Getting off the ground: Managing entertainment as a product
This is the path I took during the creation of a podcast in one of the most unstable times in history and how we managed to merge creativity and productivity (for the most part) in our earliest stage.
So you want to start a podcast…
Admit it, you wanted to start a podcast at a point in your life, it’s normal, nothing to be ashamed of. Now here’s the thing, my co-producer Paul and I decided to move forward with this (somewhat) misguided impulse. This is where we’re standing in the timeline.
Part 1: Knowing what to know
I’ll tell you what I want, what I really, really want.
Every product is created to solve a problem or satisfy a need. Sometimes you need to create said need “Apple-style” but that takes time and money, we only counted with 50% of those requirements (you can take a guess at what half). Luckily enough we didn’t need to go to such lengths because the need was staring right into our faces.
We start with an idea we kept hearing over and over again:
“I want to listen to a funny improv podcast but I can’t find any”
Strange idea, right? Most people think improv is “supposed” to be funny and yet, they can’t find the fun ones. This seemed interesting enough to deserve further research.
Here’s the abridged deconstruction of the problem based on feedback from within and outside the community (20-ish people sample):
Fantastic. Now we had loads of assumptions to work with. I decided to divide them into these categories:
Feelings:
- I can’t find funny improv podcasts
- Improv podcasts are boring
- Feels like nothing happened
Structure:
- The story doesn’t advance
- The characters don’t change
- The story is interrupted by jokes
At the moment, it seemed like the story itself was more important than how “joky” the podcasts were, that explained the contradiction, it wasn’t the number of jokes in the content but how complex the story was that made the “fun” happen.
But it was all speculation up to that point, I still needed to test those ideas. So it was time to find some popular improv podcasts, put on my headphones, and test them out.
Testing, testing, 1,2,3…
Now here’s the thing, you can’t measure feelings so I moved onto measuring observable data, and just like with literary analysis, the structure is observable. Therefore the theory is:
“Evolving plot, growing characters and uninterrupted flow make for a ‘fun’ podcast”
How do I test this idea, especially when we don’t have any product? We simply needed to observe the people who’re doing something similar, this way I can kill two birds with one stone; I can look for what seems to be working and what doesn’t and I can also keep tabs on our potential direct competitors.
That’s a job for Market Research Man® (a.k.a me)
Let’s look at the competition.
There’s approximately a sextillion improv comedy podcasts, so I decided to sample the most popular.
What you see above is the “nerd line” it will show you where some coding stuff begins and ends, it isn’t anything fancy or particularly advanced but if you don’t care for those subjects, you can safely skip it, you’ll just miss out on the process but not on the results.
How do I find out which podcasts are popular? That happened to be trickier than I initially thought. Comedy podcasts are very popular, but improv specific; not so much. Here’s the approach I took to figure out their popularity.
Situation: As I couldn’t find any hard data on the number of download improv podcasts get, I went to the next best thing, audience opinions. Luckily, there’s more than a few Top X lists on improv podcasts, so I took those as my data set.
Assumption: If any particular podcasts appear in multiple lists, that would show how popular it is.
Simple enough, let’s put it to the test
from pandas import DataFrame
from collections import defaultdict
import re
import numpy as np
import pandas as pd
import collections
import csv#------------ Input and Output file ----------------
filename = r'popular_podcasts.xlsx'
saved_name = 'most_popular.csv'
def find_most_popular(input_name,output_name):
df = pd.read_excel(input_name)
number_of_columns = len(df.columns)
column_names = []
for list_n in df:
column_names.append(list_n)
#-------------- Separates columns into different lists and puts them together into one -----------------
lists_of_names = []
for n in range(number_of_columns):
column = list(df[column_names[n]].str.lower().str.replace('(“|”|\xa0|[^\x00-\x7F]+)','', regex=True).dropna())
lists_of_names.append(column)
#------------- Find Most Popular -------------------
most_popular = {}
for list_n in lists_of_names:
for item in list_n:
if item not in most_popular:
most_popular.update({item:1})
elif item in most_popular:
most_popular[item] +=1
#------------ Create CSV file -------------------
with open(output_name, 'w') as f:
for key in most_popular.keys():
f.write("%s, %s\n" % (key, most_popular[key]))
find_most_popular(filename,saved_name)
I created a web scraping script that helped me gather the data and push it into separate lists, nothing fancy but it did the trick, here’s the data.
My assumption seemed to be right, the result of my algorithm coincided with the podcast’s ratings on Apple Music
And the winners are…
Congratulations to our winners, you’re all way more popular than us (for now).
Why are they popular? We looked at that too, but that’s another story for another time.
But first, it was time to delegate, test our hypotheses, and get our hands dirty. Here’s the plan:
Everyone involved in the production of the podcast would get a number of episodes from the top 3 most popular improv podcasts, and count the appearances of our 3 metrics.
- Advancements in story
- Changes in characters
- Joke interruptions
Assumption: a high number of story and character evolution and a low number of interruptions will coincide with how popular the podcast is.
I decided to pick random early, middle and late episodes (should be a fair sample)
Here’s the data:
Mistakes were made
As you can see, I was wrong…
There is some sort of correlation between my prediction for the result of the metrics for Off Book (OB) and Improv4Humans (I4H), but I was off about Hello From the Magic Tavern (HMT), way off.
This meant there was an indicator I haven’t even thought of, one of those “brack swan” sort of thing audiences don’t usually think about. their appeal wasn’t structural, it was an overarching quality, the “magic” ingredient was:
“Consistency”
OB and I4H created one-of stories in which anything could happen, HMT had recurrent character, settings, and themes, so their quality as improvisers (Every one of them excellent) seemed to be less important than consistency. And improv is, by nature, a chaotic form, so that sort of stability is often overlooked. The familiarity that consistency brings combined with the unpredictability of improv seems to be the key.
So after studying the competition, analyzing audience feedback, and crunching the data, I arrived at the conclusion countless people had arrived at.
“People enjoy a perfect mixture of familiarity and novelty”
It does sound like an awful lot of work to conclude in “common knowledge” but I was happy nonetheless. For two things were proven
- Common wisdom is common for a reason
But most important
- By following data and taking nothing at face value, you will eventually arrive at an answer that reflects reality
So that was it, all we needed to do to be competitive was to:
- Create amazing stories
- Keep the consistency of an improvised product
- Do it all in our own special way, our angle
Now here’s the thing, metrics can only see backward. We still needed to create a unique experience we can call our own, We had a feeling for what was popular and what worked, it was time to get creative
