Conversion in LinkedIn fell after a redesign. How would you investigate

Problem Statement Description

Product context: Microsoft is a productivity, software, AI, gaming, and cloud company; its products include Windows, Microsoft 365, Teams, LinkedIn, Xbox, Azure, Dynamics, and Copilot. LinkedIn is Microsoft's professional network; its products include profiles, feed, jobs, recruiting, LinkedIn Learning, sales tools, messaging, and ads.

LinkedIn has recently launched a redesigned experience for a key conversion flow used by IT leaders, such as moving from discovery or engagement on LinkedIn to completing a high-value action like signing up, starting a trial, requesting sales contact, submitting a lead form, or adopting an enterprise-oriented product surface. Soon after launch, the team observes that conversion has fallen compared with the pre-redesign baseline.

You are asked to investigate the drop as a root-cause analysis problem. The focus is not to propose a new design immediately, but to determine whether the decline is real, where it is happening in the funnel, which users or surfaces are affected, and what evidence would help isolate the cause.

This is a hard RCA scenario because the redesign may have changed user behavior, tracking, traffic mix, eligibility rules, page performance, or downstream sales/product handoffs. The investigation should be especially mindful of LinkedIn’s professional context, Microsoft’s enterprise trust expectations, and the needs of IT leaders who may be evaluating productivity, cloud, AI, developer, or collaboration-related offerings.

The investigation should consider:

- How to define the conversion event, baseline period, denominator, funnel steps, and expected seasonality.

- Whether the observed drop is statistically meaningful and whether it is visible across all users or concentrated in specific cohorts.

- Segmentation by IT leader profile, company size, geography, device, traffic source, logged-in state, acquisition channel, and product surface.

- Instrumentation checks, including event firing, attribution logic, experiment assignment, tracking changes, and data pipeline delays after the redesign.

- Funnel localization to identify whether the decline occurs at landing, engagement, form start, form completion, authentication, checkout, sales handoff, or confirmation.

- Product and experience hypotheses such as changed information hierarchy, unclear value proposition, increased friction, degraded performance, accessibility issues, or trust/compliance concerns.

- Mitigation and prevention planning, including what evidence would justify rollback, targeted fixes, monitoring, alerting, and post-launch quality gates.

The goal is to demonstrate a structured RCA approach that separates measurement issues from genuine user behavior changes, narrows the problem through segmentation and evidence, and leads the team toward a confident decision on containment, correction, and long-term prevention.