Comparison of churn rates

One of the most closely guarded secrets in business seems to be the churn rate. This is the rate of attrition, over a period of time, that subscriber-based customers “churn out” (unsubscribe) of the customer base. Churn rates are often measured in monthly terms, especially in the cable and satellite television and the wireless telephone industries. Business analysts will often look at the churn rate on a quarterly basis. Likewise, an annual rate of churn is a commonly used measure.

Recently, I got to searching the Internet for a comprehensive list of churn rates, if such a thing existed. A churn rate table, across industries. But alas, nobody seems to have compiled a database measuring the percent of customers who churn out of various companies, across sectors. So, I had no choice but to create such a table.

If you try to discover reported churn rates, the first thing you’ll notice is that companies will use various time periods as the basis. Monthly, quarterly, annually. I even found one study of lawyer retention rates that spanned a four-year period. So, to try to make this an apples to apples comparison, I converted all of the churn rates I found to “per year” terms. Note, however, if a company reported a 2% monthly churn rate, I did not say that equates to a 24% annual rate (2% times twelve months). Rather, I calculated a sliding count, such that in the first month, 100 customers would dwindle to 98. But in the second month, the 98 customers would decrease to only 96.04 customers (98 minus 2% of 98). Continue reading →

Measuring performance from each website section

When measuring a section performance from a website (it doesn’t matter its profile), we could commit the mistake of thinking that the one with more total pageviews is the most successful one. There are two things to take into consideration to understand which section or sections are the most successful ones and when we could say that the performance from a particular section is not acceptable.

visitors

1- When a visitor arrives to your “home page” has exactly the same chances of visiting any of the available sections, so measuring just each section landing page (or section home) allow us comparing which one is more attractive for our visitors. Measuring all the pageviews form each section drives us to measure things with different dynamic as equal, making incorrect decisions. For instance, if we have the section “chat” it could have the same Unique visitors than other parts of the sites (remember that we normally don’t measure internal UV’s because it could make us duplicate information) but the ratio pageview per visit will be close to 1 (1 pageview for each visit). On the other hand the same ratio from the section “our products” will never tends to 1, actually will have a bunch of pageviews for each visits. So saying that the section chat is less attractive just because of the low quantity of pageviews is completely incorrect because its own dynamic doesn’t allows it, so what we should expect from this section (mostly in a community website) is that the people spend lot of time there. The idea is comprehend which sections are more attractive to ours visitors.

2- On the other hand we could compare internal content (if its dynamic allows it) from different sections. I mean compare different content, from different section taking into consideration how comparable they are. So, we could compare the home of each section with the home from the other sections, and compare pageviews from one internal content of each section with one internal content form another section.

This technique help me a lot times of making very hazardous decisions, like recommending a client to cancel a section that in total pageviews seems to be unproductive.

Successful Online Business – Sixth and Seventh Part

Decision making: When you have enough information you should analyze the possibilities and scenarios and take the decisions that minimize the possibilities of fails assuring your success.

Control: Controlling the flow of your plan is vital in order to finds out your position in the project and to anticipate deviations and, if necessary, runs your contingency plans (Never forget to have contingency plans; you could analyze your historic information to reduce your uncertainty). “All project managers face problems on Monday mornings – good project managers are working on next Monday’ problems.” You cannot work on next Monday problems if you do not control how are your project going.

Successful Online Business – Fifht part

Measurement plan: In this stage we should define the KPI’s for your Scorecard and Dashboard. It should include the company strategy, objectives, tactics, and KPI’s. You can’t manage what you can’t measure repeated Robert Kaplan over and over again.

I would add to this quote, you shouldn’t measure what is not relevant to your strategy. Probably for offline projects is different because measuring is synonymous of outrageous costs, but in online you could almost measure all the people online experience without so much effort. So one of the most valuable advices I could give you is to not measure everything. The Analytics Professionals heard daily our clients ask to measure “all the information we could get from their sites or campaigns” but we should remember that, not all the data is information (but produces cost anyways), not all the information is useful for every single project, and finally as much data we have as much noise we get.

a. Tracking architecture: Your defined the KPI’s going from the strategy level to the data level. Now you should define the tracking architecture from the data level to achieve the required information to the strategic level. It’s vital a complete understanding of each tool’s backend (ad server, analytics, mailing servers, SEM servers, etc.) otherwise you wont be able to know what does the tracked information mean. Remember that for example Unique Visitor is a “virtual” metric that will be defined by the timeout definition (daily, weekly, monthly) and the time base (monthly, rolling) and by the IP tracking (is different for a Logfile analyzer than for a Page Tagging system). Please don’t forget the Tracking architecture document.

b. Constant measuring: Measuring (and reporting) in every case has two objectives, decision making and controlling. In order to define the period time for reporting, I always recommend companies to define the required period they need for decision making and controlling, and that’s the reporting period. Having daily reports, for example, just because is possible is not only ineffective, it’s going to drives you to incorrect analysis and finally, to misinterpret the information which wont help you at all to take the right decisions.

c. “Findings” and “Key Learnings”, and recommendations: After you have the report with the analysis you should get the most important findings (understanding the present), define the key learning (generating historic for future decision making) and recommendations (next steps). The idea understands what is happening, how we could use in the future our learning, and finally, what actions we are going to take in the future to be more efficient (I suggest to learn Peter Senge’s Book “The fifth Discipline”. The problem is not that you have a tough problem to deal with, but whether it’s the same problem you had before).

d. Adjustments: While you measure your success you have a better understanding of your business so you could adjust your measurement plan, or even goals to your business reality. I recommend to not taking your measurement plan and goals as untouchable things, otherwise you will be at the end prisoner of it.

Successful Online Business – Third and Fourth Part

Brainstorming process (identify tactics): Once you understand what’s your “weakest link”, you should launch a brainstorming session (or several) in order to find the project or projects (tactic) that will best help you fix your “weakest link”. If you have a lack of information about your client behavior and attitudes then a great idea is creating a social network where they could relate each other and as a result you will get a bunch of information that will help you understand what are they doing and why.

4- Project design (tactics): Now you have the idea (The difference between idea and project is that IDEA is the “what to do” and the project is the “what to do, how, when, etc”), so you need to design your project starting by defining your project scope as much detailed as possible, otherwise you will face lot of problems in the future. I recommend 70% planning 30% execution, and if you think I’m exaggerating just remember how much time have you spent in last projects just fixing problems because of a weak planning. Remember that a “millimeter of deviation in the basement of a building could make your building falls down when finishing the construction”. “If you are failing to plan is because you are planning to fail”.

Successful Online Business – Second part

Identify RCR (restricted capacity resources): Now you should find the Restricted Capacity Resources (Based on “General System Theory”). All the projects are systems, and within those systems flows productivity (not only talking about production but also about communication, sales, quality, etc). In this stage you need to finds what is restricting your system most (project) and attack it with all your efforts, remember that if you pull a chain you will break it in just one link, it means that one link is always weaker than the rest, and when you fix it, then another one is going to be the weaker one. So you need to find the weakest link from your chain (your system, your project). It means that if you are stocking bunch of products, you should focus on your marketing mix, that’s your weakest link.

Juan Damia (analytics)

Successful Online Business – First part

Very frequent people ask me about how can they make their online projects work, and more frequent they don’t understand when I ask them about how are they aligning their online project with their Corporative Strategy. An ecommerce website, an online campaign, a community website, a SEM campaign, or whatever else are tactics that should be aligned to a Corporative Strategy. If your online tactic is not aligned with your Corporative Strategy, probably the most intelligent decision could be start again from a scratch.

1- Project understanding: Understanding the strategic plan and objectives from the company is the basement of any project. In this stage is deeply important that you have all the answers you need. I suggest writing in a paper one by one all the questions you think all the people that will be working in the project would need and then find the right people to answer those questions. Remember that you will only find the right answers if you ask it to the right person.

Unique visitor – Unique Meaning?

When people talk about unique visitors, are they all talking about the same thing? When people compare two different website by their Unique Visitors are they comparing equivalent things? Well I guess both answers are “probably not”. There are some important things to take into account to really understand what “Unique Visitor” means in each case.

analytics Continue reading →

Packet Sniffing for Web Analytics – Amazing Interview to Ivo Rehberger from Nextwell

In order to understand the application of Packet Sniffer for Web Analytics proposes I interviewed Ivo Rehberger, founder, Development Manager and the main software architect of Nextwell. He has a doctoral degree in technical cybernetics, and has published many articles in technical journals and e-zines. In NEXTWELL he is responsible for research and development.

Netwell commercializes a very interesting product called Clipen which is awfully usefull for Analytics proposes.

JUAN: What’s packet sniffer?

IVO: Generally, packet sniffer is an application (or sometimes appliance) that captures packets from a network and stores them on a hard drive. The term sniffer has probably its origin in the fact the application usually runs in promiscuous mode, when each packet present “on the wire” (i.e. passive voice for present) is captured and stored. Packet sniffers are known to be improperly used to capture user passwords or other private information especially if they are sent as plain text which is even frequent these days. Nevertheless, packet sniffer is a useful tool mainly to inspect network communication, to analyze network or network application errors, to monitor network application performance, and last but not least very efficient source of input data for web analytics.

JUAN: How does this technology apply for Analytics?

IVO: Packet sniffer is placed between client’s computer and server machine and thus it has unique ability to get complete information about interaction made between users and the server application of the website. Packet sniffing as discussed here is just the principle to acquire information needed for web analytics, this does not mean each sniffing application is automatically applicable for this purpose. So if I speak concretely about our clickstream processing engine technology called Clipen, the main mission this technology serves is to provide high-quality clickstream data that are completely processed and sessionized in near real-time. We are fully aware of how much results of an analysis depend on the quality of the input data, and this awareness (together with scalability) is reflected across the entire conception of the technology. From an architectural point of view the engine plays role as an infrastructural component of web analytics and data warehousing solutions. The most hidden and important component to achieve optimal analytical results based on complex, quality and near real-time available input data. To support this, the Clipen technology ensures both procedural and performance aspects of the clickstream data processing.

JUAN: What´s advantages and/or disadvantages does it has over Page Tagging or Log Analyzers?

IVO: Web logs are inaccurate and incomplete data source from its nature and to integrate logs from several machines of same website may be terrible task with uncertain results. Page tags, on the other hand, must be embedded into each page you want to track and this task may be really painful. Tag also tends to under-measure traffic. Also, not every on-line company wants to affect its fine-tuned e-commerce application to be affected by tagging. On Nextwell’s website I published some kind of comparison of the data collection methods. The main difference that characterizes packet sniffing used for data collection is the fact we have complete information about what’s the result of each HTTP request made by website users. Was the page downloaded completely or incompletely? How much of the page was really downloaded? How long lasted the download? As the sniffer “sits” on the network line between both communication parties (but on the server side for this case), the complete information of the client/server communication is available.

The biggest advantages of the Clipen technology are complex and high-quality data with near real-time availability, high performance and scalability, easy deployment, processing customization, content tracking (website content labeling). The technology is completely non-invasive and does not require any website changes to enable the clickstream data processing. Packet sniffing is entirely transparent for the tracked server and does not affect its operation.

As disadvantage somebody may consider fact the technology is on the server side so it may be affected by proxy servers or browser caches that do not send HTTP request to original server but serve it from cache. However, there are some proxies busting techniques that if applied within HTTP server they prevent proxy servers from unwanted caching. In addition, today websites are full of dynamic content that shouldn’t be cached anyway and that’s why on-line applications inherently act against caching of HTML pages. Influence of caching is negligible.

JUAN: How does it work?

IVO: This answer is quite technical; I borrowed it from Clipen’s user guide. Machine operating Clipen technology has to be connected to the tracked website’s network flow via particular network switch with port mirroring feature or via device like TAP. The clickstream data processing starts by creation of a network communication snoop files using Clipen sniffer. TCP connections stored in each snoop file are demultiplexed and divided into parallel tasks that are distributed to particular Clipen processing nodes to process them concurrently. Within the processing of the task there are following steps that Clipen node performs:

1. TCP streams reassembling

2. TCP streams SSL/TLS decoding (in the case of HTTPS)

3. HTTP request/response message completion

4. HTTP headers analysis and extraction

5. page/non-page recognition

6. session identification

7. user identification

8. content identification

9. storing data to Clipen database

When processed data are stored in the Clipen database they go through post-processing executed internally within the Clipen database managed by relational database management system. The post-processing is composed of the following parts:

1. Sessionization

2. Referring-referred resolving

3. Page view dwell time specification

4. Robot detection

After post-processing the output data are available for extraction from the Clipen database; the data extraction is fully driven by Clipen user’s external ETL application. Clipen is universal data production system so whatever external application may extract the processed data and use those for any purpose, typically for web analytics and data warehousing, but currently we also prepare product called Clipen Benchmark that is intended for performance monitoring of web applications and for so called “real user monitoring”.

JUAN: How do you imagine Analytics in the future?

IVO: I’m not the right man to predict the future but let’s try. What may change the future of analytics are probably the RFID technologies. RFID chips and systems are capable to bring many new applications that will shift analytics a large step ahead. On the other hand the RFID will evoke (or maybe already evokes) many issues regarding privacy. The capability to analyze very detailed information about person may also lead to more strict legal control so the technological evolution in this area may finally cause some regulation. However, it’s obvious that one-to-one marketing will get new meaning thanks to RFID technologies and analytical conveniences they will bring.

JUAN: What´s your passion?

IVO: I like sport shooting. Especially practical shooting – both handgun and shotgun – under IPSC rules. Fortunately, Czech Republic has judicious gun control laws thus shooters are not persecuted here like for example in the UK. But there is not much time for hobbies now. What is important for me is to meet friends regularly if possible and relax with them in nice pub with good beer. OK, maybe not surprise, as I live in Czech Republic where we have the best beer in the world (Note from Juan: I lived in Germany and I should say that the best beer I ever tried is Alt Beer from Dusseldorf, but I do agree to try yours just for researching proposes).

JUAN: What´s the best book you´ve read?

IVO: Probably Ken Kesey’s “One Flew Over the Cuckoo’s Nest” but there are also many others.

Packet sniffing for web analytics

Some time ago I’ve wrote the post “What are Log File Analysis and Page Tagging?” about the most popular web analytics technologies “Page tagging” and “Log file analyzers”. I received a comment from Ivo Rehberger (Development Manager, Nextwell) in which he mentioned the advantages of Packet sniffing over the other two technologies. I just omitted this third technology because I wanted to mention the two most popular, but I do agree with Ivo, this technology is as good as the other two and is important at least to mention it (with pros and cons). Continue reading →

Analytics 2.0

Much more than just tagging

Category Archives: Analytics Tips