Using AI to Generate Accurate Monthly Executive Reports from Godcaster Data

AI assistants like ChatGPT, Gemini, and Grok can be powerful analysis partners — when properly structured.

Because Godcaster exports include structural nuances (attribution fields, action types, device indicators, session reconstruction limits, timestamp precision, and play-counting rules), how you prompt the AI directly affects accuracy.

To produce consistent, board-ready reports, Godcaster recommends a Two-Step Reporting Method:

• Step 1: Data Audit & KPI Computation
• Step 2: Executive Summary Narrative

Separating computation from narrative improves reliability across AI tools.

How to Use This

• Export your monthly performance CSV from the Godcaster dashboard
• Open your AI assistant (ChatGPT, Gemini, etc.)
• Upload the CSV
• Copy Step 1 below and run it
• After results are generated, copy Step 2 below and run it
• Do not shorten the prompts, as the structure is what ensures accuracy

STEP 1 — DATA AUDIT & COMPUTATION (START COPY)

You are analyzing a monthly performance export (CSV) from our Godcaster digital station.

Before summarizing, first perform a Data Audit:

List all column headers exactly as shown
• Trim/normalize header spacing if needed
• List all distinct values in the Action column
• Confirm whether Episode and Program identifiers exist on:
• Plays (play, resume)
• Share events
• Support events
• Insert events
• Identify any structural characteristics of the data (e.g., station-level vs program-level activity)

Rules:

Analyze only the data provided
• Do not fabricate missing attribution
• Clearly state limitations
• If session IDs are absent, reconstruct sessions by grouping events by IP + UserAgent chronologically and break sessions after 30 minutes of inactivity
• If timestamps lack minute/second precision, state session reconstruction limitations
• If insert events are not present, explicitly state that no insert-click rows exist

Now compute:

PLAYS (Listening Volume)
• Total plays (play)
• Total resumes (resume)
• Total plays + resumes
• Note: Plays are de-duplicated to once per listener per hour (do not attempt further deduplication)

ENGAGEMENT (All Interaction — Developer Definition)
• Total engagement actions (all rows / all Action values)

Also break engagement into categories:
• Listening: play, resume
• Browsing/Navigation: body:, search:, rss
• Response/Action: share, support:*, insert
• Ownership/Retention: pwa, follow
• System/Load: init

SUPPORT ACTIVITY
• Total support clicks (all support:*)
• Breakdown: support:program vs support:player
• Rank top programs by support clicks (if attribution exists)

INSERT PERFORMANCE
• Total insert clicks (if present)
• Insert click rate = insert clicks ÷ (plays + resumes)
• Rank top programs and/or episodes by insert clicks (if identifiers exist)

CONTENT PERFORMANCE
• Program-level rankings (top 10 by plays + resumes)
• Episode-level rankings (if identifiers exist)
• Top programs by:
• Engagement
• Shares
• Support clicks
• Insert clicks (if available)

DEVICE & PLATFORM
• Device breakdown (Mobile / Desktop / Tablet)
• Platform behavior if available

BEHAVIOR FLOW
• Live → On-Demand transitions (if reconstructable)
• Session patterns (if reconstructable)

AUDIENCE ESTIMATES
• Unique IP count (low estimate)
• Unique IP + UserAgent combinations (high estimate)

Show calculation logic for each KPI.
Do not write executive narrative yet.

END STEP 1 COPY

STEP 2 — EXECUTIVE SUMMARY (START COPY)

Using ONLY the verified metrics generated above:

Create a concise, board-level executive summary for [Station Name] for [Month, Year].

Use these definitions consistently:

• Plays = play + resume (deduplicated once per listener per hour)
• Engagement = all listener interactions (all Action values)

Include:

AUDIENCE REACH
• Estimated listener range (low and high)
• Overall listening activity level

LISTENING PERFORMANCE
• Total plays
• Total resumes
• Total plays + resumes
• Top-performing programs
• Top-performing episodes (if reliable)

ENGAGEMENT OVERVIEW
• Total engagement actions (all interactions)

Clearly break out:
• Listening activity (plays + resumes)
• Browsing/navigation activity
• Response actions (shares, support, insert)
• Ownership/retention actions (pwa, follow)

SUPPORT / RESPONSE ACTIVITY
• Total support clicks
• Breakdown if relevant (support:program = program-level, support:player / support:player_rss = station-level)
• Where support activity is concentrated
• What this indicates about listener response

INSERT PERFORMANCE
• Total insert clicks (if present)
• Insert effectiveness relative to listening (insert clicks vs plays + resumes)
• Where inserts are performing best

• If no insert activity exists, state:
“No insert interaction was recorded in this reporting period.”

OWNERSHIP SIGNALS
• PWA activity (app install / owned listening behavior)
• Follow activity (if present)
• What this suggests about deeper listener commitment

DEVICE USAGE
• Mobile / Desktop / Tablet distribution
• Any notable platform behavior patterns

LISTENER BEHAVIOR
• Live → On-Demand patterns (if available)
• Any meaningful listener journey insights supported by the data

STRATEGIC OPPORTUNITIES (2–3)
Provide grounded recommendations based on the data, such as:
• Increasing conversion from listening to response
• Expanding insert usage on high-performing programs
• Improving navigation toward high-engagement content
• Strengthening calls-to-action where engagement is already strong

DATA LIMITATIONS
Clearly explain:
• Whether actions occur at the program level (specific show) or station level (player)
• Differences between support types (program vs player)
• Lack of session IDs or reconstruction constraints
• Timestamp precision limitations
• Any constraints on interpreting listener behavior

Rules:

• Do not introduce new calculations
• Do not overstate conclusions beyond what the data supports
• Clearly distinguish between:
• Listening (plays)
• Total engagement (all interaction)
• Response actions (support, insert, share)

END STEP 2 COPY

Note on Privacy

AI assistants analyze only the CSV file you upload. They do not access your Godcaster dashboard, identify individual listeners, or track personal identities. All insights are based solely on aggregated export data.