Texts.Analyzer

Method

How to Compare Two Weeks of Text Messages

Two weeks is the smallest window where a chat comparison is honest. Shorter and you are measuring a mood; longer and you cannot remember what else was going on in your life.

Here is how to compare two weeks of messages without drawing conclusions the data cannot support.

Why two weeks, and not two days

A single day is noise. Any two people can have a quiet Tuesday, a frantic Friday or a day of no replies at all. A fortnight smooths out the weekday and weekend split and gives you enough messages for the ratios to settle — roughly 60 to 200 messages for a typical one-to-one thread.

It also has a practical edge: two weeks is short enough to remember what else was happening (work deadline, travel, a cold), which is the context you need to interpret the numbers.

Pick four numbers, and stick to them

Comparing everything at once produces a paragraph of anxiety. Four numbers are enough:

  1. Messages per person — the volume split.
  2. Sessions opened per person — who restarts the thread after silence.
  3. Median reply time per person — with gaps over six hours excluded.
  4. Questions asked per person — the cheapest measure of engagement.

Write them down like this:

Metric This fortnight Last fortnight
Maya: messages 84 96
Sam: messages 61 58
Sessions opened 9 / 3 12 / 2
Median reply 12m / 3h 8m / 41m
Questions 21 / 7 26 / 15

Reading the table

In this example, volume barely moved, but reply time for one person went from 41 minutes to 3 hours and questions fell from 15 to 7. Two of the four numbers moved in the same direction, and the volume did not — which is exactly the pattern worth noticing: the effort is thinning even though the output looks stable.

Contrast that with a fortnight where every number moved a little in different directions. That is noise with a story attached, not a trend.

The three rules for comparing

  • Same length, same channel. Two weeks against two weeks, from the same chat. A month against a week measures the month.
  • Two of four, or it is not a pattern. One metric moving is normal life.
  • Context first. Before reading anything into the table, write what else changed in those two weeks. If a real-world explanation exists, it is probably the answer.

How to run it yourself

Export the chat, paste one fortnight into the analyzer, and write down the four numbers. Then paste the previous fortnight and do the same. Ten minutes for both, and you end up with two rows of a table instead of two hundred remembered screenshots.

The analyzer reports the first four metrics directly. If your export has no timestamps, reply time will be blank — use the WhatsApp .txt export for the timing columns.