Accepted connections can look like traction while measuring something much weaker: how often users defer the first message. For a networking product, the useful event begins when one person starts a conversation that produces a reply, meeting, introduction or deal.
In Hanoi in 1902, French colonial officials faced a growing rat problem. Their response appeared measurable and practical: pay people for each rat killed.
The authorities eventually accepted a severed tail as proof. That decision made counting easier. It also separated the recorded event from the outcome they wanted.
When the proof becomes the product
People began presenting rat tails while releasing the living animals. A live rat could reproduce and generate more bounty payments later. Michael G. Vann documents the episode in The Great Hanoi Rat Hunt, including reports of people raising rats for the reward.
The programme could count tails precisely. It could not assume that each tail meant one fewer rat in Hanoi.
That is the shape of the problem in the activity log at 4:17 PM. There are 186 accepted connections. The number is real. Each acceptance represents a person who saw another person and allowed a connection.
But what happened next?
If neither person sent a message, the product may have made the emotional task feel complete before the practical task began. The user found someone relevant, clicked a button and received a small confirmation. The uncomfortable part remained untouched: writing a message with enough context to risk being ignored.
The metric rewards the easier action. The customer came for the harder outcome.
Acceptance and conversation are different commitments
Accepting a connection requires little explanation. Sending a message forces a decision.
Why this person? Why now? What do I want from them? What can I offer in return?
Every one of those questions creates friction. Product teams often remove friction by shortening forms, adding suggested contacts and placing a clear button beside each profile. Those changes can increase connection acceptance while leaving the real constraint intact.
The constraint may be uncertainty about what to say.
It may be fear of appearing transactional. A founder in Accra considering a potential customer in Berlin knows that a vague “let’s connect” message has little value. A specific note takes more work because it commits the sender to a reason.
The product has two choices. It can keep celebrating the acceptance, or it can help the user cross the next threshold.
That does not mean adding an AI button that writes a polished paragraph for everyone. Generic messages can increase sending while reducing replies. The new metric would look healthier, yet the conversation would still be missing.
The product needs evidence that one person gave another person a reason to respond.
Measure the chain, not the easiest link
I would start by separating the journey into observable decisions:
- A user identifies someone relevant.
- One person sends a message with a stated reason.
- The recipient replies.
- One of them proposes a concrete next step.
- Both complete that next step.
The first number can still matter. A sudden fall in accepted connections might expose weak recommendations or poor trust signals. It simply should not stand in for the entire job.
Then I would inspect the gap between acceptance and first message. How long do users wait? Do they return to the same profile? Do they draft and abandon messages? Does a prompt tied to the profile produce more replies than a generic template?
A useful experiment could ask the sender to choose a reason before connecting: request advice, discuss a partnership, compare a product problem or offer an introduction. After acceptance, the product can carry that reason into a short editable opening.
The test should track replies and next steps. More messages alone would be another convenient tail count.
This is the same discipline required when interpreting buyer silence. In Kojo’s silent buyer, the absence of a rejection does not create evidence of a sale. Here, the presence of an acceptance does not create evidence of a relationship.
Change what the Tuesday log celebrates
The activity log should make the unfinished work visible.
Instead of showing only “186 accepted connections,” it could show how many conversations started, how many received replies and how many reached a stated next step. It could also surface accepted connections with no message after a reasonable interval, while the context is still fresh.
That changes the founder’s Tuesday decision. The question moves from “How do we get more connections?” to “Where are users stopping before a useful conversation?”
The answer may be a better prompt. It may be stronger matching, clearer intent or fewer suggested contacts. It may also reveal that people enjoy collecting connections more than starting conversations, which would force a harder product decision.
Hanoi’s officials had an exact count of tails. The rats were still alive.
Open the 186 records. Find the point where the human commitment stopped, and run the next product test there.
Comments
No comments yet.