Initiation flip.

A thread is made of sessions. A session is a cluster of messages, then a long silence, then the next cluster. The person who sends the first message of a session is the initiator of that session. The initiator is not the person who replied first when both of you were already typing.

Initiation flip is the change in who initiates sessions over a two-week window, measured against her own previous baseline. If she initiated most sessions for a month and now initiates almost none, the flip happened. The flip can go the other way too. A girl who never started a thread for six weeks who now starts three in a week flipped in your favor.

The flip is about sessions, not messages. The math is about who reaches when nothing was forcing them to. The read is about whether her attention to the thread is still pulling her toward it on its own.

Open the thread in your phone. Scroll back fourteen days. Find every cluster of messages where the cluster before it ended more than eight hours earlier. Each of those starts is a session. Most threads have between one and four sessions a day. Count them.

Mark each session with who sent the first message. Her or you. You will end up with two numbers, the count of her starts and the count of your starts, over two weeks. Convert her side to a percentage. That is her initiation rate.

Concrete mock example. Last month: 38 sessions total, 23 of them started by her, 15 by you. Her rate was 60 percent. This two-week window: 18 sessions total, 4 started by her, 14 by you. Her rate is 22 percent. That is an initiation flip. Sixty to twenty-two against the same baseline is not noise, it is a pattern shift.

The flip the other direction looks the same on the math. A girl who was at 12 percent for a month and is now at 45 percent across the last two weeks flipped toward you. Same window, same session math, opposite direction.

There are three reads on a flip toward less initiation from her side. None of them are obvious from the flip alone. You need the other signals to land which one.

Read one. Priority drift. The thread used to be near the top of her attention. Something else moved into that slot. A new job, a class, a friend group, a project, a person. The thread is still allowed. It just is not the first place her attention reaches anymore. Priority drift is usually paired with a small latency drift and warm register. She still answers warmly, she just no longer reaches.

Read two. Social-conditioning correction. She started initiating more than she meant to early, noticed it, and pulled back to a level she is more comfortable with. This is common in the first three weeks of a thread. The flip looks dramatic on the math and is usually not a fade. Register stays warm. Latency stays tight.

Read three. Real loss of interest. The thread dropped far enough down her attention that reaching toward it stopped feeling automatic. This read is the one paired with cooling register, drifting latency, and tilted question asymmetry. Initiation alone does not diagnose it. The combination does.

The pattern alone is not the diagnosis. The pattern plus the other three signals is.

Stop starting threads to compensate. The reflex is to fill the silence with your own initiations because the math felt unbalanced. That is the move that buries the signal. If you start every session for a week, the session math now says zero percent her, one hundred percent you, and you cannot tell whether she stopped reaching because she stopped wanting to or because you never gave her the room.

Hold position for one week. Do not stop replying. Do not punish her. Just stop starting. Reply warmly when she does message. Then look at the same two-week window again. If she initiated something in that week, the flip was a wobble, and you can resume your normal rhythm. If she did not, the flip is real, and you have your read.

Initiation is usually the last of the four to shift. By the time her initiation rate dropped, the latency already drifted and the register already cooled a step. That is why a flip is the loudest single signal of the four. The other three usually arrived first and you missed them. By the time the flip is in the math, the cooling has been in the thread for two weeks.

Paired with latency drift, the flip is a confirmed read. Two of four shifting the same direction is a real signal. Paired with a cooling register, the flip is the same read. Paired with tilted question asymmetry, it is the same read. The combinations are the diagnosis. The flip alone is the question.

The first mistake is reading a flip after two or three sessions. Two sessions is a Tuesday. Three sessions is the back half of a Tuesday. A flip lives in a two-week window because anything shorter is noise. The pattern has to outlast the most plausible boring explanation for it, which is roughly one week of her life being busy.

The second mistake is counting messages instead of sessions. A girl who sends four messages back to one of yours in a single evening did not initiate four times. She initiated zero times. The session math collapses all of that into one session, and the initiator of the session was you. Confusing reply volume with initiation flatters the read and hides the pattern.

The third mistake is reading a flip on a girl who never initiated. If her baseline was 8 percent for the whole life of the thread and she is at 5 percent now, there is no flip. That is her steady state. Flips are measured against her own previous baseline, not against some imagined version of how an interested girl should text. Some interested girls do not start threads. Use her math, not your hopes.