A key metric for Threads spiked unexpectedly. How do you determine if it is healthy

Problem Statement Description

Product context: Meta is a social technology company; its products include Facebook, Instagram, WhatsApp, Messenger, Threads, Quest, creator tools, and ads.

You are the PM for a creator-facing area of Threads at Meta. A key product metric has spiked unexpectedly over the last day or week—for example, creator posts, replies, follows, content shares, or another engagement-related metric. Leadership wants to know whether this spike reflects healthy product growth, a temporary external event, measurement noise, or potentially harmful behavior.

Your task is to walk through how you would investigate the anomaly. The focus is not on naming a single cause immediately, but on structuring the diagnosis: what you would check first, how you would segment the data, how you would validate instrumentation, and how you would decide whether the spike is good for Threads, creators, and the broader user ecosystem.

Consider the Threads context: creator activity can be influenced by network effects, cross-posting from other Meta surfaces, news cycles, viral trends, competitive shifts, recommendation changes, safety issues, spam, or creator monetization incentives. A spike may look positive in aggregate while masking low-quality engagement, abuse, short-lived behavior, or concentration among a small set of accounts.

The experience should consider:

- What the “key metric” means precisely, including numerator, denominator, time window, and expected baseline.

- How to confirm whether the spike is real or caused by logging, dashboarding, data pipeline, experiment, or attribution issues.

- Which cohorts and segments matter, such as creators vs. non-creators, new vs. existing users, geography, platform, traffic source, content type, and account size.

- What adjacent metrics should move with the spike if it is healthy, such as retention, meaningful interactions, follows, content quality, creator satisfaction, or downstream engagement.

- What guardrail metrics could reveal harm, such as spam, reports, blocks, unfollows, low-quality impressions, policy violations, or negative user feedback.

- How to distinguish broad-based durable growth from a one-off event, viral anomaly, bot activity, coordinated behavior, or a small number of outlier creators.

- What evidence would justify taking action, continuing to monitor, rolling back a recent change, escalating to integrity teams, or communicating to leadership.

The goal is to demonstrate a clear RCA approach for an unexpected metric spike: frame the anomaly, validate the data, form hypotheses, gather evidence through segmentation and related metrics, assess product health, and recommend an appropriate next step without jumping to conclusions.