Tumblr Promoted Posts: Exploring Customer Segmentation For Pricing
I’ve written about both forms of Tumblr promoted posts previously. Tumblr is now in the midst of revenue discovery, where they attempt to put together a diversified revenue portfolio; a mixture of ad, transaction and subscription pricing strategies meant to increase revenue yet decrease revenue volatility over time. An important tenant of a well constructed revenue portfolio prices goods or services in a manner commensurate with value for each customer segment.
Tumblr is failing to adhere to this last point with its current static pricing model across all customers, where pinned and highlighted posts cost $5 and $2 respectively, independent of value as measured by impact, engagement, or potential reach. For example, a new Tumblr user with zero followers, myself with 592 followers, and MTV with say 500,000 followers all pay the same price for a pinned post today, but the promotional value for each of us is not equivalent. There are many ways to solve this issue employing a combination of fixed and variable pricing components.
The fixed pricing component sets the price floor by addressing setup costs and minimum customer value in a single number. Remember, cost represents not only the price of of setting up the promotion (negligible on an individual basis), but also the price for serving non-paying customers in a freemium model (huge in aggregate). Similarly, minimum customer value has nothing to do with Tumblr’s cost basis and everything to do with how the paid services are perceived. For our example, we’ll use Tumblr’s aforementioned current static prices as the fixed component.
The variable pricing component should be dynamic and based upon transparent, easy to understand key performance indicators (KPIs). For our example, follower count seems to be the best Tumblr KPI proxy for value via potential reach, however Tumblr could also measure the exact impact for each pinned or highlighted post by giving customers exact impressions (# of times the post has been served to unique users) or engagements (# of times a post has been unpinned or likes/reblogs from the pinned position or a call-to-action link within the post has been clicked). I’ll address Tumblr analytics in a future post.
How do we combine the two components? Start by creating a frequency by follower count histogram for all 60+ million Tumblr users to understand the universe we are playing in. There’s enough real customer data today to create another histogram using only actual promoted posts customers. We’re visualizing the data to come up with a customer segmentation scheme based upon follower count that makes sense**. Our options could be separating the universe into follower buckets or bins (eg. a descriptive distribution statistic like deciles) each of set price or making price mathematically tied to the exact follower count (eg. CPM). Given we don’t have Tumblr’s data, let’s go with a modified CPM approach where CPM math only applies for accounts with greater than 1000 followers (I’m assuming most accounts have fewer than 1000 followers). How you determine your CPM is up to you, but I suggest using data to come up with engagement targets at every follower count, and backing out the number from the targeted pro-rata costs offset by that particular stream (eg. if we’ve set the customer engagement targets at each follower count quartile - 1st is .5%, 2nd 3%, 3rd 2.5%, 4th 1.2% - and want pinned posts to account for 25% of costs, we can back out the CPM accordingly).
For example purposes, let’s just say our backed out CPM is $3. We now get the following:
Pinned Post Revenue = Fixed + Variable
Pinned Post Revenue = Fixed + CPM * (Followers/1000) for follower counts above 1000 otherwise it’s just Fixed
For the spreadsheet inclined, it’ll look something like this (zeroes out the variable component for follower counts below 1000):
Pinned Post Revenue = $5 + $3 * ([if Followers > 1000, Followers, 0]/1000)
So, all users with fewer than 1000 followers would pay $5, but MTV, with 500,000 followers, would pay $1,500. This pricing model now seems reasonable and commensurate with customer benefit. Additionally, the above complexity can be clearly messaged to the customer via a singular number with a simple explanation.
The number of levers available for customer segmentation pricing are numerous. To keep things in perspective, when exploring pricing, just recognize that product and revenue portfolio design go hand-in-hand to drive value at a premium. You can alter any aspect of either to achieve your goals, just err on the side of messaging clarity on behalf the customer and fiscal sense on behalf of your costs.
** A deeper analysis would be a histogram of Tumblr Notes actions by position in the stream - 1st, 2nd, 3rd, 4th, etc. This way, you can come up with real data regarding the “lift” from a pinned or highlighted promotion by comparing apples to apples. What better way to market the value f promoted posts than revealing the value. Eg. “Pinned posts receive, on average, 350% more reblogs, 500% more likes and 200% more replies than unpinned posts. Highlighted posts receive on average 50% more engagement against posts in their same position in the stream.”
Tumblr Notes Design Ethos: How Threaded Replies Can Work
In the hopes of getting someone to fix Tumblr’s notes — either inside Tumblr or an outside developer — I am offering the following proposal for a how I’d like Tumblr notes to work. (By the way, anyone who is interested in implementing, give me a call.)
Today’s notes work like this:
- A user, say gbattle, reads something of mine, let’s say in his Tumblr stream. He takes an action, let’s say he replies to it (which I have enabled), and that gesture is reflected in the list of notes associated with the post.
- Later on, others looking at the post — either in the stream or on the post’s page — can see his reply along with all the other social gestures left behind by others, including other likes, rebloggings, and replies.
The problem is that I can’t reply back.
Imagine a product called Notr, that replaces the notes section of Tumblr themes. It would interact with Tumblr’s API to fetch notes, but it would also keep track of the relationship between notes implied by nesting. Where Tumblr’s notes system is inadequate, or blocks the creation and management of notes, Notr would conserve the notes in its own database.
Note the little talk balloon next to gbattle’s reply, which is provided by Notr here. I could re-reply to gbattle by clicking the balloon, and typing in some text:
And then others could also reply to that thread:
The implementation would be something like Disqus, but integrated with the Tumblr notes system, to the extent that it is possible.
Obviously, it would be simpler if Tumblr would implement notes this way, and we could all drop the amazingly annoying use of unintegrated Disqus. Alternatively, Disqus could implement this as a version of the product.
(This post is related to this gripe, from earlier today.)
It’s impossible for me not to comment on comments where I am the subject matter.
The teeth of Stowe’s gripe fit my wounds also.
Namely, it’s well known that David Karp/Tumblr’s design ethos involves rewarding positive/quality communication via accumulating one’s commentary on his or her blog as a form of ownership. This ideal held firm for likes and reblogs, but does it hold true for replies?:
- We can only show our positive support via likes, and what we like is publicly viewable via a stream - there is no gesture-based form of negative expression (thumbs down, dislike, flag, zero stars, etc.)
- We can reblog and give our commentary, but that response lives on our tumblr, hence we reap what we sow in terms of quality
- We can write a reply in the notes, but it is not “owned” by us (the repliers) and lives with the original content author’s content, we can also edit or remove any reply we’ve submitted ad nauseum
It would seem Tumblr replies break the beauty and essence of the original design ethos, so why hasn’t the trolling onslaught begun? This is where the nuance of Tumblr’s design comes into play. Given one can only execute a reply from within the Dashboard and Tumblr enforces either a bi-directional following relationship among the author and replier or you have followed a user for more than two weeks, a de facto authentication/ordainment exists for the replier. The author can still elect to remove replies or block individual replies/repliers.
Finally, with respect to all Tumblr Notes, Tumblr reserves the right to throttle delivery by serving them up in batches within the Dashboard.
All of the above is necessary backstory before discussing Stowe Boyd’s reasonable suggestion.
For Stowe’s threaded replies to work and also adhere to the Tumblr Design Ethos, a few enforcements are necessary:
- Authentication for involvement in reply threads are based upon the relationship between the replier and the original content post author, not among the repliers. Specifically, each replier must have a bidirectional following relationship with the content owner or be a follower of the content owner for at least two weeks. Hence, in Stowe’s example, he should absolutely be allowed to reply to me (gbattle) on an indented, threaded basis.
- Just as all Tumblr Notes can be throttled, any subset of threaded replies can also be throttled to improve visibility in the Dashboard. Users can optionally expand threaded replies to reveal thread details.
- No replier can reply to his or her own reply. This safeguards against thread spam and unnecessary threading.
- Once a threaded reply has been started, the parent reply can no longer be edited. The preserves context for the threaded discussion.
Like Stowe, if Tumblr, Missing-e, or an independent browser extension developer would like to discuss the details on implementing the above, I’d be happy to help.
Note: Given the above restrictions, I don’t think Disqus integration into the Dashboard, however cool, would support everything necessary.
Let’s review what I suggested regarding highlighted/promoted posts on February 12th (back when it was only $1, not $2):
For my $1, I’d love to make sure that I get exactly one impression per user such that the post isn’t simply emphasized visually, but guaranteeing the impression via enforcing each unique delivery. That’s something I’d pay for, such that I don’t have to worry about the timing of my post on Tumblr because my followers are guaranteed to see it at least once.
The idea here is that user is paying to break the reverse chronological flow of the Tumblr Dashboard, arguably, the single most engaging feature of the web (if you take all of the time-on-site statistics and recognize that for Tumblr, 99+% of that time spent is in the Dashboard vs. Facebook where it could be in games, Timeline, News Feed, apps, stalking, etc.).
There’s a huge value in breaking the need to time the receptivity of your posts (some have built entire businesses around optimizing this). I also mentioned the next phase of this product after paid impressions should be encouraging reblogs for the viral spread of content across the social web:
For my $1, I’d love to have a way to encourage reblogs, which could be done by allowing a certain number or time window of reblogs to also gain the promotional certification. Each generation of reblog, the promotion fades into the background until, say, 4 or 5 levels deep, it disappears completely - the utility of the promotion is complete. Or, maybe you can add more money to extend the levels of promotional ability - $1 for every level. If I want to enable 5 reblog levels of promotional juice, I pay $5. I can buy as much AdWords or Sponsored Stories juice as my wallet can handle, why not create the authentic sharing version of this within Tumblr?
Given Tumblr’s current pin implementation, I think I would simplify my suggestion. Every reblog of a pinned post receives 24 hours of highlighted post status. That way, the pinning isn’t inherited by people who don’t follow you, but you and your rebloggers receive some promotional lift. Allowing the original poster to pay for added pin time to their post and the subsequent highlighted reblogs allows there to be a variable component to what a user pays. For example, the US Open should be able to run a 10 day promotion of a particular tennis championship post and the reblogs at purchase time. It need not be linear either, 10 days might be $500, not $50. Segregating your individual vs. corporate customers via fixed and variable pricing is huge when putting together a revenue portfolio.
My current contrarian suggestion for Tumblr to explore revenue possibilities is exploring how to break the singular content type post (photo, text, audio, video, chat, link) and allow for heterogeneous posts containing a mix of all types. Think of it as a collection of content rather than a post in a stream. The idea of a collection of different content on one page, with a beginning and end, creates a narrative platform that is missing on Tumblr that others have flocked to Pinterest et. al. to express themselves. You’ve mastered curating to the stream, now curate to a singular URL (which, of course, can itself be put into the stream).
And finally, it’s about time for Tumblr to start tracking outbound links like Facebook and Twitter. Understanding the viral spread and offsite behavioral intelligence around your content is critical to going from a creative platform to a discovery platform.
Just my morning $.02.
[climbs back under sheets]
If I’m looking for gold nuggets contained within mountains, don’t hand me tools to aggregate more mountains and call that a useful discovery tool! Hand me a sieve already dammit!
Lauren Indvik via Mashable
Thirteen percent of online Americans use Twitter, up a full 5% from November 2010, according to a Pew Research Center study released Wednesday.
Much of that growth came from younger adults, Pew found. Nearly one in five U.S. Internet users ages 25 to 34 use Twitter, up from 9% in November. Fourteen percent of users between 35 and 44 also now use the service, up from 8% a half-year previous.
But Twitter isn’t exclusively the domain of young adults; 8% of participants between 50 to 64 and 6% over the age of 65 consider themselves Twitter users, respectively.
Perhaps most intriguing is the demographic gap: 25% of African-Americans and 19% of Hispanics use the service, compared with 9% of Non-Hispanic Whites. Perhaps linked to the use of entertainment and cultural leaders?
Stowe, that’s a correlated explanation, but not a causal one. Let’s go deeper.
The real explanation is as historical as the results are pervasive. African Americans and Hispanics have always over-indexed in terms of Twitter usage, or any usage primarily driven by mobile devices (see PEW’s research study regarding SMS usage). It’s not simply over-indexing by percentage of DAUs/MAUs for the demographic, but there’s been reports that African Americans account for 25% of all U.S. Tweets in terms of volume while being only 13% of the population. This last fact can be seen anecdotally when witnessing the pervasive number of urban themed hashtag memes.
As I’ve said previously, Twitter is the black 4chan.
Why? Entertainers and cultural leaders are the effect, but not the cause. We have to go back to the 80s & 90s, back to the earliest phases of the consumer facing web and the so-called “digital divide” between white and African American/Hispanic communities in terms of computer/internet adoption. The problem with the digital divide argument is that it wasn’t inclusive of all digital usage and only measured computer usage. Urban communities have smaller homes, with apartments that might accommodate but one computer. If present, that computer, typically, would be in a public area of the home - a family room, a living room - thus all usage would be “public.” However, at the time, most of the internet’s usage growth occurred in private, where people tried on identities, utilized pseudonyms, connected with strangers, consumed the unthinkable, and engaged in fantasies in a world that began and ended at the dial tone.***
Beepers, Motorola 2-Ways, Sidekicks, Blackberrys and now smartphones bridged this privacy gap, allowing urbanites to enjoy in device driven fantasy. It’s from this you get beeper codes, technology as fashion accessory, and friend/follower accumulation as proxy for social proof. If it worked for suburban doctors, it also worked for the urban street pharmacists and those who postured as either one. Overlay the ostentatious, aspirational, entrepreneurial and self-promotional culture of hip-hop and you create a gigantic viral loop from the bedroom to the streets to the clubs to the small and big screens propelling mobile technology to iconic status in black America. Clearly, when Twitter hit in 2006/2007, you have an entire community predisposed and trained on short form communication for nearly twenty years. Though there was a lag, the real African American usage on Twitter took off in 2008/2009 with the native Blackberry applications that were always ad-supported as hip-hop culture is synonymous with embedded brand messaging.
So, the dirty secret of the mobile phone and app industry is that African Americans (and Hispanics, females, and southerners demographically according to PEW) are a leading indicator for mainstreaming mobile social features. There’s a reason why carriers created AMP’d and Boost Mobile instead of Skinny Jeans or Fixie+Facial Hair Mobile - historical content consumption/generation patterns, embedded knowledge, and fertile established viral graphs.
*** After SixDegrees introduced us to the concept of a social network in 1997, the next three social networks of size were all minority focused ones created by Community Connect in 1999 - Asian Avenue, Black Planet and Mi Gente - all predating Friendster by 3 years. It should come as no surprise that this early adoption translated to over-indexing on future social network services also.
Fred Wilson has a great post this morning on hiring best practices that has sparked some great suggestions. I wrote a comment (pasted after the jump). Differentiate your company by focusing on the details - creative companies attract creative people, operationally efficient companies attract operationally efficient people, and so on.
Shorten the gap between the culture you espouse and the ethos you practice.
Your hiring process and practice says everything about your company. As many have detailed, you will pass on many more people than you hire, and those people you reject will talk and either be advocates or detractors based upon their experience. What Fred detailed are indeed best practices, but remember, best practices don’t define the exceptional, they define the minimal acceptable activity. That’s where you start, not end.
So how do you become exceptional in the age where hiring is an arms race for talent?
Though on-demand and just-in-time work well for scaling technology resources, you cannot do this with human resources. You have to be looking at all times (and in all places) independent of your hiring needs, because talent doesn’t just appear on your doorstep when you need it. This involves a CEO taking stock of the culture and what he or she will or will not accept.
- Is your company vision and values reflected in your process?
- What does the job spec say about your company’s identity outside of the role itself?
- If you expect speed and efficiency from your employees, does your process reflect that value?
- Are you willing to hire based upon availability of stellar talent rather than immediate need, and if so, how do you integrate them into your company?
- Are you willing to work with talent remotely?
- If you build remote teams, how often do you visit them and how often do you bring everyone together?
- What types of non-dollar rewards/benefits are you willing to employ to acquire great talent and align them for success?
- Post the offer, what is your onboarding process for new hires - make everyone learn code or do customer service or a leadership seminar/adventure - what about the onboarding process binds everyone together?
- Will you poach people from other companies directly?
- How do you deal with other companies poaching your people?
Knowing who you are as a company is a candidate filter for both sides. If you twist in the wind to accommodate, people will know and respect you less. Not that there’s no room for negotiation, but your values should define the boundaries.
In Defense of Hype: Haterade Might Kill You
1. I am not a short seller.
It isn’t my job as an entrepreneur, and it’s not my job when investing, to pick the losers. So it’s not worth spending a lot of energy trying to prove that a particular company isn’t going to make it, most aren’t. Much more important is trying to figure out those rare few companies and few individuals that might break through, and the ripple effects if they succeed. I try and focus my effort spending time there.
There are both ethical and structural reasons why Nabeel is not a short seller (and I love that he used that specific term to emphasize picking losers for profit). Ethically, it’s just in bad taste to bad mouth every startup out there when you are in the business of attracting deal flow. Structurally, there is no market to capitalize on one’s ability to pick losing startups, so no matter how much one wants to attack Twitter, AirBnB, Pinterest or the rising star du jour, that position isn’t actionable in anything other than hot air (or journalism). Hence, much of the negative banter is a fool’s errand when the only logical response is either investing in or, better yet, creating the better mousetrap to express your position.
This reminds me of a great phrase: “There’s a reason why you’ll never see statues of the world’s greatest critics.”
My Favorite Tumblr Feature
The Tumblr ethos dictates a design discipline whereby, over the past 5 years, they have removed more features than are on the current site. Tumblr founder/CEO, David Karp, has said on record that due to the often ignored maintenance cost of feature creep, subtraction must remain a high priority action item on the product roadmap, even for proven popular services.
Many might guess that my favorite Tumblr feature is the reblog, with its embedded attribution and annotation, as it was the first of any of the RE-series of features released by a service itself. Specifically, yes, the Tumblr reblog predates Twitter’s official retweet feature in terms of release date by years and reblog is the direct predecessor to Pinterest’s repin (I’m sure there are many other re-gurgitated versions I’m forgetting too).
However, there is one feature, one singular, beautiful feature, that I have fallen in love with over the years. So much so, that I wish this feature lived on every site. It is so simple, yet so useful, that it is hard to imagine Tumblr without it.
The Up Button.
Yes, the little button that appears in the upper right corner of the Dashboard and, upon clicking, scrolls you to the top of the page. Seems silly? Too simple? Let’s explore the hidden beauty of this little button, why Tumblr has it, why no other top 25 site would dare employ it, and who might benefit from offering it outside of Tumblr.
This button’s beauty is not stand alone. It is paired, of course, with the infinite scroll capability in the Tumblr Dashboard. Infinite scroll enables Dashboard users to engage in a level of direct, in-page addictive consumption few other sites accomplish due to pageviewitis - the myopic need to generate pageviews to serve banners, skyscrapers, badges and whatever the IAB standard unit dujour happens to be for content monetization purposes. The Up Button and infinite scroll keep the Dashboard acting as an application proper rather than as a gigantic redundant pamphlet (find a top 1000 content site that doesn’t force awkward, annoying article pagebreaks for pageviews). Given the rampant abject gaming of the metric, Karp & Co. decided long ago to subtract pageviews as a relevant growth metric for their service.
Tumblr realized the maintenance cost of pageviewitis to Tumblr is user joy.
Even the most casual Tumblr user has, after a long Dashboard session, looked for the little Up Button on other long-paged sites, only to realize “duh, I’m not on Tumblr.” You would think by now, given Tumblr’s popularity, some company would create a universal Up Button that masquerades as an analytics add-in like GetClicky, Google Analytics or Chartbeat, or as a natural feature for say Disqus or Facebook comments or a generic Chrome/FireFox/Safari/IEXX browser extension, but one have.
Or maybe Tumblr should release one itself.
Or maybe I should.
[:insert image of me opening of VIM here:]
Analyzing Dropbox For Teams Product & Pricing
I recently interviewed with Dropbox and was asked to analyze Dropbox For Teams pricing for a 20 minute phone discussion. Given I found the exercise and discussion fun, I decided to post my results here [Full disclosure: it wasn’t a mutual match and there was no NDA, ergo my posting this, however, out of respect, I won’t share the interviewer’s name or response to my analysis].
Let’s set the playing field. Dropbox, a cloud storage SaaS solution, has 45 million customers, of whom, 1 million are business accounts. It utilizes a classic social freemium model of increasing engagement whereby:
- Basic accounts receive 2GB of free storage
- Basic accounts can earn up to 8GB total storage in exchange for referrals (250MB for each new free user) and testing new products (infrequent, varies)
- Pro accounts offer 50GB or 100GB of storage, payable monthly ($9.99 or $19.99 respectively) or annually ($99 or $199 respectively)
- Basic and Pro accounts have individual storage quotas; Teams has a shared storage quota
- Teams is $795/year for 5 users and 1TB, each additional user costs $125/year and adds another 200GB of storage
As tempting (and popular) as it is to analyze Dropbox for Teams pricing from the viewpoint of cost vs. competitor costs on a per-GB basis, this is the wrong approach. Always price SaaS on the value to the customer, not the cost to deliver, especially if you have the type of embedded network effects/marketing that Dropbox has. In Bezos-ian parlance, “start from the customer and move backwards.” [Note: as arguably the largest AWS/S3 customer after Zynga’s departure, Dropbox’s costs are incredibly low due to scale and scope economies - plenty of gas in the tank to thwart price-focused competitors … unless that competitor is Amazon itself!].
So, who is the Teams customer? How big are the teams? What’s the usage look like?
Without any supporting data, we can surmise one thing about the customer. There is an upper limit on how much data can be stored on any singular computer on average, hence there will be an upper limit on the number of maxed-out users for any Dropbox for Teams installation. Specifically, the average new laptop/desktop harddrive size is 500GB-2TB, so there’s only so much shared storage to go around as the team size increases. So, an aggressive limit might be team sizes from 5-50 members. After adding an additional 5 users to a team size of 10 and storage limit of 2TB, any additional users added are simply for shared storage considerations as the 200GB/user added doesn’t really matter given the inherent desktop/laptop storage limitations. Access trumps size in terms of value for Teams customers.
Given these hypotheses about the customer, what can we determine about price?
Price is more than just a number. Price is multifaceted; it can have a fixed component, variable component, payment method and payment schedule. From a small business perspective, the method of payment (credit card, invoice, deferred interest payment, etc.) and scheduling of payment (monthly, quarterly, annually, 30/45/60 day PO, etc.) can be as important as the numerical price given the underlying cashflows for the business. Offering more payment options may open up more opportunities, not only for cashflow sensitive customers, but also opportunities for higher pricing due to customer segmentation - more paperwork, delayed payments, more pay-as-you-go all lead to higher prices.
Access and payment options add to customer value, what about service?
Currently, Dropbox runs a self-service on-boarding for all new customers and Teams customers get phone support. Is there a white glove on-boarding service to be sold? One of the many things I learned from being beat up by famed salesman/investor Jordan Levy from Softbank Capital was the golden rule of $25K - if you cannot earn at least $25K/year/customer, you cannot afford to have an in-house sales team. So the question is, is there a way to package Teams for much larger customers as an Enterprise product containing many Teams to hit that $25K/year/customer target? Could Dropbox beef up the the price if they added tiered SLA’s, PCI-DSS, HIPPA, Sarb-Ox, and other compliance services? Though an Enterprise service presents longer sales cycles, the resultant customers would also present the least variable/most sustainable cashflow for Dropbox. Hmmm…
Anything else to consider?
Finally, on price, let’s re-examine that variable pricing component of $125/additional user. As I have detailed, this singular variable pricing sets a value floor for the customer as they don’t gain anything from the additional space, only the shared access. I would explore creating tiered variable pricing that represent the actual value to the customer, which isn’t 200GB/year per user, but additional shared access to that max sized drive of 2TB.
Bottom Line for Dropbox Teams:
- Focus on target customer size based upon value
- Add more payment options to accommodate small business cashflow dynamics
- Add tiered variable pricing to reflect access not storage capacity
- Explore a white-glove Enterprise solution that bundles Teams