Response rate for LinkedIn messages are down by 5%. What will you do?

Trying to understand what might have caused this 5% drop - Where has the response dropped ? - Any new features we shipped recently - Which demography, user profile/segment have we experienced this drop against - Is there any definition change in how we track response rate - Has there been a surge in the incoming side of messages? - Inaccurate recommendations based on user activity Investigations & Solutions Q - Where has the response dropped ? A - In comments section, post reactions , DM's , conversations b/w recruiter and potential candidate have witnessed a drop etc. Q - Any new features we shipped recently A - For instance we added a new feature eg. LinkedIn Shorts due to which users' attention got drifted over there and they started spending more time over there instead of replying to messages, users' might still be consuming the content but not reacting to posts etc. Q - Which demography, user profile/segment have we experienced this drop against A - I'd try to understand which bucket has reduced response rate across Posts, Comments, Reaction thru data User profile : Inactive premium members , Active premium member , Active member , Inactive member After identifying the cluster I'd go deep to understand what might have been the issue, do RCA Q - Is there any definition change in how we track response rate A - If there is a change in definition than there is nothing to worry about :) Q - As there been a surge in the incoming side of messages as compared to period we are measuring? E - For eg. earlier there might be 100 messages sent and 20 were responded , and now 200 messages are sent and only 30 are responded. That's a different problem then. Q - Inaccurate recommendations based on user activity A - Re evaluate the recommendation model One of the ways I would think of increasing the response rate Recommendation engine should evolve with time : Users should be able to discover slightly new content on their feed, if they see same type of content then they might get bored.

First, I’d check if 5% is actually meaningful or just normal variation. Is this happening consistently over multiple weeks or is it a one-time thing? If it’s noise, there is nothing to fix. If it’s real, I’d think from first principle thinking. Response rate is two-sided metric, it depends on the sender and the recipient equally. The drop could be coming from senders sending worse messages, recipients becoming less willing to reply or the platform changing something in between. Three different problems, three different fixes. I’d check whether the drop was sudden or gradual. Sudden points to a product change, something in inbox ranking or notifications. Gradual points to inbox fatigue or a shift in the type of users sending messages. Then I’d segment by message type, sender persona, and recipient seniority. A 5% aggregate drop rarely tells you where the real problem is. My leading hypothesis is inbox quality degrading. If LinkedIn made outreach easier, message volume went up to faster than people’s willingness to reply. More messages, same replies, rate drops. That’s a platform problem, not a user problem. The fix follows the diagnosis. Sudden drop tied to a product change, we investigate that first. Gradual drop driven by inbox quality, we look at controlling message volume and filtering low-quality outreach.

Response rate shall be defined as for every sent message, what is the reply rate. Now if the response rate is down, it could happen due to multiple reasons. 1. The first one could be a user mix. The behaviour for sending and receiving messages is not the same across all of the users. More entrenched users are likely to send more messages. If the user share on the platform is shifting towards low engaging users the problem might arise 2. What could also break is the quality of messages also. If a lot of the messages are spam or sponsored content, responses will die down further. 3. The third reason could be a break in the funnel also - what if the notifications for messages or auto suggest for response is not working well 4. A part of it could also be a behavioural shift, but that can only happen if the drop is not sudden but gradual.