Growthology

Video Streaming Quality and Viewer Behavior

Startups and businesses that offer video on their website or on a mobile platform face a fickle audience. If videos don’t load or are frequently interrupted with buffering, viewers are more likely to abandon the video and not return to the site. These are the finding of study by S. Shunmuga Krishnan and Ramesh K. Sitaraman, “Video Stream Quality Impacts Viewer Behavior: Inferring Causality Using Quasi-Experimental Designs,” who analyzed 23 million video views from 6.7 million unique viewers (paper available here).

I am not terribly surprised by the results. We might intuitively expect video failing to stream smoothly or quickly would lead to viewers giving up on watching the video. But the authors importantly establish a causal link and offer some nuanced findings that I think are informative. My high-level takeaway is that as internet speeds increase, consumers become more impatient and expect better quality from video providers. Here are some selected highlights from the paper:

• If a video fails to start up within two seconds, viewers start to abandon the video, and the drop off is steep. However, abandonment varies by internet speed. Viewers with faster internet connections have less patience and abandon sooner. Mobile connections give a lot more slack.

• Longer videos get more slack to startup (lesser abandonment rate from delays in starting up) than shorter clips.
• Viewers who dealt with rebuffering interruptions and delays viewed the videos for less time.
• Viewers who have difficulties are less likely to return to the site, and this is persistent for at least a week.

Knowing this is important because abandonment rate, how long videos are viewed, and return viewers are all factors for generating revenue.

 

Band Camp, Judges, and Interviews

My overall point in this post is to say that interview timing and order matters a lot. Perhaps the early bird does get the worm after all. More nuanced explanations follow below after I present some stories and evidence. Now, onwards to the text…

Growing up, I played alto saxophone for a number of years. I played on my own and took lessons while also playing in the official school bands in middle school and for most of high school. And like many band members, I attended band camp—specifically for me, Midwestern Music Camp at the University of Kansas for an entire week (you stay at the KU dorms; you rehearse in Murphy Hall and perform at the Lied Center; very cool for a band kid etc.). I think I did this for four or five summers.

KU runs a very large camp, so they split campers into multiple bands depending on skill level. My recollection is that to decide skill level and chairs,¹ you audition with the department head of your instrument at KU. I’m foggy about the exact process, but I remember signing up for an audition time and spending hours in agony practicing and waiting before being able to get the audition over with. Don’t quote me on this, but I want to say that you could check-in and audition in the morning and afternoons over a two-day period (perhaps it was only one day; I can't remember).

I was not even the best saxophonist at my own high school, but overall I was fairly good. One of the later years I attended band camp, I made 2^nd chair in the top flight band, which I was estatic about. What displeased me, and to the best of my recollection the rest of the saxophonists in the band as well, was the skill level of the individual who was given 1^st chair. Frankly, this person was very mediocre. They misplayed all week long and at multiple points just dropped out of sections of songs entirely because they could not keep up with the music. I was very annoyed that this person was given 1^st chair ahead of me.

Perhaps it had something to do with the audition timing.

I’ve been a longtime reader of Ed Yong, who a couple of years ago highlighted a paper by Shai Danzigera, Jonathan Levavb, and Liora Avnaim-Pessoa that documents an odd trend in judge parole rulings (Ed Yong’s summary here; full paper here). Here is the most relevant chart from the paper:

What this shows is the proportion of parole rulings that favor the prisoner over an average day of rulings (tick marks on the x-axis note every third case). Data were collected from 1,112 hearings over a 10-month period. The dotted lines represent food breaks and effectively creates three ruling sessions each day. The chart shows that prisoners at the beginning of each session have a much greater likelihood of receiving a favorable ruling than those at the end of each session. This holds across the entire sample of judges and rulings (more controls for other variables and explanation are available in the full paper). What Danzigia and his co-authors theorize is that the judges' mental resources are used up early in each session, so they are sticking with the easier status quo decision as they tire out; the food breaks help judges reset. Whatever the interpretation, there is clearly some unfortunate bias against prisoners who receive rulings scheduled later in each session.

A recent paper by Uri Simonsohn and Francesca Gino analyzed thousands of interview timings of MBA students (gated full paper here; earlier working paper version here; press release summary here). They find that interview position matters here too. Over multiple days of interviews, what does an interviewer do if the best applicants all come on the same day, or even all on the same day right next to each other? Simonsohn and Gino find that an interviewer expects to recommend the same number of applications each day and that this constrains their ratings. Once they near this subconscious daily quota on each day, they are hesitant to approve additional applications from that day regardless of their overall merit. This means that interviews early in the day have a negative impact on interviews later in the day; if early interviews are given high scores, the subsequent applicants are more likely to receive lower scores regardless of the applicant’s characteristics. This effect grew even stronger as the day wore on.

I cannot remember what time of day I auditioned at band camp so many years ago. However, knowing myself, I would have been sleeping in and saved a couple of hours to practice at KU before the audition, so there is no way that I auditioned in the morning. If we assume I am objectively able to rate and compare my saxophone skills, I can say that I am now disappointed to now find out years later that logically it looks like my audition timing helped seal my second-best chair position. I perhaps did not look as strong in the afternoon as I might have in the morning.

I think the research I’ve summarized here extrapolates well to startups—who have to interview, pitch, and audition to various parties all the time—particularly the interpretation from the MBA interview paper that high scoring early on creates a bias against later candidates. This is particularly important to consider in the context of startup competitions or submission of competitive bids. I will weakly suggest that if you have an option, perhaps going early in a competitive scenario is better than waiting. The more powerful interpretation is something you have no control over—hope that your strongest competition isn’t interviewed or reviewed around the same time you are? Yikes, that’s depressing. Perhaps judges and reviewers should make sure to take plenty of breaks when they are reviewing and scoring pools of applicants.

---

¹ Chair is music terminology referring to skill and seating position within a band or orchestra. The best player of a given instrument is 1^st chair of their section and second-best is 2^nd chair, and so on and so forth.

Fifth Annual Kauffman Economic Webloggers' Forum 2013: Economic Weblogging, Financial Weblogging, Weblogging Economists, and Economics of Weblogging...

...all of these are topics that I think fall within the proper purview of this event. And now that they have given me the keys, I am trying to think about how to make it as productive as possible.

Here is a draft of what I might say to open the conference on April 12, 2013, at 9 AM CDT:

Thirty-five years ago my father bought my family’s first personal computer. Eighteen years ago my former roommate Paul Mende told me:

Hey, this World-Wide-Web thing is revolutionizing scholarly communication in Physics via the http://arxiv.org/ web server. You should check it out.

Fourteen years ago I noticed that both of my mentions in the then-most recent Foreign Affairs--one by Jagdish Bhagwati and one by Paul Krugman--had come not from things I had published but from things I had thrown up on my website...

Twelve years ago I decided that somebody should go all-in and attempt to win the intellectual influence game via the strategy of always-putting-something-new-and-interesting-up-on-the-web. And as a guy with tenure at an institution that seemed to me to be the global optimum, I was one of the few people in the world who could do so with no significant possible downside risk.

And today here we all are. Everybody here has at least tried hard to significantly raise the level of the debate wherever their feet have trod on matters economic, and tried hard to raise it significantly above what it would otherwise have been--to evade the dumbness filters that surrounded us and impeded communication and education. We all seem to have come remarkably close to maximizing the win, for some value of “maximize” and “win”, at least for ourselves if not for our institutions and our causes.

But the struggle is ongoing. There is still an enormous amount of headroom.

How do we keep maximizing our collective personal win? What are the tools, networks, and communities that we are going to use and deal with in the future? How will we deal with the other communities and modalities of communication on our borders. We, of course, seek total universal domination. Is that realistic? If so, how? If not, for what should we settle and how should we settle for it? And how do we maximize the societal win?

---

Does this draft introduction have a chance of triggering the kind of conversation on April 12 that it would be useful to see? If not, what should I say instead? And what else should I say to open the conference? Suggestions welcome.

The conference is invitation only, alas! For our budget and our space is very limited...

Save the Date: Economic Bloggers Forum 2013

I did not want to scoop Brad DeLong since he is organizing the event, but he has now announced so I will reiterate: the Fifth Annual Kauffman Foundation Economic Webloggers' Conference will take place on Friday, April 12.

This conference, being hosted for a fifth time, is an invitation-only event to top bloggers in the economics sphere to discuss a variety of topics in economics, finance, and journalism. We'll post updates as the event draws closer. But for now, save April 12 as the date for the live stream--we plan to stream the event here on Growthology the same as last year (video and agenda from last year is here). Given the audience, expect a lot of live tweeting during the day.

 

Negative Press, Boards, and Strategic Directions

To me, that bad press can be a catalyst for strategic change in a business is not a surprising research finding. That bad press can have no influence when a business is doing well is again not terribly surprising. These are some of the conclusions in a forthcoming paper, “Burr Under the Saddle: How Media Coverage Influences Strategic Change,” by Michael K. Bednar, Steven Boivie, and Nicholas R. Prince. The authors look at 40,000 articles covering 250 firms over the course of five years, analyzing how boards react to negative media coverage (abstract and gated copy of paper here; press summary here).

What I find a more interesting finding is that bad press has a greater effect on strategic change when a businesses’s board of directors is comprised of outsiders—independent board members with no previous ties to the firm’s management (such as being family, friends, or former business relationships).

I’m wary to post blasé “practical” advice like “be aware of the outside/insider composition of your startup’s board of directors.” I don’t find this kind of messaging particularly helpful, and I’ve sat through infinite numbers of talks/presentations with the bullet point “choose board wisely,” so I’m pretty sure people are hearing this message already. But when I read things such as Wasserman’s The Founder’s Dilemmas and see all the instances of founders being replaced at their startups (for various reasons), and then I see this research about insider and outsider dynamics in boards, I’m inclined to add some nuance to board composition and “choosing wisely.”

As your startup grows, it seems very likely that a board would grow to include both ‘insiders’ and ‘outsiders.’ These groups will have different behavioral characteristics, with one of them being the interpretation of bad news. Perhaps you can extrapolate this bad news sensitivity to a broader behavioral characteristic: outsiders place more weight on outsider information. That is, they are more willing to take cues from outsiders, for better or for worse. So as the composition of your board changes, perhaps adding more outsiders as the startup grows, be more prepared to deal with reactionary changes from any bad press. The struggle over strategy will be harder than it was in the beginning.

Is More Better?

It’s no secret that we at the Foundation think entrepreneurial activity benefits the economy. Working off this premise, we often promote policy changes that lower barriers to becoming an entrepreneur along the following lines of logic:

However, this series of events is far from obvious. Another outcome seems entirely plausible:

Essentially, more companies is far from given to be an unalloyed good so we cannot be certain that a simple increase in the number of business is a net positive.

An excellent new empirical paper by Hombert, Schoar, Sraer and Thesmar presented earlier this month at the American Economic Association’s annual meetings investigates one such instance of lower barriers (the paper, Should We Make it Safer to Start a Business?”, is not yet publically available pending publication but we will post it here when it is). They measured the effect of a decrease in cost of entry by studying a large French reform intended to encourage unemployed people to start businesses. Not unsurprisingly, they found that more firms entered after the reform went into effect. They also found the new firms not to be of lower quality than start-ups previous to the reform—they are just as likely to succeed and to create new jobs. They do find evidence of crowding out--that is the number of incumbent firms decreases in response to the influx of new firms, with the job creation from new firms offset by job destruction in dying firms. However, they also find those old firms that exit to be less productive than their young counterparts, a narrative consistent with Schumpeterian creative destruction. So on net, they find that lower barriers don’t actually create more jobs but do improve productivity.

This seems to be a clear tally in favor of the first narrative. I wonder, however, about two more things. First, while it seems clear that the effect on welfare is positive given the large productivity gain, I still wonder how these gains compare to the cost of publically financing the reform and the implications for Pareto efficiency.  

Second, surely not all barriers to entry are equal? That is, removal of some will result in more innovation while others will result in more productivity while still others may lead to the second narrative above. Perhaps the paper would benefit from a theoretical model that would clarify the mechanism.

Downsides to Failure Tolerance

Ramana Nanda and Matthew Rhodes-Kropf have a working paper out titled “Innovation and the Financial Guillotine” (where guillotine refers to investors’ willingness to kill an investment). Having your startup choose a strategy of high tolerance for failure in theory helps your startup and your employees feel like they can take risks to try something very innovative. The same goes for investors (and governments)—choose to tolerate failure to promote innovation. When Nanda and Rhodes-Kropf modeled investment choice decisions and their interaction with failure tolerance, they find that choosing a broad strategy of failure tolerance leads to funding less radical innovation and more incremental innovation. The issue is that by promoting failure tolerance, you make it harder to kill projects. Not being able to kill projects is a risky proposition.

My interpretation is that simplifying your startup or investment choices by employing a broad strategy of failure tolerance is not a good replacement for project-by-project considerations (the authors sort of get at this but don’t say it directly, so I could be wrong).