LinkedIn · Metrics · Senior · 45 to 60 minutes

How should PYMK balance growth and equity?

LinkedIn is a professional social network where members build work identities, connect with other professionals, and find jobs and business opportunities. Answer this product management case and get it graded line by line against a weighted rubric.

Metric design
Marketplace thinking
Experimentation
Prioritization
Tradeoff judgment

The situation

You are a senior PM in LinkedIn's network growth team, responsible for People You May Know, the recommendation system that suggests new professional connections. The VP of Growth and the machine learning lead have asked you for a decision memo before the next planning cycle. They are seeing two worrying patterns in PYMK. First, some highly visible members receive so many invitations that their experience may be getting worse, even if senders like seeing those profiles recommended. Second, members who use LinkedIn less often appear underrepresented in recommendations, which could widen existing network gaps and make PYMK work best for people who already have strong networks. The hard part is that PYMK has historically been judged mostly by sender side metrics like invitations sent, and any change that spreads recommendations more broadly could lower that number. Recipient value is real, but harder to measure because one person can receive invitations from many senders. You need to recommend how LinkedIn should define success, what changes to test, and how to evaluate tradeoffs across relevance, engagement, and equity.

What to produce

A 2 page product decision memo.

Your answer must

Define the core product problem as a two sided network issue, and identify at least three member groups that should be measured separately.
Propose a metric hierarchy with one primary success metric, supporting metrics, and guardrails that cover senders, recipients, and distributional equity.
Recommend one or two product or ranking interventions that are feasible within PYMK, and explain the expected first order and second order effects.
Describe an A/B testing plan, including how you would measure recipient side impact when a recipient may get invitations from multiple senders.
State a launch sequence and decision rules for whether to ship, iterate, or stop if the metrics move in mixed directions.

Constraints

Assume only recommendation and ranking changes are in scope this quarter, not redesigns of the invitation inbox or notification system.
You cannot optimize only for senders, because a connection only forms if recipients engage and accept.
Any proposal must be testable with live A/B experiments and must preserve baseline recommendation relevance for the broad member base.

Format

Use five short sections, problem framing, metric tree, proposed changes, experiment plan, and launch decision rules.

How LinkedIn actually played it

LinkedIn changed People You May Know to improve equity in who gets connections and avoid overloading popular members. Experiments cut overloaded recipients by 50 percent, with a 1 percent drop in requests sent but a 1 percent gain in sessions, and raised invitations to infrequent members by 5.44 percent. A strong example of balancing sides of a network and reading mixed metrics.

Optimize for both senders and recipients in network products.
A drop in one metric can be a net win across sessions.
Name the group you want to help and track it directly.

Built from Optimizing People You May Know (PYMK) for equity in network creation by Qiannan Yin and colleagues.