Last Exit Westend

There are plenty of blog posts out there labeled “How to increase your newsletter open rate” – and depending on from the blog post’s date you get advice that “40-50 per cent makes a great open rate” (or: 20. Or 10. It depends on which industry are in, largely. On the maturity of your market. Things like that).

Usually these claims are followed by advice along the line of: 1. Have a great subject line, 2. Make sure you send your newsletter on an appropriate day (usually: end of the week/weekend for private persons, Tuesdays and/or Thursdays for B2B), and 3. always include a personalized greeting: “Hello {First name}”.

Details vary, but most advice centers around the content, and some meta tasks like timing, dynamic field population, and subscriber base tidying.

I am slightly surprised how little effort is put on the improvement of newsletters as-a-service. After all, most newsletter recipients are bound to get largely identical repeating sales messages à la: “Our offer this week: 20% off on all products”, or: “Be the first of your friends to get the new {enter product name here}”.

So: being told over and over that you can now buy for less is a message that tends to wear off even with your most loyal subscribers. “Repetitio non placent”, as they say in Latin.

It is easy (well: is it really?) to imagine that newsletter subscriptions and user activities are following similar principles as any other life cycle model does.

Recap: after a honeymoon period shortly after signup where everything is nice and dandy some fatigue creeps in (be it related to the feeling of “same old, same old” with regard to the messages received, or a general change in user interest), before finally the likelihood to defect is becoming so high that no longer any look is thrown at your newsletters. Such an inactive subscriber is very close to one that never subscribed to your service in the first place.

If we can take this model for granted we can try to come up with particular actions in the newsletter program architecture and timing. The idea with this is to either prolong the honeymoon phase, or to decreasing the marginalization effects that the messages are generating over time.

For doing that reasonably we need to understand what makes the specifics of each phase. Which basic considerations are appropriate and what data is available to support our findings?

Let’s start with some basic considerations about the newsletter tool you’re using.

Assuming that you are using a newsletter tool you access via a browser (as opposed to the email client you have on your own computer) it is very likely that some built-in functionality is at your service, helping you to make the needed distinctions.

You should be able to follow the signup date for any particular user – that is helpful to determine the amount of newsletters a person has already received.

For any newsletter and any subscriber you should be able to see which links were clicked. That helps you to find the most prominent links for each newsletter, and it will help you to determine the activity level and fields of interest per each subscriber.

Finally, you should be able to see if certain email addresses produce bounces – for a set of different reasons the newsletter cannot be delivered to your subscribers’ email addresses – they “bounce”.

Let’s continue with some considerations about the particular hurdles you have to cross along the life cycle. I am modeling the life cycle stages in a generic way for any particular user/service here.

1. Right after signup a certain level of interest from any subscriber can be assumed. This interest can be increased or decreased over time, depending on the subscriber’s perceived value.

2. If all goes well, the subscriber will find your messages relevant and interesting. In other words: the post-signup dissonance is minimal. (I made this term up, deriving it from the term “post-purchase dissonance”, normally used for describing often-occurring mixed feelings about the usefulness of a purchased product in relation to its price).
During that phase you may gain important insights on which items in the newsletter were clicked by a new subscriber. Assuming a user’s explorative mindset after signing up for a newsletter this helps determining which topics are likely to resonate with him/her.

3. Sooner or later the increase in interest will get smaller. Lower click rates and response frequencies are the result. This point marks the beginning of a transition phase where finally…

4. … marginality kicks in.
This marginality can have different reasons (but they are all “situated” in the subscriber’s mind. They cannot be monitored directly): the feeling that the user “has seen it all”, a change in user’s interest (or life situation), a feeling of redundancy in the messages you are sending out, or a feeling of a lack of relevance.

5. As this deterioration in user engagement continues we will see longer and longer periods of inactivity. At this “fatigue” stage we will see newsletters which have no click activity for a given user at all, paired with high click frequencies on other newsletters. We as well can expect a longer latency time (newsletter sendout on Friday at noon, but clicks on it are only made on Sunday evening).

6. This oscillation in user’s response to newsletters may continue for quite a while and may vary according to the moon phase, user’s resistance against bad weather, or to the topics presented in the newsletter. However:

7. Sooner or later subscribers will no longer bother to open or read the newsletters received. At this stage the user has defected from the service.

These seven stages are not marked by clear boundaries. Tendencies over time may show emerging or arbitrary patterns – and in some phases it may only take very little effort to re-activate the subscriber. Other users may rush from phase to phase or show completely erratic behaviour.

The point is: the actions to be taken are very specific in any phase of the life cycle. To tell the phases from each other (they are not identical with the stages) I have included a graph below which models the seven stages of the user life cycle on a so-called “cusp surface”, adapting ideas of the French mathematician René Thom, and two authors named Zeeman and Renfrew (from their book “Transformations. Mathematical approaches to cultural change”).

Both horizontal dimensions of the graph are marking user perceptions of Interest and Marginality (the “parameter space”). The resulting three-dimensional surface depicts the relevance perceived by the subscriber. The path drawn on the twisted surface marks the user life cycle in time. Literally subscribers are “walking the line” Not all on identical paths and with the same timing, but pretty much along similar marks.

The projection of the cusp path to the two-dimensional I/M plain shows two things: (1). a significant rectangular “U-shape” (of the path), and (2) a greyed area which is labeled as “Bifurcation set“.

While the U-shape consists of three clear stages (I will use them later to group the counter measures into them), the “Bifurcation set” is a bit trickier to grasp.
Without even try to be tangent to the mathematical principles behind it, I am sure that a very pragmatic and “graphical” explanation will do for our purposes. In case you are familiar with the concept anyway, just skip the next paragraph. If you want to read more, get Renfrew/Zeeman’s book. It’s really interesting.

Within this bifurcation set, the maximum level of subscriber’s indeterminacy is given. On the three-dimensional surface we see that this area marks a set where for any unique coordinates in the parameter space (I/M) multiple points are given on the cusp surface. The “oscillation” between different relevance levels is to be taken literally, as the “true position” of the subscriber cannot be properly determined within the bifurcation set.
At the same time the subscriber may be considering the newsletters as “highly relevant” and “barely relevant” – it simply marks the assumption of certain indistinguishable criteria and probabilities for clicking or not clicking newsletter links. The randomness of the subscriber action is the point here – and this shall best not be confused with increase or decrease in user interest.

The three distinct stages on the U-shape can be labeled as following:
(A) growing user interest (either purely driven by curiosity and willingness to explore, but most often fueled by the first received newsletters themselves), (B) increased user marginality, and (C) increased likelihood for user defection.

As mentioned before, certain actions from the newsletter publisher can accelerate or slow down the transitions from phase to phase. Picking up the distinctions between strategic, tactical, and operational decision-making along with the stage definitions will give us a valid contrast folio against which we can formulate and valuate our attempts to minimize subscriber defection.

Honeymoon phase
The strategic goal can be defined as “minimizing the post-signup dissonance”. On the tactical level we concede a need for “building interest with regard to message content and message context.”

After all, the subscriber is about to learn about your offering and products (content), as well as about your service, about the terms and conditions of delivery, and about the sidekick offerings your company has in stock, i.e.: What other subscriptions are there? Why should users create and maintain their user profile? (this would be the context).

What would be operationally the worst thing to do was the repetition of identical messages and similar offers.

After all, a novelty factor deteriorates pretty quickly, if the only changing variant in your communication would be whether you have “Sunglasses for sale with a 20% discount” this week, and “Winter boots for 20 % less!” next week.

If you, in other words, are repeating both the communication scheme and the benefit week over week people will learn that pattern quickly. Instead of ordering at your shop after having seen the repeating message for a couple of weeks, they may rather go and check whether your discounted prices are any lower than the prices of a well-established competitor.

Increased user marginality phase
Strategically, the goal for this phase would be “to minimize the value perception decrease”.
Tactically, the appropriate question is: “How can we re-focus the subscriber?”, and the corresponding answer would be: “By offering strong selection criteria to re-gain relevance”. These selection criteria are to be strongly centered around content and categories.

Operationally, the worst thing you could do is to do random line-extension offers.

Imagine: you have tried to build a connection with your subscriber by focusing on your core competence of selling fashion. A lot of related featured fashion products can be thought of, but with a line extension offer for “winter tyres” you surely wouldn’t support the recipient’s focus.

In other words: if you are adding arbitrarity to your communication through contextless contents in this phase the odds your message will be well received are strongly against you.

Defection phase
Undoubtedly the trickiest phase, the strategic goal would be formulated as “regaining significance through a strongly personalized core offering”. Tactically, it would make perfect sense to now grant significant personalized incentives with clear benefits.

Operationally, the worst you could do in this phase is to grant generic “lawn mower” discounts to all of your subscribers of a certain age – and even worse it would be if you got caught with a discounted offer which isn’t better than the street prices paid for the goods you are discounting on.

Well – all of that is pretty obvious, once you think about it. Specific situations requiring specific actions, not a general de-valuation of your offers, products, or communication.

One question remains, however: how the hell could we tell one phase from another?
Simple answer: by utilizing usage data per newsletter and historic data per user.

Your newsletter tool should give you plenty of information about any of your users’ actions (if not: consider a different tool!) – and if you are doing ROI tracking on your newsletters (both in the newsletter tool and in your favourite web analytics system), you should have a valid purchase history on newsletter level and on user level.

So: start from what you know. You should know from your users at least: 1. at which point in time they signed up for your service, 2. how many emails they have received/opened (although the latter is a somewhat obscure metric!), 3. which links they have clicked in any of the newsletters, 4. which of the clicks led to a purchase.

As you are having that data both on an individual subscriber level as well as on the level of any particular newsletter issue you have two data sources to compare. Cutting through these four metrics groups, and knowing that all reasonable newsletter systems offer tools for maintaining your subscriber base already gives you a pretty decent toolkit for segmenting and filtering your newsletter recipient group and for targeting your messages related to the phases outlined here.
Start using that. Tailor your contents to your insights and filters. Experiment and play with it. Make your subscribers matter.

Finally: some examples on what to look at. Consider a user receiving a monthly newsletter from a travel agency. Having special offers for flights and hotel packages makes a central pillar for the commercial success for such a newsletter program.
Imagine one of your subscribers is always looking at the flight offers you have in for London (or for Luton. For Lisbon. Whatever!). My advice is: use that historic data set for future offerings – but not right away. Do it in the “marginality” phase.

Group your offerings by world region then. By destination country, if you want. By city, if you have to. Give those who have booked a trip to London through one of your newsletters more than once a related discount offer in a later stage of their life cycle (Defection).
Look at the most common city destination you have in your click/revenue statistics. Does this make a sufficient user base for a “special” newsletter sent to those a bit further down the road of their life cycle?

What’s the point constantly offering flights from Europe to Asia to those who never ever have clicked on any destination link which lies outside of Europe? Well – leave it up to them whether it matters or not. Let them refine what offers they are interested in. Help them select based on their previous choices. Provide relevant content with respect to their preferences.

Although you may argue about missing a sales opportunity with the “flights to Asia” thing – keeping a user retained through the relevance of your offering matters a lot more these days, I believe.

Can you give a presentation on a Web Analytics Association event without mentioning the term “Web Analytics” more than once?
Well – see for yourself in this Youtube video (3 parts, about 30 minutes total).
First part here:

This presentation mentions pygmies, dragons, and in general revolves around some core ideas in medieval and renaissance map-making.

In retrospect, I have to apologize for the rather frequent usage of the word “indeed”. I couldn’t read my speaker notes due to a display fuckup and had to make up for it somehow.

Thanks to Pete for uploading the video and to Anders for smiling at me for the entire course of the presentation.

“Hi. We are planning to open a bunch of hairdresser salons in various locations in Germany. We’ve dug out some data and are about to evaluate some good locations at the moment. We were wondering whether you would want to help us.”

- “Sure. What’ve you got?”

“Well – here’s some data for you:

<<Amount of hairdresser salons in Germany: 73 500

Amount of hairdresser visits per head [sic!] and year, 2008: 5,75

Amount of money spent by males on average per hairdresser visit, in Euro: 15,32
Amount of money spent by females on average per hairdresser visit, in Euro: 42,75

Agreed hourly wage for a hairdresser in the federal state of Saxony, in Euro: 3,06
Agreed hourly wage for a hairdresser in the federal state of Hesse, in Euro: 7,99 >>”

- “Errm, OK. What next?”

“That’s how far we’ve gotten, really. Are you able to help us?”

- “I don’t know yet. Let me evaluate. I’ll get back to you tomorrow.”

“Appreciated. Tomorrow then. Bye.”

What I have found after one hour of research – finally from the same data source:

<<Amount of hairdresser visits per head [sic!] and year, 1998: 15 >>

Data. I mean: fuck it, y’know?

[*All data in <</>> from: brandeins 12/2009, "Die Welt in Zahlen", Hamburg. That magazine is much appreciated and shall be considered wholeheartedly recommended for well over a decade by now (NB. Its's published in German, entirely). For that particular reason: this is an entirely unauthorized translation by the author of this blog post.
**All annotations in square brackets ([...]) are added by the author of this blog post.
***This post was assembled from my unused materials for the presentation about “Data and Context” at the “Web Analytics Wednesday” held at the “World Analytics Association” event in Sanomatalo, Helsinki, March 30, 2010]

And, yes: If we would be truly serious about adding all needed disclaimers to the data we deal with on a daily basis we would need to have a completely new nomenclature which can distinguish between at least three levels of depiction.
****All asterisks (*) in this post are just randomly added and don’t necessarily refer to any particular semantic or syntactic entities of this post in part or in total.

7. Greed (Avaritia)

[editorial note: since I began to write this article in early December, quite many announcements have been made which seem to fit this topic perfectly. As a side effect, the article's original intent (ranting/bullying about certain common misconceptions and vanity power play which exploit structures in Web Analytics) has vanished more and more into the background, giving way to looks at more concerning emerging trends within the field of Social Media, its measurement, and the consequences. For that particular reason this post turned (a) bit epic, (b) less amusing, and (c) is somewhat lacking the structural coherency from the previous posts.
As we know from Ted Nelson: "Everything is deeply intertwingled.". I hope you will find it worth the time nevertheless.]

Depending on your career ambitions, you may desire getting the title of a “Social Media Guru” (for job descriptions and inspiration, try this search) – or you may insist on having a sign on the door of your office which says “Head of Consumer Insights”. Combined with a yearly salary of £150.000. While you still keep your lackeys, the Analyst, and the hairy IT guy on your roster.

No matter what you want to be called – it is important to understand that yet the Analytics industry is not mature enough that anybody really could derive a 360 degree view on what consumers and customers do by just looking at clickstream data from one data source.

Nevertheless all the commercial toolmakers are claiming that their products are the best – and if you ask for quotes from toolmakers or external consultants you for sure can expect to be charged premium rates.

Your boss may faint if you’d ask him for a salary of £150.000 p.a. straight, but if there’s a good probability you could bring in an additional revenue of £2M a year, he may consider this a fair deal. This simple ROI logic is as well applied by tool providers and service providers.

“ROI” has become a mantra in ad spending over the years. Meanwhile, ROIs are calculated for design makeovers, for usability assessments, and technical platform renewals. The logical next step is to calculate ROIs for employees, and for bought-in third party services. And to pay everybody according to the revenue they generate.

For Web Analytics vendors and other insight providers the famous “Million Dollar hole” that can be spotted from conversion data (cf. for an example the introduction in Steve Jackson’s excellent “Cult of Analytics” book) can be a true door opener, if – and only if – it’s in the client’s central interest to (a) save money or (b) generate more revenue.

Labeling this attempt to find money as “greed” is slightly over the top in most cases: web site budgets have been shrinking for a decade now (unless you are a well-funded startup), and the only way to revert that trend is to tie web site, agency and consultant spending to the revenue generation process.

Agencies and consultants are of course trying to get their share by aiming high: “If the conversion rate (CR) goes up 5 per cent the additional revenue will be eight million Dollar a year”. Although that is undoubtedly true, the big IF in the beginning of the sentence is often overlooked – particularly if your shop has a CR of 12,24% already at the moment. It will get quite expensive to gain enough quality traffic or to further improve the conversion funnel.

Putting up a brand new web shop on your high-volume consumer site with several millions of visits every month requires Web Analytics tracking – that much is for sure. But if you have to pay your Web Analytics provider by visit or page view volume you occasionally can’t help but calling this pricing model “daylight robbery” in many cases.

While the daylight robbery is happening particularly for customers with sites in Europe at the moment, yet another trend comes over from the US.
Particularly with Social Media we recently notice a repetition pattern from the late 90s: Volume metrics like “Impressions”, and non-monetary micro-conversions (“Likes”, “Posts”, “Comments”) are all over sudden becoming the new cool metrics.

We see a shift from the paradigm that web sites have to acquire their traffic to the paradigm that brands are invading the social networks literally for free (as traffic is already there) and are measuring their success in doing so with content consumption metrics. These metrics are clearly focusing on brand building, but the channels are true exchange and communication channels. This implies the need for constant monitoring. To be precise: Social Media traffic costs money, and the nature of the spending is totally different than with Search Engine Marketing. At least if you don’t want to ruin your reputation as a brand.

The short-term reward for companies actively maintaining their presences in social media networks is an increase in “followers” / “Likes” / or whatever other micro-conversion events. The amazing increase in people being present in social media networks generates similar dream increase rates as in the early days of the web (“Twitter followers up this month: 250%, “Likes” on our FB page: +82% in the last 10 days. Wow!”).
We’re, in a sense, going back to the good old days of feelgood metrics at the moment (cf. part 2 of the series) – but in early January 2011 Goldman Sachs has reportedly invested 500 million Dollars into Facebook, seeing a valuation of 50 billion Dollars for Facebook altogether. This coup may change the Social Media game entirely.

As is indicated in various places on this blog monetization is a method for making potential revenue real. Still: the reported 50 billion Dollars may or may not be a hypothetical value – but it creates a particular dynamics and focuses attention towards Facebook. The same is true for many other companies.
If, let’s say, an Analyst from an Investment Bank checks the market capitalization for a particular company and finds out it is underrated, according press releases and investor advice is sent out. Investors start investing in such a company, stock prices rise, and those who have bought in early enough are able to cash in while price and the demand for stocks is still increasing. At least for a while.
If you have followed the economic news in your particular country you may have a precise idea about the mechanism. If you haven’t, rest assured that this happens without any consideration of the “real” value of the company.

The German sociologist Niklas Luhmann described the structure of this economic shift already back in the Eighties: the mode of observation is switched from the first order (“observing companies”) to the second order (“observing observers of companies”). If Goldman Sachs considers Facebook valid to invest 500 million Dollars – and these guys surely aren’t mad – wouldn’t you be wise to invest in Facebook, too?

Monetizing each Facebook user with 100 Dollars may be a bit too optimistic. Or then: it may yield some unpredicted side effects in turning up some more funny money for Social Media gurus: if the average Facebook user has a value of 100 Dollars across his/her entire lifecycle, and a Facebook Fan can be purchased at around 14 to 18 cents – wouldn’t it be wise to recruit more Facebook users through your company’s web site? And couldn’t you easily monetize on getting Facebook fans recruited through your corporate web site and any occasional campaign as well (if you’re B2B)? If it turns out that Facebook will gain another 10 million visitors within the next three months – wouldn’t that increase the value of your investment, too?
And: having more fans may enable your company to get a bigger share of the potential value of existing and future Facebook users … well – you got the idea by now.

The more people join Facebook, the more attractive it gets to have a Facebook presence. Every Facebook user has on average 130 friends, and if one likes your Facebook page, you reach all of his/her friends as well (as the notification about that “Like” is appearing as a wall post). Calculating the true Reach for Facebook is a bit tricky, but still pretty much resembles a Ponzi scheme (for an actual blog post during the writing of this article, check out this article from Joseph Perla. I haven’t read it entirely myself but will do so after this article is done. I hope it’s as good as the first couple of lines indicate).

There’s one apparent downside to the whole easy idea of monetization and generating fans for your company’s presence on Facebook, though. It’s an unnoticed change in the look and feel of the Facebook presence itself, caused by platform updates, as it happened for some of the sloppily maintained presences on August 23, 2010.

A recent accidental update of Facebook (December 16, 2010) was revealing a tab-free design for FB pages, removing all custom-made landing pages for Facebook presences. According to Mashable, Facebook claims not to change the pages to a tab-free design any time soon. In August 2010 the development of FBML applications was declared to be phased out. Instead the return of the deprecated <iframe> tag was promoted. Newer developments like OAuth 2.0, Graph API, and the JavaScript SDK) have yet caused some uncertainties about if and how the support for existing fan pages will be continued in the future (for details on this topic, check FB’s Developer roadmap, here).

As we’ve come to learn Facebook is trying to monopolize access to user’s data on the one hand (that’s yet another flavour of “greed”, no?), on the other hand there are plans to limit what companies can do within their presence on the Facebook platform.
At the same time as an investor appeal is generated, Facebook itself is setting up side bar elements like “Find More Friends”, where users can plough through their email contacts across various services, or find other acquaintances based on their hometowns, current cities, mutual friends, workplaces, colleges and universities.

That may be tempting for some. Others (particularly the German privacy commissioner Johannes Caspar) are generally suspicious about data collection and exploitation of data not owned by the respective companies, and yet some others are seeing FB’s “Open Graph” as a chance to make “FB traffic convert similar to the patterns that we see for Google traffic.” (Jesper Åström, in “The Facebook Marketing Guide 2010″).

Holy mackerel! It seems as if the easy, occasionally boring, and somewhat geeky world of equipping your own tiny little web site with Web Analytics tracking code and making a career from that has turned into a multi-million (or billion) Dollar game by now. Legacy tracking and IT systems, web site contents from the late 1990′s being run on outdated platforms, business people not talking to IT departments, Analysts not properly marketing their capabilities to the business, the rise and fall of compound metrics for Web Analytics and Social Media measurement – this all has been and still is a commodity.
The described ROI angle is still valid and will stay relevant, but the money won’t return that easy as online behaviour is changing, and as a whole new range of devices for browsing and curated channels are emerging as you read these lines.

You may still get away with a corner office, a fancy title, your own staff, and a fixed salary of £75.000 – for now. Chances are getting slimmer that you will be able to keep more for long. And if you’re lucky enough: Enjoy it while it lasts.

And still nobody will be clicking on Facebook ads. Sorry, folks.

Post Scriptum 1: The mentioned Johannes Caspar considers the usage of Google Analytics in its current form illegal in Germany (cf. an according article here, and states that a test legal case against a large company using the service is in the works. Germany is not America, so the “steep fine” will not be as outrageous as it would be in front of a US court, but it could be enough to grind down the whole Analytics industry in Germany in 2011. At least it should be enough to create a massive uncertainty in the market.
Post Scriptum 2: Goldman Sachs has decided to not offer Facebook shares to investors from the US. More details from a search engine near you.
Post Scriptum 3: Facebook has decided not to share user addresses and phone numbers with Facebook apps. More details on your favourite news site or blog.

… and Five Things to Learn” is a brilliant roundup and an excellent outlook for the current and emerging state of Social Media Measurement.