---
name: podcast-growth-review
description: Diagnoses weak podcast episode or clip performance and recommends whether to repackage, fix the opening, change the clip route, adjust topic choices, or wait. Uses the show's own results at matching ages and separates reach, clicks, and viewing. Use when someone asks "why did this episode flop", "should we change the thumbnail", "views are down", "people leave early", or "our clips get views but no listeners". For a single A/B test report, use podcast-packaging if installed.
metadata:
  title: "Find out why an episode flopped"
  summary: "See if it was the title, thumbnail, intro, or topic."
  section: "Podcasting"
  authors: "samarth, bob-studio"
  icon: "chart-trend-up"
  color: "purple"
  order: "5"
  status: "published"
---

# Find out why an episode flopped

Help the team choose a next move. Find what may be holding growth back. Show the evidence for it. Suggest a change only when it could help with that problem. Waiting can be the right choice.

## Read the context

Read `SHOW-PROFILE.md` in the show's working folder if it exists. It's a short notes file about the show (audience, look, past titles, and lessons) that the Podcasting skills share. Use its audience, past results, and test lessons. If absent, keep a working note: audience, show promise, format, look, recent titles, and known results. Mark guesses. Do not require another skill or a finished profile.

Find the user's goal. A full episode, a discovery clip, and a clip meant to send listeners to an episode need different checks. Review one episode unless the user asks about a wider pattern.

| Input | Action |
| --- | --- |
| Public link only | Inspect the package and accessible opening. Use public views as limited context. Ask for the smallest private report that would separate likely causes. Do not invent click or retention data. |
| Transcript only | Check the promise and opening. Give content risks as hypotheses. You cannot diagnose reach or viewing from text alone. |
| Screenshot or export | Read labels, date range, filters, and units first. Flag cropped fields. Use only visible values. |
| No material | Ask which episode or clip is in question and what feels wrong. Request its link plus a same-age comparison if available. Give a short data request, not a made-up diagnosis. |

## Get a fair comparison

Use three to ten similar releases from the same show when available. Match format, length, topic type, and age after release. Note the gap in fame if one guest is well known and another is not.

Compare the first seven days with the first seven days. Never compare a new episode with a year's views on an old one. Use the median, meaning the middle result, to reduce the effect of one huge hit. Name the comparison episodes and any weak match.

Check where views came from. In YouTube Studio, the creator dashboard, traffic sources show how people arrived. Browse includes Home and other browsing pages. Suggested means recommendations beside or after videos. Search reflects a typed query. External means a link from outside YouTube.

Note paid ads, posts from the guest, and news events that brought views. Compare the same source where the report allows it. If the mix changed, check each source before blaming the package.

## Read only the numbers needed

Start with reach, clicks, and viewing. Ask for more only when it changes the next move.

| Measure | Meaning | Useful question |
| --- | --- | --- |
| Views and episode age | How many views within a stated time | Is this below a fair comparison? |
| Impressions | Counted appearances of the thumbnail on YouTube | Did it get a chance to be chosen? |
| Click-through rate, or CTR | Share of counted impressions that led to a view | Did the package earn the click in that traffic source? |
| Average view duration | Average time watched under the report's view definition | Did those viewers keep watching? |
| Audience retention | The graph of how much of the audience stays through the video | Where does interest drop? |
| Clip-to-episode evidence | Available linked traffic, starts, or tracked streams | Did the clip serve its chosen goal? |

Do not divide all views by impressions to recreate CTR. Some views have no counted thumbnail impression. Use the platform's reported rate.

For clips, compare like formats and lengths. Use engaged views and "Stayed to watch" where available. Engaged views pass the first seconds; "Stayed to watch" is the share that does so. Check the report's definition and date before comparing with older totals.

## Find the likely limit

| Pattern | Check next | Possible move |
| --- | --- | --- |
| Low impressions, healthy clicks and viewing | Episode age, audience size, competing topics, and guest reach | Wait, support discovery, or test a wider truthful frame |
| Normal impressions, weak CTR in a matched source, healthy viewing | Clarity of the title-thumbnail promise and relevant guest proof | Test a new pair |
| Healthy clicks, sharp early drop | Exact opening, ad placement, audio trouble, and gap between promise and content | Repair the mismatch |
| Good start, later drop | The segment at the drop; repetition, tangent, or skipped section | Improve that edit or future structure |
| More impressions, lower CTR, more useful watch time | Whether the episode reached new viewers | Keep the package if the wider reach is healthy |
| Clip reach grows, episode starts do not | Working link, audience fit, and reason to watch more | Fix the route or select a better-fit moment |

Use this table to form a possible cause. It does not prove that cause. A dip in the graph can mean a skip. A peak can mean a replay or a shared link to that point. Watch the section before naming the reason.

If the opening is suspect, state what the title and thumbnail promise. Quote the first line that begins to meet it. Propose the smallest fix: move a supported line earlier, trim unrelated setup, or narrow the package. Mark any new host script as proposed wording.

For a live episode, separate changes available now from changes for the next upload. Do not recommend deleting and uploading again as a routine intro fix. Check the available editor before promising a change to published footage.

Check whether the topic had a brief news window or serves a lasting need. A lasting episode can grow later. Do not use that possibility to dismiss clear evidence of a mismatch today.

## Choose one next move

Rank possible causes by evidence. Name the strongest other cause. Say how sure you are: high, medium, or low, with a reason. If the facts still point to more than one cause, ask for the one report that would help most.

For a packaging change, supply one supported title-thumbnail pair and a test question. Keep the episode edit fixed for that test. If native testing is available, use its reported watch-time result. A rise after a manual swap is weak causal evidence because the audience may have changed.

For a clip problem, write the exact prompt to watch more or name a better moment to cut. If the data cannot show who went from a clip to an episode, say so. A rise in episode views after a clip is only a clue.

For a topic problem, suggest one better-fit angle for a future episode. Reposition the whole show only when a pattern spans releases. Do not let one slow episode undo its identity.

Give a review date and the evidence that would change the decision. Use the next matched age checkpoint, often seven days for an early review and 28 days for a slower episode. Wait for a running test to finish before reacting to its daily lead.

## Return a decision memo

Lead with: "Next move: [specific action or wait]."

Then give:

1. The comparison set and a compact table of relevant numbers.
2. The likely cause, evidence, strongest alternative, and confidence.
3. One concrete change, or the reason to hold steady.
4. The review date and what would count as a useful result.
5. One dated lesson for the show profile, labeled as a guess or supported result.

For the invented *Better Questions* sleep-tracking episode, suppose day-seven views lag similar episodes. Browse impressions are typical, Browse CTR is 2.1% against a 4.5% median, and viewing is typical for that source. Treat the package as the leading test candidate. The figures support a test; they do not prove the image caused the gap. Check guest and topic differences before raising confidence.

If installed, use `podcast-packaging` for a full creative or test pass, `podcast-clips` for a new batch plan, or `podcast-show-positioning` for a repeated audience mismatch. Hand over the relevant evidence and one question. Do not send the user through every sibling skill.