Amplitude

Renta sends rows from your data warehouse to Amplitude as one of two objects: events and user profiles.

An event row carries the person's User ID, the event name, and any columns you route into event properties or user properties. A user-profile row carries the User ID and the properties written onto that person's Amplitude profile.

Authentication

Renta authenticates with an Amplitude project API key. A key works with a single project, so create a separate destination for every Amplitude project you sync into.

Connecting the destination takes two values from Amplitude:

1. In Amplitude, open Settings > Projects, select the project, and open the API Keys tab. Copy the API key.
2. Note the data region of the project, US or EU. Amplitude runs the two regions as separate installations, and a key belongs to only one of them.

Renta checks the key when you save the destination. The check writes nothing, so connecting a destination leaves no trace in your Amplitude project.

Create a pipeline

Once the destination is connected, create a pipeline to send a data model to Amplitude.

Pipeline configuration

On the final step, choose what Renta writes to Amplitude and which columns fill it.

What Renta sends

Pick the Amplitude object and the data model behind it.

1. What would you like Renta to send to Amplitude.
   Choose the kind of data this pipeline writes.
2. Data model.
   Choose the data model to sync. The columns available for mapping below come from this model.

Field mapping

Connect the columns of your data model to Amplitude's built-in event fields.

1. Mapping rows.
   Pick a warehouse column on the left and the Amplitude field it fills on the right. Required rows are created for you and cannot be removed.
2. Add mapping.
   Add a row for each further field you want to fill.

The dropdown on the right offers Amplitude's built-in fields, listed here by kind.

Two of them are mandatory:

1. User ID is required for both objects.
   It has to come from a column, because a fixed value would write every row of the sync onto a single Amplitude profile, and Amplitude cannot re-map or split a user ID afterwards.
2. Event name is required when the object is Event.
   Map a column that holds the name, or assign it a Static value to send one constant name for the whole sync. The names $identify and $groupidentify are reserved by Amplitude, as is anything starting with [Amplitude]. Renta never substitutes a name of its own, so an event always arrives under the one you set.

Insert ID is Amplitude's deduplication key and rarely needs mapping. When you leave it empty, Renta builds the key from the content of the row, and that is what makes a repeated run safe.

Event and user properties

Amplitude accepts arbitrary property names, so these two blocks work differently from the block above: you type the property name instead of picking it from a list.

1. Event properties.
   Attributes of the event itself, such as the plan or the amount. Pick the column on the left and type the property name Amplitude should store it under.
2. User properties.
   Attributes stored on the person's Amplitude profile, such as the CRM status. On an Event pipeline they travel with the event, so they are queryable in charts straight away.

An Event pipeline offers both blocks. A User profile pipeline offers only user properties, and at least one of them is required: without a property there is nothing to write to the profile.

Sync behavior

Set what happens to deleted rows and name the pipeline.

1. When a source row is deleted.
   Ignore is the only available option. Amplitude ingest is append-only, so an event that has already been sent cannot be retracted.
2. Integration name.
   This is the name of your pipeline. Edit it to match your preferences.

Schedule

Choose how often the pipeline runs.

For Daily schedules, set the Update time (the exact time each run starts), then click Create pipeline.

The new pipeline appears in the Pipelines section under the Reverse ETL tab, where you can track its runs and progress.

Sync strategy

Renta syncs incrementally. Each run compares the data model against the state recorded by the previous run and sends only the rows that are new or changed. The first run therefore delivers the whole model.

Every event carries a deduplication key that Renta derives from the content of the row, unless you mapped Insert ID yourself. Identical content produces an identical key, so a delivery that Renta repeats after a network failure does not become a second event in Amplitude. Changed content produces a new key, so the update always lands.

Run log

To see what a run sent, open the pipeline, click the run in the Runs table, and expand an entry under Run actions. An Amplitude pipeline records an amplitude_partition_delivered action for each batch it delivered, carrying Amplitude's answer for that batch.

The remaining variables describe how the batch was split into requests and where it sat among the run's rows.

The monitoring documentation covers the rest of the run details.

Limits

1. Rows Renta refuses to send.
   Renta drops a row whose User ID is empty, shorter than five characters, or set to a placeholder Amplitude refuses, such as null, none or 0. The rest of the batch still goes through, and the run log reports how many rows were dropped.
2. Runs that stop early.
   A run halts when rejections pass 30% of a single batch or 10% of the run as a whole. This catches a mismapped identity column before it fills Amplitude with unattributable rows. The pipeline resumes from where it stopped on the next run.
3. A key Amplitude refuses.
   If the API key stops working mid-sync, for instance because it was rotated in Amplitude, the run stops at the first refused request instead of sending the rest of the data to a key that cannot accept it. Update the key on the destination, and the next run resumes from where this one stopped.
4. Value length.
   Amplitude truncates text values longer than 1024 characters. Renta sends the value as it stands in the warehouse rather than shortening it for you.
5. Device ID.
   The field is not offered. Amplitude permanently merges any two people that share a device ID, so a repeated or recycled warehouse value would collapse unrelated profiles into one. Amplitude derives the device from User ID instead.