Communicate to Somchai

Step 1: Revisit Somchai's requirements

Open materials/client-email.md again. Somchai asked for four things:

1. Combine the three data sources into one coherent dataset
2. Show which property differences are real and which are noise
3. Account for the fact that these are different markets
4. Tell me what actually drives guest satisfaction

The findings report must address all four. Not in technical terms -- in language that serves a board meeting.

Step 2: Draft the board summary

Direct AI to produce a findings report structured around Somchai's requirements, not around the statistical methods.

Lead with what the board needs to know, not how it was calculated. The board does not need to know you ran a Kruskal-Wallis test. They need to know that the satisfaction differences across properties are real but smaller than they assumed.

Step 3: Review for honest evidence

Read the draft AI produces. Look for two common problems:

First, does it include effect sizes in plain language? "Statistically significant" is not enough. Somchai needs to know the magnitude. "The property you stay at explains about 4% of the variation in guest satisfaction -- season and room type matter more" is honest. "The differences are statistically significant (p < 0.001)" without magnitude is not.

Second, does it communicate the limitations? Every analysis has boundaries. If assumption violations limited confidence in some results, the report should say so. If the data only covers 18-24 months, the seasonal patterns might not generalize to unusual years. Limitations belong in the main body, not in a footnote.

Direct AI to revise if the communication is not honest about uncertainty and practical significance.

Step 4: Translate specific findings

Practice the translation from statistical output to board-ready language:

• "F(4, 3348) = 8.73, p < 0.001, eta-squared = 0.04" becomes "The satisfaction differences across properties are real -- not random noise -- but they are small. The property explains about 4% of the variation in guest satisfaction."
• "Season x Property interaction significant (p = 0.02)" becomes "The property differences depend on the season. Koh Samui leads in winter months when beach tourism peaks, but Chiang Mai catches up during its own high season."
• "Lasso coefficient for rate_per_night = 0.00" becomes "Room price has no measurable effect on satisfaction once you account for room type and property."

The report is the deliverable. Not the notebook. Not the methodology memo. The report is what Somchai presents to the board.

Step 5: Handle the Koh Samui renovation

Koh Samui had a renovation closure for several weeks. If you discovered this in the data and handled it during the analysis, the report should note it: "Koh Samui's averages exclude the renovation period when no guests were present."

If you did not account for it, now is the time to go back and check whether those zero-booking weeks affected any averages or comparisons. This is the kind of detail that Somchai will notice -- he knows about the renovation, and if the numbers do not account for it, the report loses credibility.