph nodes by clicking on it and interactively zoom-in / zoom-out the graph layout.

You can also plot step_matrix (read more about step_matrix function here):

data.rete.step_matrix(max_steps=16,
                      thresh = 0.2,
                      centered={'event':'cart',
                                'left_gap':5,
                                'occurrence':1},
                      targets=['payment_done']);

or you can explore what types of behavior clusters are present in your dataset (read more about exploring behavior clusters here):

data.rete.get_clusters(method='kmeans',
                       n_clusters=8,
                       ngram_range=(1,2),
                       plot_type='cluster_bar',
                       targets=['payment_done','cart']);

Users with similar behavior grouped in the same cluster. Clusters with low conversion rate can represent systematic problem in the product: specific behavior pattern which does not lead to product goals. Obtained user segments can be explored deeper to understand problematic behavior pattern. In the example above for instance, cluster 4 has low conversion rate to purchase but high conversion rate to cart visit.