Role: You are an expert email performance analyst and data storyteller. You are data-savvy, but you communicate in clear, simple language that non-technical stakeholders can easily understand.
I am providing a CSV with subject line performance data from email campaigns [UPLOAD CSV]. Each row represents one email send and includes at least: subject line text, number of recipients, open rate, click rate, and conversion rate.
Please analyze the file as follows:
First, clean the data and remove outlier sends based on the number of recipients. Use the 5th and 95th percentiles of the “Recipients” column and keep only rows within that range, so very small tests and unusually large blasts do not distort the results.
Then, inspect the subject lines and, based on the actual content of this dataset, decide which features are worth analyzing. Do not use a fixed template. Instead, infer useful features from the data itself. These can include, for example:
Length in characters or words
Use of punctuation marks, special characters, digits, emojis, uppercase, numbers or percentages in the copy
Urgency or discount words
“Power words”
Brand-specific phrases
Any other patterns that seem relevant
Then, create the features you consider meaningful and run correlation analysis between these features and the 3 key performance metrics: open rate, click rate, and conversion rate.
From this, identify the most important patterns. Turn these into clear hypotheses about what seems to help or hurt performance. For example:
Whether shorter or longer subject lines work better
Whether including numbers or percentages helps
Whether emojis or exclamation marks are beneficial or risky
Whether certain words or structures are consistently associated with higher engagement
Finally, write the results as a short executive-style summary that a project manager could share with non-technical stakeholders. The output should be insight-focused, not technical. Avoid code, formulas, and statistical jargon.
The final answer should include:
A one- or two-sentence description of the dataset and how outliers were handled
A concise explanation of the main patterns you discovered, grouped around the 3 metrics (opens, clicks, conversions), in plain language
A set of practical recommendations for how to write future subject lines for this brand, directly derived from the patterns you found
Optional: a short note on limitations (e.g., correlations are not causation, or patterns that look suggestive but are weak)
Keep the response compact, readable, and directly useful for decision-making.
