I worked with two companies in the past month that needed a way to model viral adoption as part of their effort to understand the impact of a strategic decision to deliver part or all of their technology under an opensource license. Having such model enables them to understand the critical variables and success-factors, set targets for them and then measure as they embark on their strategies.

I set out to make a generic model that can at once be used at both companies. Hoping it may be of use to others as well, I will publish it here. I had much benefit from articles published by David Skok and Andrew Chen and try to use their terms and insights as much as possible.

The basic assumption for viral growth is that every new user of a product becomes an ambassador for that product and will convert others to start using that product as well. There are two variables that allow you to create a model for viral growth:

1. Viral Coefficient: the total number of new users generated by an existing user
2. Viral Cycle Time: the amount of time it takes before all these new users have been generated

Viral growth can happen in many ways, from word-of-mouth, to active solicitation and anything in between. This model makes no assumptions about that, it just measures the results. Here is an example in which the unit of time is set in weeks, the starting number of users is 1000, the viral coefficient is .3 and the cycle time is 3 weeks:

The model reflects that the viral impact starts in the first period (week) after a user has started using a product and lasts throughout the duration of the cycle. If new users join in the first week, they start working their viral magic in the week after that. So, the 1,000 users in week zero will deliver 300 new users (coefficient is .3) over 3 weeks (cycle time is 3), or 100 new users in week 1, 2 and 3. The 100 new users in week 1 deliver 0.3*100=30 new users,  10 in the weeks 2, 3 and 4. The 110 new users in week 2 deliver 33 more users, and so forth.

As the coefficient is less than 1 in this case, the number of new users per week trends to zero over time. If the coefficient is 1 or larger, the viral engine will keep on giving.

The pattern for the number of new users per week (in C15:L15) is that each number is equal to the sum of the numbers of the previous three weeks, multiplied with the viral coefficient and then divided by the viral cycle time. In the next image you can see how that pattern is turned into a spreadsheet formula:

The formula in cell H24 is:

=SUMIF($B23:G23,”>=”&(H23-$B$20),$B24:G24)*$B$19/$B$20

This formula defines the range of values that can be summarized as *all* previous weeks ($B24:G24), but only includes the numbers in the weeks for which the week number is 3 or larger (columns E, F and G).

Viral Formula Excel Model (Click to download)

You can download the Excel spreadsheet with this model (see above). I have also published the model as a Google Spreadsheet. Take a look and play with the values for Viral Coefficient and Viral Cycle Time, to see how they impact the growth of your user base. To inspect the formulas, select a cell and then hover the mouse over the formula area on the right bottom of the browser window.

In the next entries I will add further metrics to this model that have an impact on viral distribution.

]]> http://www.markdevisser.com/2010/02/a-spreadsheet-model-for-viral-growth/feed/ 2 http://www.markdevisser.com/2009/11/lots-of-innovation-in-search-engines/ http://www.markdevisser.com/2009/11/lots-of-innovation-in-search-engines/#comments Sun, 22 Nov 2009 00:24:43 +0000 mdv http://www.markdevisser.com/?p=71

There is no product category that appears to be as locked up as Search. Google’s market share is a massive 72%, with the spoils going to Bing, Yahoo, Ask.com and a host of others. Yet there is a flurry of innovation going on in search and it is one of the very few segments of the technology market that is attracting abundant venture capital investments.

The innovations are driven by the emergence of the real-time web. It used to be that search engines crawled sites daily to update the indexes, but now the aim is to capture information, index it and present search results in real-time. Google’s innovation was page-rank to determine what information was the most relevant, which worked fantastically for webpages. That model needs an update for the content of the likes of Twitter and Facebook, where the relevance is determined by the number of friends of the poster, the number of times a message is repeated, the time at which it was posted and at what location.

And beyond that trend there is still the hope that one day the search engine will not respond to the character strings that I type, but to the question I have in my head, with full understanding of my interests and context. Also in that field, semantic search, progress is being made. In fact, for real-time search it is critical that these efforts succeed because 140-character messages give too little context to enable meaningful search results with traditional approaches.

I asked a number of the incumbent and new search engines for information about Al Gore, which I thought would be a topic with history, but also sure to be a lively topic in the real-time web (Update Nov 22: I wrote this on Saturday afternoon, Al Gore appeared on Saturday Night Live, quite a coincidence. I reshot the screen shots for this story on Sunday):

As a baseline, let’s ask Google:

Unless you have lived under a rock in the past decade, this mix of relevant web pages, advertisements and news about the topic will look familiar to you. But even in this well-known page there is a lot of innovation going on, click for instance on the “Wonder Wheel” in the left sidebar (click “show options” if you do not see the sidebar), it will show you topics that Google thinks are related to the topic “Al Gore”. Notice also that one can check for information on a topic that has become available within the last hour – real-time anyone?

Take a look at the search results presented by Bing:

The Bing results closely match the Google results, but then Bing adds search results in a range of additional categories like “El Gore Facts”, “El Gore Quotes”, “El Gore Biography”, “El Gore Accomplishments” and “El Gore Book”. This moves the Bing results in the direction of semantic engines like Hakia (discussed below).

Here are the results from Yahoo:

Yahoo offers about the same search results as do Google and Bing, but then it adds links to news sources like ABC and Huffington that provide remarkably good content. Yahoo also offers useful related searches.

So now take a look at some of the Real-Time search engines, starting with TweetMeme:

Tweetmeme counts the number of times tweets link to a URL and lists the URLs with the highest number of tweets that refer to them. This metric establishes relevance to the links, it makes it easier to make sense of what is discussed on Twitter. But that strength is also its weakness, I cannot learn any more than what gets tweeted, and from my experiments so far that does not teach me more about my topics than the gossip that surrounds them.

Tweetmeme shines in its own categorization. I follow its Technology category and often find useful information there that I have not found before.

Another way to find information that is currently discussed on the real-time web is CrowdEye:

CrowdEye has an algorithm that assigns a number between 1 and 100 to give relevance to the poster of the tweet about your topic. Like Tweetmeme, CrowdEye is a great site to browse and discover topics that are Hot in the Twittersphere, but it does provide less context than Google, Bing and Yahoo.

Another search engine for Twitter that I like is TWazzup:

TWazzup has many slick features, such as the ability to Retweet straight from the listed tweets, saved searches (very convenient for complex boolean searches), list of influential twitterers on the topic of your search, and more. The homepage of TWazzup is very sparse, I use this bookmark to populate it with the major news of the day.

There are a number of additional search engines which track what is being discussed on Twitter, check these links to see which one you like best: DailyRT, which prioritizes topics by their number of retweets, Topsy, another search engine that weighs the influence of the posters, and Mozzler, Scoopler and OneRiot , which take information from Twitter, Digg, Delicious and more.  And Bing has a Twitter search page with the same functionality:

Semantic Search adds capability to search engines to parse the meaning of the information it indexes. Hakia is an impressive example of this:

It provides information about Al Gore in categories such as News, Biography, Awards, Speaches, Quotes, Blogs, Interviews, Controversies, Contact Data and more, much more. So far, the information I have found on Hakia has always been genuinely useful. Other sites that offer semantic search capabilities are Yebol and Powerset, absolutely the best search for Wikipedia.

And last but certainly not least is the Ellerdale project. The website is still very much a beta, but the potential of its technology platform is easily recognized:

Ellerdale parses the internet to discover “topics”. It manages to recognize that Schwarzenegger, Governator, Arnold and California Governor are the same thing. Even more exciting is that it can apply that semantic understanding to index the real-time web. It can see that a Twitter entry about “arnold” is really about Schwarzenegger and show it in the search results for the California Governor.

It did not take long at all to see the first service pop up that categorizes lists and helps you discover them, see how much they are subscribed to, etc. It is called Listorious and offers the ability to “discover the best Twitter Lists” as value proposition. It is clear that Twitter lists will define the new aristocracy in the Twitter universe. Andrew Muller has a very good post on how this may end up concentrating power for the already influential and if you study the lists on the Listorious first page, that is already playing out: there are no surprises there at all.

Twitter is not known for feature-creep, but last week it introduced a significant new feature that will have real impact on how Twitter is used: Twitter Lists.

So what are Twitter Lists?

They are named groups of people who have Twitter accounts. I created a list for folks who write about Open Source (link) and did the same for Marketing (link). I made these list public, so anyone can see who I track in these lists. I also subscribed to a list called “foss” by Stephen O””””Grady. It is a humble start, but it gets the idea across. I will work to move my TweetDeck groups over in the coming weeks.

Twitter says that Lists are “A great way to organize the people you follow”, but I could add people who I do not follow to my lists. I suspect that this will be a big disincentive to following people in Twitter. Seeing the messages in the context of the list in which they occur will make them more meaningful than the generic stream of messages from all people I follow mashed together.

I am sure the folks at the Twitter ranking systems are already busy recalibrating their algorithms.

Apart from the convenience of the Lists for following what is being said in the Twitterverse, what do Twitter Lists mean for you if you want your message to be heard and your brand to be known?

1. Get used to the idea that the value of the number of followers you have will quickly deteriorate in terms of your social capital. If you are not on someone”s list, (s)he will not see your posts, even if that someone is a follower.
2. To be included on someone”s list, you”ll have to be recognized as an authority or at least meaningful contributor in some field. If your messages are all over the map, no one will know what list to place you on.
3. At least for now, the number of individuals who can be followed in a list is restricted, so the placement on a good list is a premium event. Thank the list owner, return the favor if possible, and promote the list.
4. Use the reports of the lists that include you to learn how you are categorized. Does that match the brand image you want out there?
5. Create lists that reflect  you brand. Choose wisely, express your thought-leadership. Do not just list the Twitter superstars like Obama and Spears, everybody can do that. Find topical experts who really contribute, and ideally ones who include positive messages about *you* (your organization) in their tweets.
6. Promote your lists on your blogs and other web assets. These will be great content generators, with 24×7 up-to-the-minute information on your topics.

Some of the major Twitter brands like Guy Kawasaki are being added to hundreds of lists per day. Soon there will be a recognized elite in the Twitter List Universe. The expected brands will be there, but many cards will be reshuffled. Make sure that your brand creates a presence there too.

A constant theme running through my marketing work for years (at Borland, Red Hat, Agitar, Zend and Sonatype) was the strategy decision whether we should focus on a Pull or Push strategy. “Pull” was defined roughly as the work done to create streams of inbound leads, brought about by giving away something valuable for free (free software, free content, a community to belong to, etc.). In contrast, “Push” was the more traditional form of marketing, interrupting targets with letters, emails, calls, events, etc

It has always seemed self-evident to me that Pull is the superior approach. It engages with potential customers who are expressing a desire to be communicated with, which clearly results in better relations than with potential customers who receive an unsolicited phone call from your aggressive sales rep. The benefits are both tangible (lower costs) and intangible (improved brand value).

The opportunities for adopting Pull strategies are growing fast with the emergence of search and social media. If someone somewhere is looking for information about a product like yours, you are now able to make sure (s)he will find your information. You can do so at low cost and it makes no sense to omit doing that, while dedicating your scarce resources to outbound campaigns to drum up new interest.

So why is it sometimes hard to get an organization to adopt the Pull strategy? There are a number of reasons: companies have a hard time giving things away for free, it does not come natural to them. Or they adopt the Pull strategy, but when the flow of inbound inquiries falls short of plan, they quickly return to the old ways, where they think they can better predict the outcomes. Or they are not very good at the tools of the trade – the tools are technical and highly analytical and are evolving fast.

I will be consulting for some companies on these topics, and in the process I will research and discuss these items here.

Oh and by the way, the industry has converged on the term “Inbound Marketing” to describe this field of activity. So I will drop the Push/Pull terminology in favor of Outbound/Inbound.