Repeat purchasers from Postgres

Introduction

Planning to use data about your repeat purchasers from Postgres to start making predictions in Faraday? Great!

Since Postgres is natively supported by Faraday, it's easy to onboard your data and use it for predictions by creating a Repeat purchasers cohort.

Cohorts are the building blocks of your predictions in Faraday, and your Repeat purchasers cohort will enable you to predict which customers are most likely to come back for more. Follow the steps below to create your Repeat purchasers cohort.

Getting started

Make sure you have a Faraday account (signup is free!) and that it's not in test mode.

Prerequisites

You'll need data representing one of the your repeat purchasers available in Postgres. This data must take the form of one of the following:

• Each row represents a single transaction event that a person could experience in order to become a repeat purchasers
• Each row represents a fully qualified repeat purchaser

You'll also need the following connections available in your Faraday account:

• Postgres — or create one first

Shortcuts

Your account may already have a cohort representing your repeat purchasers. Before following the instructions below, check your existing cohorts.

Your account may already have a dataset with Repeat purchasers or Transaction data. If you see one listed among your existing datasets, you can skip the "Dataset" section below and proceed directly to the "Cohort" section.

Dataset

Connecting your data

First, you'll add data representing your repeat purchasers from Postgres to Faraday using a Dataset.

• In the navigation sidebar, choose Datasets.
• Click the New Dataset button.
• Fill out the form
  • Choose Postgres and click the Next button. (Don't see Postgres on the list? Make sure you have the connection established—see "Prerequisites" above.)
  • Enter a memorable name, like "Transaction data" or "Repeat purchasers data".
  • Fill in the Postgres options.
  • Click the Create dataset button.
• Wait briefly while Faraday analyzes your data. It shouldn't take long.

Describing your data

Next, you'll help Faraday understand what your data means, starting with recognizing people in your data.

Identities

Your data may contain multiple identities per person, like billing and shipping. Choose one to start with.

• Click the Add identity set button.
• Fill out the form
  • Provide a memorable name for this identity set, like "billing," "shipping," or "customer." You can only use lowercase letters, numbers, and underscores here for technical reasons.
  • For each identity property listed on the left side of the table, use the dropdown on the right to see if there's a column in your data that contains this kind of data for your chosen identity set. (If not, you can leave that dropdown blank.)
• Click the Finish button.

If there are other identity sets in your data, feel free to use the Add identity set button and repeat the process.

Behavior

Now you have to make a decision about your data:

A. Does each row represent a single transaction event that a person could experience in order to become a repeat purchaser?

B. Or, alternatively, does each row represent a fully qualified repeat purchaser?

Option A: each row represents a transaction event

• Click the Add event button.
• Fill out the form:

  • Check the dropdown for an existing event that resembles a transaction. If you find one, choose it.

  • Otherwise, choose Create a new event stream.

  • Click the Next button.

  • If you're creating a new event stream, enter a memorable name, like "transaction." You can only use lowercase letters, numbers, and underscores to name event streams.

  • Under the Timestamp section, use the dropdown on the left to choose a column in your data that contains a "timestamp" for when that row's Transaction event occurred. You may have to use the dropdown on the right to choose the correct format for your timestamp.

  • Skip the Value section.

  • Skip the optional Properties section.

• Click the Finish button.

Option B: each row represents a single repeat purchaser

• Skip the Events section

• Click the Add trait button.

• Fill out the form:

  • In the Trait name textbox, enter a memorable name like "count".
  • Choose the column best representing the number of occurences.

• Click the Finish button. You're done connecting your data to Faraday! Now let's use it to define a Repeat purchasers cohort.

Cohort

Your Repeat purchasers cohort is a formal, fluid representation of your repeat purchasers. You'll use this cohort, along with others, as building blocks to configure Faraday to make powerful predictions about your repeat purchasers and others.

• In the navigation sidebar, click Cohorts.
• Click the New cohort button.

Filtering

Here you will specify the qualifications a person needs to meet to be part of this cohort. If everyone in the dataset should be a member, you can skip this section.

Follow the instructions for option A or option B - whichever one you picked above.

Option A: each row represents a transaction event

• Click the Add event button.

• Fill out the form:

  • Choose the event you created (or selected) in the "Dataset" section above.
  • Click the Next button.
  • Expand the Frequency section.
  • Under Event occurred at least N times, enter 2.

• Click the Finish button.

• Skip the Traits section.

Option B: each row represents a single repeat purchaser

• Skip the Events section.

• Click the Add trait button.

• From the list of traits, choose the count trait you created above (you can filter by "User defined" or the search bar to find it easier).

• Click Next

• Fill out the form:

  • From the dropdown, select Greater than (>). (If this option does not appear, then double check that the column you chose to represent count is an integer.)
  • On the right, enter the value 2.

• Click the Finish button

Finishing touches

• Enter a memorable name, like "Repeat purchasers."
• Click the Save cohort button.

Your new cohort is now ready to use!

Conclusion

With your Repeat purchasers cohort created from a dataset based on Postgres repeat purchasers data, you're ready to use your cohort in a prediction!