Learning, Data, and Love

  One of the primary theses of my research is that people are not like films, people are like other people. As such, recommendation engines that try to base what you would like to watch on what you have already watched are inherently flawed. Even if this were the right starting point, the algorithms currently doing this work are likely to value the wrong things when comparing films. It is very possible to love Iron Man and hate Iron Man 2, but not according to these current systems.

So what should we be comparing? My belief is that people's behavior tells us what other people they are like, not what films they will like. I'm much more likely to watch a movie recommended by friends who share my tastes than by an advertisement, critic, or stranger. But if we are to accept this as the starting point, can we define people well enough to make the right match?

Well, that is certainly what I'm exploring in the film world, but some very smart man seems to have figured it out already in a not-so-different space: Love.

chris mckinlay

UCLA mathematician Chris McKinlay is profiled as doing just that in the Wired article How a Math Genius Hacked OkCupid to Find True Love. In short, he used data he collected by mining OkCupid to separate all female users into seven distinct cohorts. He could then look at the properties of these groups (age, religion, etc) and determine which fit the profile of someone he would want to date.

I certainly recommend reading the whole article (as the story of a mind who tries to find love by sleeping on a mattress in the cubicle of his office creates a compelling dichotomy,) but for the purposes of film marketing suffice it to say that this level of increase in marketing effectiveness is a virtual goldmine. McKinlay went from 100 high level matches out of 80,000 women (.1% effectiveness) to 10,000 high level matches (an insane 12.5% effectiveness rate!)

And it wasn't that he was lying to anyone. He just determined the right questions to answer that were the most relevant, and how important each of those questions were to the given cohort. Isn't this what marketing should be? Get the right message in front of the right people and your conversion rate soars.

At the risk of spoiling the article, I found this quote from McKinlay's now-fiancee to be the most illuminating.

“People are much more complicated than their profiles,” she says. “So the way we met was kind of superficial, but everything that happened after is not superficial at all. It’s been cultivated through a lot of work.”

The way we introduce people to movies for a marketing perspective is always going to be somewhat superficial. The key is to make sure that after that introduction has been made, there appears to be the right motivations behind it. This gives the audience a reason to trust you and your message, and creates a virtuous cycle.

While my research may not take us all the way to the promised land (and it certainly won't get anyone a date,) I think it begins to push the ways in which we evaluate and understand audiences and their preferences. Most importantly, experiments like this should show distributors, marketers, and even studios the benefits of exposing their data. After all, when do you think was the last time any of them hired someone with a PhD in mathematics?

Real Film Data (Part 3: Online)

To this point, we’ve looked at why it’s hard to get data about film audiences from both theatrical and VOD companies, be they exhibitors or distributors. The misalignment of incentives is the primary motivator for the lack of collection of and the hesitance to distribute real numbers and reporting.

Fortunately, there is a new form of film distribution that is all about the data. Companies like VHX*, Gumroad, and Vimeo are enabling filmmakers to control the pricing, release strategy, and perhaps most importantly, the data that results from sales. This is a huge change in ethos from the business models we have been talking about

*Disclaimer: I have been working with VHX since the beginning of the year, specifically as a Data Analyst. The screenshots below are from their service largely because of the ease of access, but other services provide the same types of reports. 

What Data Looks Like

The images here are from the view a filmmaker would see in their dashboard. This is the top level report that shows the number of sales per day over the course of a week. It also shows important aggregate data such as “All-Time Revenue” and “Total Sold,” although these number are also things would would get from a distributor who offers your film on a VOD platform.

So what’s the big difference?

Screen Shot 2014-04-13 at 3.44.03 PM

With a traditional distributor, these numbers come in quarterly (if not longer,) by which time it’s too late to make any changes to your marketing, publicity, or release strategy. But with real time data, you can see your sales as they happen, meaning you can take insights away about how effective everything it, and even where advertising is working and where it isn’t.

You may be thinking, “If my film isn’t doing well, what can I really do?” Because of the amount of control you have, there’s actually quite a bit you can do.

What Pricing Looks Like

The first thing is the most obvious, you can lower your price. But rather than just dropping the price for everyone, you can run things like flash sales. By offering a coupon that discounts the price of the content, you incentivize people to buy the film before the price goes back up. It turns out that this is a strong motivator for audiences to purchase.

You can also tie bonus content to the movie and charge a premium price. Some films, such as the documentary Stripped, have massive amounts of additional footage that they add to a more expensive product, allowing super fans to get more content while still allowing everyone else to just see the movie. Other films, such as Camp Takota, leverage services like Shopify to sell merchandise along side digital downloads of the film.

There’s another important question companies like this hope to answer. Once people start buying your film, how do you spread awareness without spending boatloads of cash on marketing?

What Your Audience Looks Like

Once you get your film out there, whether it’s the “free” publicity from a film festival, active social media engagement, or just straightforward advertising, it’s important to know who is coming to your site and who is buying your film.

Screen Shot 2014-04-13 at 3.44.11 PM

The image here shows one way that the VHX platform tracks that, showing the source of traffic to your site, the conversion percentage for each source, and for all countries. This is a great way to get a top level view of where you audience is located, but there’s also a great way to know exactly who your audience is.

When someone buys the film, the filmmaker knows two important things: their location (at least their state/country) and their email address.

This is incredibly powerful for engaging with your viewers when it comes tim to grow your audience. You can send the people who purchased your film a discount code that they can share with friends, thus reaching people who your advertising may never get to. You can also carry this audience over to your next film, so you’re not starting over from square one with each film.

If you want to read more about great community engagement, read the Indie Game: The Movie case study. It might change your life.

What Isn’t There… Yet

I have painted a rosy picture of this world so far, but it is not quite there yet. The industry classifies this type of distribution as Electronic Sell Through (EST,) and it makes up a very small part of the industry’s gross revenue. This is partially because filmmakers and distributors are not looking at this as an important way to sell a film, but also because it is not a way that audiences are used to consuming their content.

It can also be more difficult to drive an audience to a specific film website as opposed to a market place. Think of it as the difference between the person who went to Blockbuster looking for a specific movie as opposed to the person just browsing for something interesting (you may have to be over 29 years old to understand that.)

That said, these problems are not unsolvable, and in fact are well on their way in the right direction. Shortly after championing the new models of distribution in his speech at the Edinburgh International Television Festival, Kevin Spacey self financed and self released his documentary Now: In the Wings on a World Stage, while Joss Whedon released his film In Your Eyes on Vimeo On Demand for rental.

As more people within The Business start to understand the benefits of this type of release, I imagine the problems presented above will fade.

Data: An Epilogue

What is next for us. Now that we know the truth about what data is out there, we can begin to explore what we can do with it. The next step for me as part of the CRI program is to start the quest toward defining a way of talking about this data that makes sense to films and filmmakers. The first step toward that will be next week.

Forest for the trees: Risk categorization criteria

Over the last month I’ve been wandering through a forest of data - exploring and compiling various pieces of information related to 1350 films that received a US theatrical release in 2012 and 2013. This data will be used to train the initial rubrics that will guide my separation of films into four risk levels.  While I’m still in the process of finding and organizing this data, I wanted to reach out to readers to ask for your thoughts on my approach. As a reminder, I’m initially dividing the index into 4 sub-indices, each with a distinct level of risk. See my last post for more detail. My job, therefore, is to properly select criteria that correlate to risk (in terms of films’ financial returns). The 3 criteria I believe will be most important follow:

Marketing budget

I believe this is one of the most important factors contributing to a film’s success or lack thereof. You can tell a unique story through an amazing, well-crafted film, but if you don’t properly get the word out about it, it won’t be seen by many people. Unfortunately, I’m discovering that accurate information about films’ marketing budgets is quite difficult to find.

I’m planning to compile information from filmmakers, distributors, ad agencies and journalists to triangulate/approximate marketing budgets for specific titles. However, this may not be fruitful in a lot of cases. And it’s hardly a scalable approach. So, dear reader, where else might I look for information about marketing budgets? Extra points for accurate, scalable sources.

Production budget

A film’s budget impacts everything about it. If a story hasn’t been sufficiently supported through its budget, an amazing story and/or cast won’t shine through. Nor will a marketing budget be as effective, regardless of its size. This is not to say that all films have to have a large budget – many very low budget films have still performed very well, both critically and financially. While I believe it highly likely there will be strong correlation between film budget and ability to drive revenue, I also believe films with distinct profiles have distinct budget levels that will indicate propensity for success at the box office and beyond.

Sources like BoxOfficeMojo are fantastic aggregators of data like film budget. However, it’s not assured the recorded budgets are correct. Additionally, I’ve found sites like this don’t record budget for hundreds of low budget indies. Where might I find budgets for these smaller films? How might I confirm the accuracy of reports made by data aggregators like this?

Attachments

Films are not commodities. Films’ casts, directors and producers undoubtedly have enormous impact on the finished product. Therefore, I’m positing that attachments themselves have significant impact on the potential performance of a film independent of its production budget or marketing budget.

The creative players in a film are some of the easiest pieces of information to discover with certainty. However, unlike budgets, attachments are not numerical in nature. To approximate a numerical system, I’m taking on the highly imperfect task of devising a scoring system within which to place key attachments, based on the financial success of films in which they previously participated. I am trying to create a science in something that is obviously unscientific. Any opinions about factors I might want to consider along the way are highly encouraged.

I should note that I will be tracking other sorts of data as I organize my training data. Critera like distribution strategy (wide, platform, limited, day-and-date, ultra-VOD, etc.), source (studio, mini-major, established indie filmmaker, unknown, etc.) and origin (existing blockbuster book, original material, etc.) are likely related to level of financial risk. However, these may end up being largely duplicative of the above, more complex criteria. Also, I’m limiting the data pool at this early point in hopes of not muddling other correlations that may prove strong.

I’m lucky to have had interested mathematicians and data scientists reach out to help me discover and organize meaningful conclusions from my raw data. Engaging with folks who read these posts has been incredibly helpful. I want to talk with even more people – so, please, comment on this post or email me directly at colinwhitlow@gmail.com if you’d prefer. I’m open to all ideas and questions.

Ubiquity

My final post for the Real Film Data series is coming soon, but in the meantime I'd like to share a post I worked on for VHX. For those who don't know, VHX is a platform for filmmakers on which they can release their own content, their way, for the price they see fit. I've been working as a Data Analyst with them since January and have had the opportunity to leverage their knowledge into some interesting insights.

One of those is in the following post. It pertains to how you should release a film and across which platforms. Turns out, so long as you do it right, it helps to have your content everywhere.

Exclusivity is dying. Long live Ubiquity! [Click here for the original post at VHX]

Real Film Data (Part 2: VOD)

In my last post, I discussed the difficulties of getting data from theaters, primarily due to misaligned incentives. It is not due to malicious intent per say, rather that the costs of data collection and analysis do not contribute substantially to their bottom line. In the distribution window I will talk about today, I'm going to continue to follow the money. This will give a picture of where the incentives lie and why filmmakers get left out.

Warning: This is going to be a long walk, but important in understanding where filmmakers and their representatives have different objectives. So hang in there! For those of you who aren't here to read, the next little bit a nice summary.

tl;dr - Platforms like iPhones, Netflix, and Comcast have the data we want, but as filmmakers our only link to them is through a distributor who has no interest in risking the loss of their value proposition by pushing for data transparency. Thus, it's up to filmmakers to find different strategies for getting data about their audience. These strategies are things are what I will be exploring in the next quarter of my fellowship.

VOD

The Video On Demand space consists of two separate models, Transactional VOD and Subscription VOD. The former is a one time payment that allows for the purchase or rental of video content. This includes both cable providers such as Comcast and Time Warner Cable (soon to be the same company) and online marketplaces like iTunes and Google Play. Subscription VOD consists of Netflix, Hulu Plus, and Amazon Prime who offer a library of content for a monthly or yearly fee.

Now that we've gotten the boring stuff out of the way, let's move into the fun part: cash money!

None of the platforms mentioned above will just let you throw your film on their service. These companies like to have some arbiter of the content that makes it on to their service, both to reduce their overhead in managing the films and also to ensure only the highest quality films are shown.

There are two ways to get your film on a VOD platform, either through a distributor or through an aggregator, each of whom has a different way of making money off of you.

Distributor

Distributors that work with indie film in the VOD space are numerous. Their method of acquisition is to (sometimes) offer a minimum guarantee to the filmmakers in exchange for rights to put the film up across multiple platforms. This MG is an upfront payment to the filmmaker that must be recouped by the distributor before any revenue from the sale of the film returns to the producers. Furthermore, distributors must also recoup the costs of advertising, marketing, and other services in advance of returns to the filmmaker. Finally, they take a split of those profits as well, generally between 25 - 40%.

If you sell your film to a distributor like IFC, for instance, they are likely going to ask for worldwide rights for all media. This means your digital VOD rights as well as any theatrical, physical media, television, etc. Other distributors like Gravitas work almost completely on digital VOD platforms, so you will be able to retain some of your rights to other venues, but likely with no minimum guarantee.

Aggregator

If you have the money to pay upfront, an aggregator like Distribber or Quiver will take your film, ensure quality to the necessary standards, package it with other films, and put them on various the VOD platforms. Because you pay them, they do not own any rights to your film and do not take share in profit participation. 100% of the sales that come in go directly to the filmmaker.

The cost for these services are usually between $1,000 - $2,000 and increase depending on the number of platforms you'd like the film be for sale/view. If you want in available on all satellite and telco platforms, this adds roughly $5000 to the cost.

Example

What is the benefit of going with distributor vs. that of going with an aggregator? In a frictionless market, this question is complicated but not impossible to estimate. Let's use online transactional VOD only and assume platforms are taking 50% of the sales. For a movie priced at $10 dollar, there is a clear point where one see the benefits of an aggregator over a distributor.

In this case, selling more than 1,600 copies results in a better outcome by going through an aggregator. Something certainly attainable for a film with decent marketing, publicity, and word-of-mouth.

So this is the answer, right? Plug in these numbers, get an answer as to how many units you'd have to sell, and make your decision. How easy!

Not Even Close

In the example, I mentioned something about "frictionless markets." Basically this means that all methods of putting content out there are equal and none is subject to any benefit or harm that isn't experienced by the entire market. Turns out, this is where distributors exhibit their leverage.

As businesses, they have relationships with platforms and can lobby for better placement of their titles and, depending on the platform, more favorable terms. And the difference between landing on the front page of iTunes and being buried under hundreds of other films can make a huge difference to your sales. So while 1,600 copies may be the break even point, a good distributor will shift the playing field in your favor to ensure you get more sales than you would without them.

You may be asking by now, "What does any of this have to do with DATA?!?" Well, now that we have walked through the value proposition that a distributor offers over a DIY model, we can talk about how that value proposition keeps them from getting filmmakers the data they need.

Why You Get Nothing

There are three points here. First, distributors' marketing arms are not concerned by and large with the film's audience, but rather with the platforms (iTunes, Amazon, Cable Providers, etc.) In order to pay off their value proposition, distributors have to be able to convince these other companies to act in ways that are advantageous to them, such as placing their titles in front of more people. Notice I said "more" people, not necessarily the "right" people.

Secondly, and stemming from the first point, is that distributors have not been pushing for data regarding the audience that purchases the film. This is because the platforms run by cable companies focus more on service uptime and connectivity than user interaction. Those run by tech companies flatly do not share their data.

And perhaps most important of all, were a distributor to expose to a filmmaker ways to measure and connect with an audience, the would immediately weaken their own position. Distributors would like filmmakers to believe that they have some expertise in finding an audience for a film and would suffer greatly if the creators could speak directly to an audience.

For these reasons, the data filmmakers want is either not being collected or not being released. Information about a film's audience on VOD is therefore kept by entities that the filmmaker never directly works with (platforms), and their ambassador to these companies (the distributor) has no incentive to push for transparency.

Unite!

As filmmakers, what can we do? First, we can begin to share the data we do have. Knowing what data is out there, and how little there is, will begin to motivate new filmmakers to demand more. Secondly, we can migrate to platforms that are more transparent with their data. Which, coincidentally, will be the topic of next week's post.

Stay tuned!

Real Film Data (Part 1: Theatrical)

The theatrical film market is dark and filled with terrors. At least I think that's how the quote goes. By "dark" of course I mean in regards to data. The purpose of this blog post is to look into who has data about audience theatrical viewing habits and whether filmmakers are able to access it. While we live in a world where a good number of our actions are tracked, calculated, and processed so that our actions will be better understood, 94% of people still walk up to a theater, pay for a ticket, and watch a film without another soul in the world knowing they did so.

Privacy advocates may love this homespun tale of a world without information, but to filmmakers it is incredibly troublesome to find yourself lacking the knowledge of who your audience is.

Currently, there are a few groups of people who have information about theatrical audiences:

1) Exhibitors - First off, it's important to understand how theaters make money. Concessions! Because studios take almost all box office revenue from the first week of a film run, theaters have to charge a massive amount for food and drinks just to stay in business. That's why the single most important metric theaters track is called "per-head," which measures amount spent at the concession stand.

If you are a member of a rewards program, however, congratulations! You've just exchanged your data for the occasional free bag of popcorn.Companies mostly use loyalty programs to keep customers coming back, but also to collect and understand their purchasing behavior.

As a filmmaker, you get none of this.

2) Ticketing Websites - Sites such as Fandango and MovieTickets.com* will presale tickets to a film, therefore knowing quite a bit about you (email, geographic information) and will use this to directly market to you. They do a pretty scatter shot job, as most correspondence is through non-targeted email campaigns. That said, they have demographic, geographic, and direct contact information for actual customers, and may even be able to draw more information about them through site interactions.

As a filmmaker, you get none of this.

* Fun Story: Fandango is owned by Comcast; MovieTickets.com is a joint venture by many companies, including AMC theaters, Viacom, and Time Warner. If you think they are in any way interested in sharing data with filmmakers, well, just wait until we get to TV...

3) Tech world - Finally, some people that might be interested in helping filmmakers. Some are specifically film based like Moviepass. Their subscription service allows users to pay a monthly fee for the ability to use a pre-paid debit card to pay for movie tickets.

In order to use the service you have to have a smart phone with their app, which uses your location to "check-in" to a theater, thus allowing you to purchase the ticket with their card. This gives Moviepass direct information about who is seeing which films and where in real time. They also have an in-app ability to invite friends, which allows them to track word-of-mouth. Unfortunately it seems like Moviepass is more interested in selling this data to studios and theaters than directly to filmmakers. You can check out a great primer on Moviepass and their value add by clicking here.

Taking a step back from film-centric apps alone, location-based apps like Foursquare could be helpful, allowing mobile users to check in for specific events at a given location.

Imagine that for a second. Someone checks in to a theater for your movie. You can then see where else they have checked in begin to make strong conclusions about what type of person this is. In the aggregate, this allows someone to forecast what types of people see what types of films and be able to types of places they most frequently eat, shop, and spend time.

I hope that the use of existing data sets alongside newly created technology will be able to utilize what is available from these newer models of data collection.

--------------------------------------------------------------------------------------

Of course the largest issue with all of these methods is that they take place after the film has been released. They might be able to help a smaller platform release target its marketing as it grows from city to city, but at that point a shift in strategy can do more harm than good.

However, if we begin to study the actions of an audience over time using platforms like the ones mentioned above, we can aggregate data to expose patterns of behavior. Once we do that, we can define films in such away to exploit those patterns and put films in front of their audience first without breaking the bank on advertising.

The real meaning of "data" in film (a prologue)

I begin most of my conversations about film these days with one topic in mind: data. As anyone who has read Moneyball will remember, access to data doesn't necessarily make for better decisions. Thus, my research is not just about getting to the data, but making sure that what we get is useable. And we have already seen some larger companies begin to do this.

Google, Apple, and Netflix make strategic and product decisions around their vast data resources. Hollywood studios have largely ignored the proliferation of data analysis, even as several individuals in the industry have made a half step toward increased data transparency. While this makes for interesting reading, what does this mean for an individual filmmaker?

Over the next week, I will be breaking down data by window (theatrical, VOD, online, etc.) and by who currently has access to that data (distributors, exhibitors, filmmakers, etc.) In doing so, I will illustrate the sheer lack of information in the industry in total. Then comes the scary part.

I will show how little of this data, both financial and marketing, trickles down to the individual filmmaker. We should care about two things after a film's release: who saw it and for how much. My next few posts will show just how in-the-dark most of us are.

There are 5 venues I will be discussing: -Theatrical -Physical Media Sales -VOD (both transactional and subscription) -TV (pay TV, cable, network licensing, etc.) -Direct Sales

Each of these exhibition avenues is a different market and poses a different set of problems. As I go through each of these areas, I'm going to treat them with a hacker's eye. I want to see how close much of the traditional system can be replicated piecemeal.

My primary question is: How can one release a film in as many places as possible while maintaining control over both revenue streams and individualized data? And what are the side effects, both positive and negative?

By the end of this week, I hope to have clearly laid out where the industry stands. This will show how far we need to go to get what filmmaker's need to be successful.