Media Theory

Two Businesses: the case for data

I’m a huge fan of Rob Long’s podcast Martini Shot. But I couldn’t disagree more with his most recent one: Two Businesses.

On it Long argues, humorously and rather unconvincingly, that Amazon should sell its studio to CBS because running a successful online retailer and grocery store (Whole Foods), is nothing like running a television and movie studio, because:

“a successful grocery store chain and a successful online retailer need to be masters in margin shaving and supply chain management and on time delivery and customer service, none of which have anything to do with producing successful movie and tv shows.”

As if television and film are businesses that have money to burn these days. I understand what Long is getting at: content is volatile. But I think he’s missing what innovators like Amazon and Netflix are going to bring to the business.

Even if Amazon or Netflix are no better at creating breakout shows than the traditional studios, they are willing to consider that the production process itself is ripe for improvement. And if these new media companies are able to improve their productions by even a small amount, those successes will compound over time. Much like in poker, the measure of skill is in the long game. You can’t judge a player’s skill by one hand, you need to consider how a poker player plays a thousand hands; the same goes for the production companies and studios.

There has been a lot of talk about the crazy expensive deals Netflix has made with television producers Shonda Rhimes and Ryan Murphy may seem like business as usual. But with no inside information myself, I’d guess that it’s the data they’re after too. After all, data is the most valuable resource in the world nowadays.

I’m sure that Netflix believes that Rhimes and Murphy are going create hit shows for their subscribers, otherwise Netflix wouldn’t have paid such a premium. But Netflix also knows that Rhimes and Murphy are going to establish experienced producing teams that will generate a ton of production data for Netflix to study. Data that will completely accessible now that these shows are being produced in house. Data that will allow Netflix to learn how to apply efficiencies and create hit shows from the ground up at cost.

(I imagine that David Letterman’s team is going to get some interesting questions from the Senior Data Scientist – Studio Production in the near future.)

George Pelecanos’s short story, coincidentally called, The Martini Shot illustrates a common situation ripe for innovation that Netflix is probably already pursuing. The main character, a successful television producer, describes a location shoot where the Director of Photography is ordering another needlessly complicated shot that will never be used in the edit, but pushed the crew into overtime anyway:

“But we knew Lomax’s MO. He leaned toward artsy, with shots that made no sense in terms of POV, angles and footage we’d never use when it came time to cut. The secretary’s arrival, easily accomplished by a walk into frame, would be complicated by his insistence on bringing her in with a dolly shot, which meant laying down track and more lighting, which meant time. We’d get behind, and the rest of the day we’d be playing catch-up, and consequently the last scene or two would suffer. We’d worked with Lomax before. He made the days longer than they had to be, but he was all right.”

Not only is this wasteful, but needless overtime is also dangerous.

If producers were able to correlate production payroll with their editing timelines an entirely new dimension of decision making would become possible. Producers would be able to look at shows at a cost per shot, per scene, and per sequence basis. Conversations about how to make shows would change significantly. And if producers are not going to see the value in improved efficiency and accountability, then bond and insurance companies will. Amazon, the master of logistics, is the right company to use data to drive down costs, which over enough productions improves the likelihood of profitability.

This shift will require thinking about the process production process differently. Much like how Elon Musk’s Gigafactory announcement reframed how we should be thinking about factories:

“The biggest epiphany I’ve had this year is that what really matters is the machine that builds the machine – the factory.”

Margin shaving, supply chain management, on time delivery: all of these things will be executed very differently in television and film production. But their core goals: reducing waste and improving efficiency are much needed in every business. I wouldn’t bet against innovators like Amazon Studios.

Management Media Theory

What the Bravo docu-soap can learn from Netflix (part-2)

“The biggest epiphany I’ve had this year is that what really matters is the machine that builds the machine – the factory.” — Elon Musk, on the Gigafactory.

Netflix is starting to think about the television production factory. Others should take note because the ability to reduce waste and improve efficiency are useful skills that deliver significant advantages over time.

In order to develop a LEAN production model for television production, we need to break down our products into components and analyze how we’re putting each piece together. I’m using the Bravo-style docu-soap as a frame work because their franchise offers experimental opportunities not commonly found in television. In part 1 of this series I wrote about the confessional interview. In this part I’ll propose some thoughts about the notes process.

Experiment 2: consolidating notes


Consolidating the cut delivery and notes turnaround into ‘blocks’ can reduce wasted editor time and increase the quality of attention given to review cuts.

Notes from the Network

Within the unscripted production community just whispering “Bravo notes” is enough to conjure dread in the hearts of even the most seasoned story producer. In addition to the quantity of notes received with each cut, Bravo is notorious for asking for additional cuts beyond what’s formally agreed upon, i.e., the notorious Rough Cut 5 or the Fine Cut 7.

As my colleague FarFromReality jokes: “For those of us in the trenches – particularly in Post production – saying you are about to work on a Bravo show is to say, ‘I am going to be continuously abused for the foreseeable future and my production company is probably about to go into a deficit.’” But the truth is that nobody really likes notes from any network.

As I’ve previously illustrated, network notes turnaround is the area where production companies are likely to accumulate overages. Making matters worse is that the easiest option for the production company to minimize overages has a negative outcomes for the staff: telling story producers and editors to go on unpaid hiatus until the notes come back.

On the network’s side, I suspect executive satisfaction is low among those inundated with multiple cuts from disparate series. Multiple studies have clearly demonstrated the advantages of paying deep attention and not multitasking. The cognitive load of switching between series and chronology (watching episode 6’s rough cut before episode 4’s fine cut) must certainly take its toll on the quality of notes given.

Therefore it is only logical to look for ways to simultaneously reduce the chance for overages and improve the quality of the collaboration between the network and their producing partner.

Linear Weeks Grid

Linear Weeks Grid stacked
Sample staggered post schedule most unscripted shows use.

Right now most shows are edited on a staggered schedule. (Episode 2 starts editing a week after episode 1. Episode 3 starts editing one week after episode 2 and two weeks after episode 1.) While this seems like it would be efficient at first glance, when you look at how this plays out in the post schedule, staggered scheduling creates weeks in which disparate cuts are delivered with no regard to a human’s ability to retain information or switch contexts.

Post Schedule dashboard overview
Staggered schedules create situations where a network executive is expected to watch & note: episode 4’s rough cut, episode 1’s second fine cut, and episode 3’s fine cut.

Block scheduling

What I propose is experimenting with the block scheduling of cuts and the notes process. For example: Production company delivers episode 1 – 4 rough cuts all at once. Then the network executive reviews all of the cuts together and gives notes all at once.

Linear Weeks Grid Netflix blocked
Block scheduling would deliver a series of cuts and notes all at once.

The goal is to reduce notes turnaround and improve the quality of notes as well as strengthening the network-production company relationship, by giving showrunners the full attention of the network.


I don’t know how either of these experiments would end up. But the point is to try and discover opportunities to reduce waste and improve efficiency. I believe Bravo is in a unique position because their franchise shows are ripe for experimentation. But LEAN thinking is available to all of the networks and production companies. The question is which parties are going capitalize on these opportunities first?

Management Media Theory

What the Bravo docu-soap can learn from Netflix (part-1)


One of the advantages Netflix will enjoy in the future is that they are willing to rethink the entire production process. One look at their recent job postings will demonstrate that they are looking for opportunities to minimize waste and increase efficiency:

Senior Data Scientist – Studio Production

Netflix is seeking a versatile data science practitioner who is ready to tackle data analysis and modeling challenges in a refreshingly new problem space – Studio Production Science and Analytics.

Recruiting Researcher – Creative Content

This researcher will be responsible for uncovering candidates not already known to us, no small task in a business that can be especially insular.

Production Engineer, Live & Multi-Camera Production, Studio Technologies

Create blogs, documentation and other support resources to educate creative partners and vendors about our requirements and initiatives.

I believe Netflix could do more (notice that the content team is noticeably absent from their tech blog) but I’d like to see Hollywood’s old guard embrace Silicon Valley’s attitude of experimentation at the production level. Therefore, in the spirit of lean production, I offer the following two experiments to the Bravo-style docu-soap.

Why the Bravo docu-soap?

Firstly, unscripted programing (from documentary through all of reality TV’s sub-genres) are edited together from footage that is easily categorized. Vérité, b-roll, time lapse, confessional interviews, car cameras, cast cameras; all of these shots are easy to identify, and more importantly, the relationship between unscripted shots is easier to understand and quantify than scripted. Therefore, it is possible to analyze the final episodes and draw useful conclusions without image analysis.

Second, the Bravo franchises in particular present a unique opportunity. Although much pop culture criticism has been written about the differences between each of the Real Housewives series, generally speaking they are all the produced similarly. Each series is composed of the same stylistic elements, which make it easier to compare production variations within the franchise, as well as with other unscripted programing.

Experiment 1: Delay interview recording


By delaying the interview recording until the very end of post production, when the story producers (and network executives) know which questions to ask, the show’s production could eliminate unnecessary expenses and improve creative consistency.

Anecdotal Interview Statistics:

Bravo interview
Lisa Vanderpump in the iconic Bravo “confessional” interview.

The Bravo style docu-soap records cast “confessional” interviews multiple times throughout the production. Usually 4 or 5 days of interviews for each month of shooting. In addition, productions will record an additional 2 or 3 rounds of week-long “pickup” interviews after production has wrapped. In addition to the location expenses (studio rental), each interview day requires a minimum: Camera, Audio, Hair & Makeup, Story Producer, and Line Producer. Interview footage probably has the highest ratio of footage recorded vs used, which makes it an extremely wasteful and highly open to optimization.

Ep 02 percent
Episode 2 Act 2 Interviews are 16% of the sequence.
Ep 12 percent
Episode 12 Act 5 interviews are 13% of the sequence.

In addition, as the series progresses, the distance between when a scene’s vérité was recorded and its corresponding interview recorded grows. In my sample sequences episode 2 act 2 interviews were recorded 28, 68, and 72 days later; while episode 12 act 5 interviews were recorded 97 and 117 days later.

Ep 02 days after
Episode 2 interviews were recorded two months after the vérité.
Ep 12 days after
Episode 12 interviews were recorded over three months after the vérité.

I believe this happens because later interviews are addressing specific notes from the network. Since the field producers know what questions they’re asking the cast, the answers are often more concise.

By consolidating interviews until the very end of post, it is possible to imagine that each cast member could be interviewed over 2 days, instead of the usual 5 – 7 days over the course of a season. If you consider a cast of six, consolidating the interviews until the end of post could save up to 30 shooting days.

My goal in posting these experiments is to prompt Producers to start thinking about how their shows are made at the production-level. Our tools are not making it easy. But it’s only when we start to breakdown each episode into its components that we can start to create a Hollywood version of the LEAN production model. In Part-2 I’ll offer another experiment.

To Be Continued…


Time to Process a Hour of RAW Footage

Building on top of my previous post about the need for NLE analytics; I finally thought of a good tool to measure workflow that the NLE should automate: time to process a minute of raw footage. It would be expressed as a ratio.

For example:
  • an xdccm workflow that involves loading from a PDW-1600 deck would have a ratio of 1:1.
  • an Arri workflow that requires transcoding ProRes4444 footage to DNX36 might be 2:1.
  • a time-lapse workflow that involves Lightroom and After Effects could be 12:1.

I think this measurement is helpful because it can inform the budgeting and planning phases of post production by answering questions such as; is it worth spending the money to have a DIT transcode footage on set? How much time AE time will be required to properly load/ingest this footage? It will also make it easier to compare workflows

I will attempt this on my next show. I would even like to include the time to preform peripheral tasks such as labelling and backup copying. I suspect that there will be a big variation between processing lots of little files and a few long files. Anyway, I think this is the first useful measurement and indicator I’ve come up with since I’ve started asking myself this question in late 2009. I’m excited to try it out in the future.

All that said; this is something that a computer would be able to spit out immediately and I hope that someone at Adobe and Avid are listening.


NLE analytics

Compared to other businesses, unscripted production operates with such little data. It feels like ratings are the only time when anyone is held accountable to any sort of number. But in the world of analytics, this seems crazy to me, and I think the NLE companies should step-up their game in this area.

One feature I’d love to see added to an NLE is a table that breaks down a sequence by source tape. This table would report that 3 minutes of a sequence came from tape A, 2 minutes from tape B, and 5 minutes from tape C, etc etc. When supplemented with additional meta-data, this would be a rich source of information for producers.

I imagine that this would be useful to producers because it would help them identify trends such as; did their show draw from the source material evenly, or did they overwhelming rely on one or two sources? Perhaps one cameraperson was more effective at shooting usable footage. For a television series this information could be compiled into seeing bigger trends; such as were some types of shots completely unnecessary?

For example; a few years ago while working on a competition show, I manually dug through sequences and budgets and told a production company executive that the cost of Loading, Grouping, and Storing the car camera footage for the entire season was approximately thirty thousand dollars. And yet we only used the car camera footage for one scene in the entire twelve episode season. How they continued to use a car camera in future seasons without a serious conversation about the purpose of this type of footage was beyond me.

Another area ripe for analytics is the talent interview. At the end of a docu-drama season, how useful would it be for the producers to see where most of their interview bites came from? I could see myself analyzing which field producers created the interview bites most heavily drawn on by the story producers.

In a business of shrinking budgets, efficiency is more important than ever. I’m not looking to have analytics drive decisions, but they should at least inform them. (And perhaps curb some of the crazy.) The primary tool of our trade, the NLE, is grossly behind in this area and I’d like to see Adobe or Avid rise to the challenge.