How to read a WhatsApp chat without guessing

1. Read the record before you read the relationship

Read a WhatsApp chat by separating what the export records from what it cannot reveal. The file can show who wrote, when messages arrived, how the conversation was divided, and where its timing changed. It cannot show a person’s hidden feelings or explain why a gap happened.

Start with a one-to-one conversation you took part in and are allowed to access. A timestamped export gives timing indicators the information they need; the iPhone WhatsApp export guide walks through that step.

Then use the same order every time:

1. Check whether the sample supports the measure.
2. Read the definition before the result.
3. Compare several indicators instead of isolating one message.
4. Keep the observation separate from your interpretation.

2. Nine indicators, each with a narrow job

Tellune computes nine deterministic indicators from the conversation you confirm. Each answers one limited question.

• Message balance reports each person’s share of confirmed messages and words. An even split is not automatically better.
• Who starts counts the first message after at least eight hours of silence. It needs at least five separate sessions.
• Reply pace takes the median time from one person’s message burst to the other’s reply. Replies after 24 hours are excluded, and each person needs at least four timed replies.
• Follow-ups count a separate second burst sent before the other person replies. The measure needs at least eight bursts and does not treat several quick messages in one burst as several follow-ups. The double-texting guide explains that boundary in detail.
• Long waits count waits of at least 24 hours. The final open message is marked as still open rather than judged.
• Late night is the share of timestamped messages sent from midnight to 5 a.m. in the confirmed timezone.
• Rhythm reports active days, the longest run of active days, and the largest quiet stretch. It needs at least 15 timestamped messages.
• Curiosity counts questions using question marks and common question openers. Detection is approximate, and the measure needs at least 15 messages.
• Depth trend compares average messages per session in the earlier and later halves of the window. It needs at least five sessions.

The names are deliberately plain. “Late night” is a time window. “Depth trend” is session length. Neither is a claim about intimacy.

3. Combine measures without turning them into a score

One number rarely settles the question that brought you to the chat. Use a small cluster of measures that describe different parts of the same pattern.

Question you can check	Measures to read together	Boundary
Who carries the conversation?	Message balance · Who starts · Curiosity	A lopsided split is not a relationship score
How does the timing work?	Reply pace · Long waits · Rhythm	A gap records time, not motive
Are follow-ups changing?	Follow-ups · Who starts · Rhythm	A count cannot explain why someone wrote again
Are sessions changing shape?	Depth trend · Curiosity · Message balance	Longer sessions are not proof of closeness

Suppose message balance is uneven but session starts alternate and both people ask questions. Those are three observations, not a hidden overall score. The right reading is the literal one: volume, initiation, and questions describe different parts of this sample.

That restraint matters. A single result can be accurate while the story attached to it is not.

4. Missing data is a limit, not an invitation to guess

Tellune marks an indicator unavailable when the import does not meet its sample requirement. It does not estimate a missing value.

Use that limit:

1. Confirm timestamps. Timing measures cannot be reconstructed reliably from text that has lost them.
2. Read the sample line. Five sessions can support session-based measures; reply pace needs four timed replies from each person.
3. Check the timezone. Late-night share uses midnight to 5 a.m. in the timezone you confirm.
4. Accept unavailable results. A short conversation may support message balance while leaving reply pace or rhythm unavailable.
5. Compare like with like. A later result is useful only when the same measure is available under compatible definitions.

More text is not automatically better evidence. The relevant question is whether the data supports the measure you want to read.

5. Some details sit outside this method

Punctuation, emoji, read receipts, and one unusually short reply are not Tellune indicators. The product does not have a verified rule that turns those details into a judgment about a person.

You can notice them. You cannot use Tellune’s measurement method to claim they reveal intent, personality, or relationship health. If a detail has no defined measure and boundary, leave it outside the report.

6. Know what happens before you run a report

Privacy depends on the import path and the stage of analysis. WhatsApp exports and pasted text are parsed on the phone for participant mapping and the nine indicators. Screenshot extraction is a separate, best-effort path that temporarily sends compressed images after confirmation. Optional coaching can send the chosen goal, indicator values, and the exact minimized excerpts shown before consent.

The raw WhatsApp export is not sent for those indicator or coaching steps. You can choose a numbers-only path without cloud coaching. The stage-by-stage privacy explainer covers what can leave the phone and when.

Tellune’s preview shows two indicators. Premium unlocks the complete nine-indicator read and Rematch. The iPhone release is still in preparation.

7. The useful next move is another observation

Choose one defined measure that matches your question. Write down its current result and its limitation. If you check the conversation again later, compare the same measure under compatible definitions.

That can show whether the recorded pattern moved. It still cannot tell you what another person feels or what caused the change.