It took me a long time before I got back to completing this series. I joined Akiban Technologies and that has gotten me very busy in the recent months.

In the previous post I presented the model for viral distribution of a free product.

Now I'll add the premium product and the variables that model the conversion from free to premium:

Conversion Rate

This is expressed as a percentage of the number of users of the free product in a particular time period. There is an important assumption in this – conversion numbers are driven by the inventory of free users, not by how long the users have been using the free product. It is a great simplification of the model, and in fact most often more accurate than models that try to predict the time involved to migrate from free to premium.

Customer Life Time Value

How much can you invest in programs to land a new paying customer? Surely not more than the total revenue that the new customer will bring over the time that he will be a customer. That amount is referred to as the Customer Life Time Value (Customer LTV). (In fact the calculation needs to be adjusted for timing issues: the cost generally precedes the revenue, so a company has to finance these upfront costs.)

Assume a scenario in which 1,000 customers are acquired who deliver $5,000 Average Revenue Per Year (ARPY). If the Annual Retention Rate (ARR) is 90%, the number of customers in year 2 will be 900. As you can see from the following spreadsheet this original group of 1,000 customers will dwindle to 6.36 after 50 years (delivering $31.813 in year 50). Adding the annual revenue from this dwindling group over 50 years and dividing it by the number of original customers gives you the 50 year value per customer at $49,714.

A workable formula to calculate the Customer LTV (without the 50-year constraint) is the following:

(Customer LTV) = (Customer ARPY) / (1 – (Customer ARR))

In my example this calculates to 5,000 / (1 – .9) = $50,000

Customer Life Time Value

Customer LTV Excel Model (Click to download)

Marginal Costs

If an airplane is about to take off and some chairs are empty, how much would it cost them extra to deliver the flight if they would allow an extra passenger to come on board? That extra coast will not include fuel and salaries, these costs would be made with or without the additional passenger. The extra costs include the food (s)he eats and the administration costs involved in processing the ticket. In economics terms, marginal cost is defined as the change in total cost that arises when the quantity produced changes by one unit. In the software industry, where distribution of product is generally done by sending bits across the internet, the marginal cost is mostly the cost of support, if the extra customer experiences problems.

Customer Acquisition Costs

The costs of acquiring a customer vary enormously from business to business. If you have to call your potential new customer, take him to lunch, meet his technical and finance teams, train his staff and install his acquired products, then obviously the cost of that acquired customer is higher that for the ecommerce commerce that relies exclusively on Google to be found and has no sales staff.

The easiest way to calculate this amount is by adding all sales and marketing costs, and dividing it by the number of new customers. More detailed models separate out the general branding programs and public relations, and attribute only the direct marketing and sales costs to this category.

The comprehensive model

Now that I have reviewed all the variables to are required to describe the revenue and profit modeling for a Freemium business, here is the spreadsheet model that brings it all together. It adds detail to the marketing costs and engineering costs, but that is all fairly self-evident. Take a look, and let me know if this is useful to your business. I will love to hear your comments and would be happy to answer any questions.

The Spreadsheet model for a Freemium Business (Click to download)

In the previous post I introduced a spreadsheet that can model viral growth. In this post I want to introduce two additional concepts into the model, to make it conform better to what we see in the real world. These concepts are Total Available Market and Retention Rate:

Total Available Market (Carrying Capacity)

Funny things happen with viral models when the Viral Coefficient is greater than 1 and the Viral Cycle Time is short enough. The model will 'hockey stick' to very high numbers because of the exponential growth that is embedded in the model.

Take a look what happens with the model from the previous post if the Coefficient is set to “2″ and the Cycle Time to “1″:

According to the model you will have 2 million users by the end of week 10 and pass 1 billion in week 19. Before you panic and start recruiting additional support staff, let us introduce the concept of Total Available Market (TAM) into the model.

Assume our product is a tool that will be used by Java Developers. Our estimate for the total worldwide number of such developers is about 10 million. It is clear that the total number of users of our tool will never exceed 10 million, and the viral effects will slow down as the number of users grows to a meaningful percentage of that 10 million.

The way to include this in the model is to multiply the number of acquired new users with the TAMFactor, which is defined as:

TAMFactor = (1 – TCU/TAM)
(TCU means “Total Current Users” and
TAM means “Total Available Market”)

TCU/TAM will be close to zero in the early phase of the viral cycle, and therefore the TAMFactor will not slow down things too much. But by the time TCU is 25% of TAM, the factor slows the viral spreading down to 75% of the normal growth.

Here is the updated model:

You can see the formula for cell H26 (=H$25*(1-G$27/$B$22)) and its impact: it slows the number of virally acquired users from 63,636 to 63,237, a modest slowing. But as you can see on the chart below, that changes by week 13: in that week the unadjusted number shoots off the chart, and in the adjusted chart the trend flattens and never gets above 9 million.

Retention Rate

In any community of users, some percentage will drop out. Users may change jobs, change technology, loose interest, become dissatisfied, merge with other users, be promoted, demoted, whatever the case may be. In this model we will define the Retention Rate as the percentage of users that continues to be a user a year later. In the software industry 80% is generally seen as a good number.

In the spreadsheet model we convert the annual Retention Rate into a weekly Attrition Rate with the formula:

Weekly Attrition Rate = (1 -(Annual Retention Rate^(1/52))

Here is the updated model with the attrition calculated for cell H28:
The updated chart shows that later in the cycle, growth declines and the ongoing attrition starts driving the overall number of users down:

Here is the Excel spreadsheet, so you can review the model and experiment with the values of the key variables to learn to understand their impact.

TAM and Retention Excel Model (Click to download)

In my next post about this topic, I will expand the spreadsheet model to incorporate a free product that depends on viral distribution to succeed and a commercial product that will be purchased by a subset of the users of the free product. This is referred to as the Freemium model. It will model the conversion from free product to evaluation of the premium product, conversion of evaluation to purchase and the impact of evaluation time.

Stay tuned.

The discussion started with a question by Patrick Murphy of SiliconCloud:

Meta Keywords – Yes or No?
What are your views on meta keywords? I know Google have come and said that they do not use them but from my experience and clients experience they still have an influence. Can I ask what your experience or view is?

So far, there are 113 answers. The consensus seems to be that it does not help much for Search Engine Optimization (SEO), but it does not hurt much either. Here are some of the contributions that I found very helpful:

Kevin Micalizzi, Community Manager at Dimdim, points out what the policies of Google, Yahoo and Bing are with regards to Meta Keywords:

Google doesn””t use them:
http://googlewebmastercentral.blogspot.com/2009/09/google-does-not-use-keywords-meta-tag.html

Yahoo still recommends them:
http://help.yahoo.com/l/us/yahoo/search/indexing/ranking-02.html

Bing says you still get some credit for them:
http://www.bing.com/community/blogs/webmaster/archive/2009/07/18/head-s-up-on-lt-head-gt-tag-optimization-sem-101.aspx

Matt Sullivan, from HubSpot gives some perspective on the relative (lack of) importance of meta keywords:

Ultimately there are 5 main areas for On-Page SEO: 1. Title Tag, 2. H1 Tags, 3. Content, 4. URL, and 5. Image information (file name, alt-text, and surrounding text).

Keep in mind, On-Page Optimization only accounts for roughly 25% of your total optimization. The other 75% comes from the number and quality of your inbound links.

Fadi Semaan adds to this:

I need to add that content is a king so do not forget to use your keywords in the H2 Tags (Google gives it better rating than H1) and the text itself (content of the page)

Some posters expressed the concern that creating Meta Keywords makes it easy for your competitors to find out what your best keywords are. Christopher Regan of HyperDisk tells you not to worry about that anymore by pointing out that your SEO strategy and budget is out in the open anyway:

As to your competitors knowing what you are targeting they can use SpyFu.com and easily check on your PPC buys, then also use other tools for keyword/phrase density scans.

(If you have not checked out Spyfu yet, do it now: it is a real gem!)

Tom Lynch from Astek Consulting also points out good resources and best practices (lightly edited):

If you have limited budget and time, focus on unique HTML titles for every page that are descriptive of the content on the page. Lead with the targeted keywords and tag with the corporate ID (only if needed.)

Keep it under 65 characters including space and punctuation because Google clips it after that in the SERPs. Use tools like wordtracker.com and keyworddiscovery.com to determine the “right” keywords. Words people actually search for. Shoot for the highest search traffic counts with the least amount of competition pages. Wordtracker has added some nice features like all in url and all in anchor making it easier to hone which one to focus on.

Take a look what you have indexed on Google first by doing a search “site: http://www.put-your-url-here.com” and see what is indexed on Google. If you have thousands of pages on your site and only a few show up you have problem. It may be your CMS is not search engine friendly – got to fix that first…

But meta keywords are moot, moot moot! Meta description on the other hand – important. Not for scoring the page but often used as the snippet text in SERP””s. Must be unique for each page, keep it under 150 characters including spaces and punctuation and in addition to the page synopsis try to include a call to action.

David Hallmark at CrystalVision points to need to pick keywords that are (near the) top in their niche:

This whole conversation began as a simple question on Meta Keyword tag and has evolved into more. Trying to optimize for a generic term is well over so mostly we are left with the low hanging fruit or niche terms. Gather enough niche terms and you can begin to rank well for the generic terms as they relate. In the meantime you will have cornered your market.

On-page techniques won’t matter if you are going after terms that are too competitive. In our view anyway generic terms are tire kicker terms and actually don’t do any better response wise as a targeted term. “Real Estate” vs “Hawaiian Real Estate” makes a world of difference. (2.1 billion vs 146k results) Then again, “Hawaiian beach real estate” has ZERO competition!!!!!

There are 10’s of millions of wide open keyword terms to rank high for. Your keyword list is where your energy should be concentrated on. The problem with most lists is what’s good and do-able and which are tire kicker terms? A problem I’ve fought for years with internally and with clients. Without sounding “salesy” we developed a tool to do the really heavy lifting of your keyword research: http://www.keywordnichetools.com/

Simply put, knowing where the openings are you can toe wedge into even the tightest of markets. In the example you can rank high very quickly for 43 terms or spend months or years trying to rank well for 4. Your call but you knock those 43 out of the park. The other 4 will fall into place on their own! Or by that time you won’t care!!!

I am sure that Patrick Murphy feels he got his answers and then some!

I have created a Diigo list with all links (click here), including the cached pages of the LinkedIn forum for those of you who are not members yet (does anybody like that still exist?). Happy reading!

Just the day after I posted the article about Google Reader an article was published on ReadWriteWeb (RWW) with the title ”RSS Reader Market in Disarray, Continues to Decline“.  The story states that the number of unique visitors to RWW who use Feed Readers has been declining and that the Feed Readers have made little progress recently in terms of features and functionality.

No doubt, the tools interacting with social media feeds have been in the center of attention in the past year. But their hype does not yet make them as useful as the best feed readers. And the data to support the article comes from just one source and does not yet represent a trend.

The comments that the article receives, demonstrate that for most people the news readers are truly useful and are the primary tools used to follow what is happening. This is very true for me as well. The report about the death of RSS Readers seems to be greatly exaggerated.

The writer of the story, Richard MacManus, states that he follows Twitter and Twitter Lists for news multiple times per days and only uses Google Reader for specific searches. Implied is that the Twitter model is somehow superior.

RSS News Feeds and Social Media Streams are two distinct content streams that are kept separate by their technology, not by substantial semantic differences:

• Twitter allows you to subscribe to an RSS feed of the results of any search at search.twitter.com. Why no RSS feeds for individuals and/or lists?
• Google Reader could support the Twitter API and present Tweets to the user like all other XML content.

Either of these two approaches would allow us to mash the two content streams and use one reader for both. It would greatly simplify my life, and I guess that of many others as well.

Zack Urlocker posted an article on “The Open Source” about one of the scams that the internet marketing technologies have spawned.  He writes about the practice of setting up large numbers of review sites that publish product reviews, invariably very positive, full of links to the reviewed product.

The reviewers make money because they receive affiliate revenues from the vendors of the products they promote.  In some cases they make enough to be able to pay for Google Adwords to promote their reviews and so increase traffic and thus grow revenues further.  On a site like http://payperpost.com/ this concept of sponsored conversations is openly advertised.

So what is wrong with this practice?

It is that many of these reviews never disclose that are paid by the vendor for their reviews.  The reader is duped into thinking that an objective third party had a good experience and therefore wrote the review.  In some cases the reviewers post their reviews on many sites, each with slightly different names, words and graphs, so they can create the impression that the product is widely used and liked.

Perhaps even more wrong and more important to the vendor is the fact that the links from the reviews to his product pages grow the page rank of these pages, so he will get a better position on search results and thus more search traffic and increased sales.  An honest competitor may get very frustrated seeing his valuable content get crowded out by these bogus reviews.

So should this be stopped and if so by who?

The FTC has announced its intention to play a role when it introduced “Guides Concerning the Use of Endorsements and Testimonials in Advertising”, which went into effect December 1^st (you’ll find the full text here if you have the stomach to read its 84 pages).  Unfortunately, these guidelines are deeply flawed, as Jeff Jarvis describes better than I could.  It is as hard to regulate what people say on the Internet as it is what they say in face-to-face speech, the Internet is not the Medium it is the Place where they communicate. The owner of a coffee shop cannot control what two people say to each other over a cup of joe.  The response to the FTC on this front has been very unanimous: stay out!

Google cannot control the content of web sites either, but it does have the ability to score pages lower if they detect fake review sites.  This is not easy, as there are many uses of affiliate linking that are ethical and legitimate.  But if Google gets it right most of the time, then the value of these links goes down, the page ranks are normalized and the financial incentive for the scammers goes away.  It would work like a spam filter, but for web pages.

I also think that the people who search for product information will grow wiser over time.  If one reads a glowing review about something, it makes sense to check the rest of the site on which it is published to see if the authoritative.  Same for the reviewer, what other reviews did he write?  When the public gets wise, the scam dies (it has been a long time since I received email from my cousins in Nigeria that want to transfer millions to my bank account).