Successful paid marketing campaigns rely on conversion tracking in ad platforms and reporting systems. Tracking not only provides a feedback loop to marketers to demonstrate which campaigns are most effective, this data is also essential for the machine learning algorithms powering ad platforms to learn and serve more relevant ads to users.
However, conversion tracking is notoriously hard to implement correctly.
Because most conversions are configured client-side (as opposed to server-side), conversions can behave unpredictably when changes are made to HTML markup or user experience. And if you’re responsible for monitoring the health of conversions across multiple campaigns and accounts, it’s nearly impossible to do this manually.
The negative impact of broken conversion tracking can’t be understated.
It can range from marketers having low confidence in campaign performance reports, to Google Ads proactively pausing bidding for placements if the algorithms determine that campaigns are failing to convert. That means wasted ad budgets and time.
Identifying issues with conversion tracking early—before they can impact campaign
Get To Issues Early
At Outshine, we’ve developed systems to ETL data from ad accounts and Google Analytics to data warehouses. While the main purpose of these warehouses is to enable accurate reporting, we realized that the conversion data could also be used to generate alerts when issues with conversions arise. In addition to ETL pipelines, we now use a data transformation platform called Dataform to model and maintain data in our warehouses.
Here’s a guide on how to use a key feature of Dataform—assertions—to alert you when issues with conversion tracking arise.
The Method
Assertions in Dataform are SQL queries that check the state of your data based on pre-defined logic. In its simplest form, the example below alerts a user when a specific metric exceeds a threshold (in this case when ad spend exceeds $1,000):
SELECT campaign
,spend
FROM `table`
WHERE spend > 1000
If the query returns any rows (i.e. the WHERE condition has been met), the assertion has failed, and the user will be alerted. Within Dataform, assertions can be scheduled to run daily (or whatever frequency you choose), and users will be notified by email or Slack if the assertion fails.
To monitor conversion tracking, we wrote assertions to alert us if Google Analytics goal completions drop below a certain threshold within any given day. To do so, we use SQL’s date_diff function to report on the previous day’s goal completion counts.
config {
type: "assertion"
}
WITH aggregate_goals_by_day AS (
SELECT
date,
SUM(goal_completion_1) AS content_download,
SUM(goal_completion_2) AS demo_request,
SUM(goal_completion_3) AS webinar
FROM
datasource
WHERE
DATE_DIFF(CURRENT_DATE, date, DAY) = 1
GROUP BY
1
)
SELECT
*
FROM
aggregate_goals_by_day
WHERE
content_download < 1
OR demo_request < 1
OR webinar < 1
The beauty of assertions in Dataform is they are scalable across multiple clients and accounts. Assertions save analysts valuable time because they remove the need to manually monitor metrics within the ad platforms, and push notifications to us only when something has gone wrong.
Assertions can be customized to meet the needs of the client or account and can use whatever thresholds, metrics, or groupings that require monitoring.
Other Use Cases
Monitoring key metrics:
- Conversion rates between marketing funnel stage (lead to MQL, for example)
- Violations in service level agreements
- Ad spend
- Changes in sales pipeline metrics (MRR, new pipeline generated)
Building QA checks into your data modeling & reporting process:
- Detecting duplicates or nulls in data
- Checking key uniqueness
Creating dependencies in the data modeling process:
- Preventing other jobs from running if assertions fail
The Rise Of Push vs Pull Analytics
We believe that the developments in the world of business intelligence in 2019 mean that more and more data teams will be moving to “push” models. Creating systems that push insights to people, rather than relying on data professionals to pull information from technology.
Monitoring conversion tracking using a tool like Dataform is a perfect example of where the modern data team doesn’t passively wait for marketing, product, or management teams to ask for answers to questions. Instead, we can proactively bring solutions to problems a team may not even know they had.
