An ML based mostly strategy to proactive advertiser churn prevention | by Pinterest Engineering | Pinterest Engineering Weblog | Could, 2023

Pinterest Engineering
Pinterest Engineering Blog

Erika Solar ML Engineer | Advertiser Progress Modeling Group; Ogheneovo Dibie Engineering Supervisor | Advertiser Progress Modeling Group

Old, rustic boat sinking in ocean — Photo by Jason Blackeye on Unsplash
Picture by Jason Blackeye on Unsplash

On this weblog put up, we describe a Machine Studying (ML) powered proactive churn prevention answer that was prototyped with our small & medium enterprise (SMB) advertisers. Outcomes from our preliminary experiment counsel that we will detect future churn with a excessive diploma of predictive energy and consequently empower our gross sales companions in mitigating churn. ML-powered proactive churn prevention can obtain higher outcomes than conventional reactive handbook effort.

Like many ads-based companies, at Pinterest, we’re intently centered on minimizing advertiser churn on our platform. Historically, advertiser churn is addressed reactively. Particularly, a gross sales individual reaches out to an advertiser solely after they’ve churned. This strategy is difficult as a result of it’s extremely troublesome to “resurrect” a buyer as soon as they go away the platform. To deal with the challenges with addressing churn reactively, we current a ML-powered proactive strategy to advertiser churn discount. Particularly, we developed a mannequin that may predict the probability of advertiser churn within the close to future and empowered our gross sales group with insights from this mannequin to forestall in danger accounts from churning.

On this weblog, we cowl the:

  • Churn prediction mannequin’s design and implementation
  • Experimentation within the managed North America SMB section

Our group constructed a ML mannequin to foretell advertiser’s churn probability within the subsequent 14 days. We use the Shapely Additive Rationalization (SHAP) bundle to estimate the mannequin’s options’ contribution to the churn prediction. We offer the mannequin churn prediction together with prime contributing options to gross sales. Gross sales makes use of this data to prioritize their effort to mitigate churn for advertisers in danger. We’ll discuss every part in additional element within the following subsections.

Mannequin Structure

The preliminary model of our mannequin is predicated on a snapshot Gradient Boosting Choice Tree (GBDT) structure. We selected GBDT for the next causes:

  1. GBDT is a extensively used mannequin with good efficiency on small to medium sized tabular information* (our information matches on this description).
  2. SHAP works properly with GBDT to estimate options’ contributions.
  3. Mannequin characteristic significance is straightforward to generate with GBDT.
  4. It may possibly additionally function an excellent baseline mannequin for future mannequin enhancements, e.g. a sequential mannequin.

*Snapshot means we use all the data out there as much as a given timestamp to foretell the churn chance within the subsequent 14 days with respect to that timestamp.

Goal Variable

After thorough evaluation and session on the enterprise wants, we determined to make use of the next goal variable definition (see Determine 1).

7/01 to 07/07 is 7 day spend >0. 07/07 to 07/21 is 14 days. 07/21 to 07/27 is 7 day spend >0 ? If yes, then Label 0: active. If no, then Label 1: churn.
Determine 1: Goal Variable Definition

For our use case, we distinguish between an lively and churned advertiser as follows:

  • Energetic advertiser: spent within the final 7 days
  • Churned advertiser: no spend within the final 7 days

We solely predict the churn probability for lively advertisers. Particularly, we predict if they may churn within the subsequent 14 days.


There are over 200 options used within the mannequin. These options are aggregated throughout completely different statistical measures–e.g. min, avg, max and so forth — over a variety of time home windows such because the previous week / month previous to the inference dates. We additionally embody week over week and month over month change options to mirror current developments. These options may be grouped within the following classes:

  • Efficiency: impressions**, clicks, conversions, conversion values, spend, price per 1000 impressions, price per click on, clickthrough charge
  • Objective: objective attainment ratio, distance to objective
  • Funds: funds and utilization
  • Advertisements supervisor actions: creates, edits, archives, customized experiences
  • Property: gross sales channel, nation, trade, tenure, dimension, spend historical past
  • Marketing campaign configuration: focusing on, bid technique, goal kind, marketing campaign finish date

**View greater than 1 second.

Function Contribution

We use the SHAP library to estimate the characteristic contribution to mannequin chance output. Sigmoid of the sum of the options’ SHAP contribution is the same as mannequin chance. From SHAP characteristic contribution, we will know what the important thing drivers are of excessive churn chance. We then spotlight them for the Gross sales group to forestall churn.

We use an offline educated mannequin to deduce lively advertisers’ churn chance every day.

Churn Threat Class

To assist the Gross sales group higher perceive the that means of the mannequin output, we classify accounts into three classes based mostly on their churn chance: excessive, medium, and low churn danger. Excessive churn danger captures the accounts which are principally more likely to churn with excessive precision. Medium churn danger captures the accounts which have a decrease probability of churn. Low churn danger incorporates the ‘wholesome’ accounts which are unlikely to churn within the subsequent 14 days. We choose the thresholds to outline completely different churn danger classes in keeping with the Gross sales group’s request of desired precision and recall. Extra particulars may be present in Experiment End result.

Our first experiment was centered on SMB accounts in North America which are managed by Gross sales Account Managers (AMs). We cut up the advertisers randomly into remedy and management teams throughout the experiment inhabitants. For the management group, we don’t make any adjustments to the present Gross sales group procedures. For the remedy group, we supported the Gross sales group to forestall churn with the next data:

  1. Churn Threat Class: Excessive / medium / low churn danger
  2. Churn Motive Class. We labeled the detailed churn causes into coarse churn classes to ease understanding. The Gross sales group carried out investigations utilizing churn classes as instructions.
14 Day Churn Prediction Model — Overall Churn Risk High. Churn Category is Performance and Campaign Setup / Best Practices. Absolute Change in 14d Churn Risk % D/D is -11% down.
Determine 2: Churn Info Widget

Experiment Success Metrics

Our experiment was evaluated based mostly on the next standards:

  1. Mannequin predictive energy, i.e. how properly our mannequin is ready to establish advertisers which are more likely to churn
  2. Efficacy of churn prediction in churn discount

Mannequin Predictive Energy

With a purpose to decide the mannequin’s predictive energy, we in contrast its on-line efficiency on the management group (i.e. AMs who didn’t have entry to the churn predictions) to what we had noticed offline throughout growth (i.e. our out-of-sample analysis). Particularly, we measured mannequin efficiency based mostly on:

  1. Mannequin high quality: We in contrast the AUC-ROC and AUC-PR noticed on-line to offline.
  2. Churn danger segmentation: In session with gross sales, we decided thresholds for top, medium, and low churn danger classes in order that:
  3. Recall in excessive and medium churn danger ought to be above 70%.
  4. Precision in excessive churn danger ought to be round 70%.

This allows gross sales to seize most accounts vulnerable to churning whereas additionally prioritizing easy methods to work by means of them, i.e. excessive churn danger first (highest precision).

With respect to mannequin high quality, our outcomes point out that the AUC-ROC noticed on-line is inside 1% of the offline AUC-ROC and the net AUC-PR is inside 3% of the offline AUC-PR. This means that the mannequin’s predictive energy in figuring out at-risk accounts is similar to what we noticed offline.

When it comes to churn danger segmentation, our mannequin’s precision, recall, and proportion of the inhabitants captured throughout the excessive and medium danger churn classes have been constantly inside 2–3% of our offline analysis. This means that the segmentation of account danger based mostly on churn probability have been in step with our offline analysis and gross sales expectations.

Efficacy of Churn Prediction in Advertiser Churn Discount

We noticed a 24% (statistically important) discount within the churn charge of excessive tier pods*** in our experiment remedy group in comparison with the management. This means that accounts whose churn dangers have been uncovered to AMs have been much less more likely to churn than those who weren’t.

*** In excessive tier pods, AMs handle about 50–70 accounts on common.

On this weblog put up, we illustrated the event and implementation of an ML-based answer for proactive churn prevention at Pinterest. We’re additionally actively investigating sequential mannequin architectures reminiscent of Lengthy short-term reminiscence (LSTM) and Transformers, which can higher seize the utilization behaviors of advertisers and reduce the necessity for handbook characteristic engineering reminiscent of week-over-week or month-over-month characteristic aggregation utilized in our present mannequin.

Advertiser Progress Modeling Group

  • Engineering: Erika Solar, Ogheneovo Dibie, Keshava Subramanya, Mao Ye
  • Product: Shailini Pandya
  • Product Analytics/Knowledge Science: Alex Simons

Gross sales Group

  • Product: Wesley Kwiecien, Grace Yun
  • Gross sales Managers: Abby (Fromm) Lubarsky

Salesforce Group

  • Engineering: Gayathri Varadarangan (She Her), Murthy Tumuluri, Phani Chimata, Gabriela Mihaila, Richard Wu

Optimization Workbench Group

  • Engineering: Phil Value, Jordan Boaz, Lucilla Chalmer
  • Product: Dan Marantz

[1] When and Why Tree-Based Models (Often) Outperform Neural Networks | by Andre Ye | Towards Data Science

To study extra about engineering at Pinterest, take a look at the remainder of our Engineering Weblog and go to our Pinterest Labs web site. To discover life at Pinterest, go to our Careers web page.