Tuesday, September 30, 2014

Exam and HW related Queries

Hi all,

MKTR class got done today. Whew.

Lot of new, experimental stuff found its way into the course this year.

Hence, 'twas a tad more stressful than I'd first budgeted for. It showed, I guess.

Got quite a few queries re the exam. Here's my attempt at answers:

Exam announcements:

  • You'll have plenty of time for the exam. 2.5 hours officially, you probably won't need that long.
  • *Only* those pre-reads which we tested for in the class AND which are from your coursepack are within-bounds for the exam.
  • I'm traveling from tomorrow, won't be available on campus. Limited email access.
  • No intention of making a sample paper solution. Pls consult your peers to compare answers.

HW Announcements:

Session 8 HW is due tonite. Try to get it done ASAP. Submitting by tomorrow midnight is also OK.

  • In part 2, take a screen-shot of your named-file network and paste as image onto your PPT. No code to graph your anonymous file was sent.
  • In HW part 2, there was a Q on finding out the network metrics for select subclusters. Ignore it. We forgot to send code to enable that one.
  • Seems folks with 1000+ friends are having trouble pulling data because the token expires before the pull is complete.
  • My advice is to borrow or use a 3G connection for such a data pull. ISB firewalls student access to file downloads and its causing much delay even to install packages.

Project Announcements:

Pls find the project related detailed document here.

Whereas in Homeworks, you've focussed on one tool at a time, the project requires you to bring to bear multiple tools for the same problem.

Important perspective emerges when stitching together different tools for the same substantive business problem.

Use any two tools - no need for large samples or big data collection. You have enough data collection options in your repertoire already.

Unfortunately, the project deadline cannot be pushed further. Pls finish and submit on time.

JSM interpretation explanation:

Some things to remember when interpreting JSMs.

One, when evaluating dots (brands) against attributes (arrows), always drop perpendiculars from the dots to the arrows.

Two, further away from the mean in the positive direction is a brand's projection on an attribute, higher its value on that attribute.

Three, the preferences (pink lines) are akin to attributes (i.e., are lines). Thus, when evaluating brands (dots) against preferences (arrows), drop perpendiculars as usual.

Four, the angle between two arrows (attributes) reveals how correlated they are.

Four, its tricky to try comparing individual preferences (pink lines) against attributes (blue arrows). If one really had to, taking the angle between them is the way forward.

Consider the Qs below from your practice exam. (click for larger image)

For Q 7.1., I'd look at Table 7.1's "Palm V" column. Highest value there, 8.5, corresponds to Light_weight attribute.

For Q 7.2., I'd look at Table 7.1.'s last row, "Preference". Highest value there, 6.6, corresponds to Compaq_IPac.

For Q 7.3., I'd look at Figure 7.1. The highest brand shown there on both attributes, is Sony_CLIE.

For Q 7.4., I'd look at the correlation table given in Table 7.2. In the row "memory", highest value, 0.93, corresponds to "software". In the figure, the same is shown as ThirdPartySW.

For Q 7.5., I'd look at Figure 7.1. The preference attribute has the smallest angle (hence, concordance) with R50's preference vector. Note: Such inferences are dicey, but we're doing it here anyway.

For Q 7.6., I'd look at Figure 7.1. for the smallest angle between preference vectors and the attributes mentioned. Again, note: Such inferences are dicey, but we're doing it here anyway.

I'd say R22 and R30 respectively. However, in the exam you can expect no Qs linking the pink lines and the blue arrows.

Hope that helps.

Good luck for the exam. And beyond.

Sudhir

Sunday, September 28, 2014

Analysis Results for the HW posted on LMS

Hi all,

I've received feedback saying that getting the R code to run has become an issue for several reasons.

MKTR's value-add is (or should be) imparting hands-on practice in:

  • problem formulation
  • opportunity recognition
  • a generalist understanding of methods available
  • tool selection
  • interpretation skills
  • and big-picture perspective.

Your getting stuck in R code execution issues doesn't serve that purpose.

So, I'm running the R analysis on my machine and sending you what I think are the results you need (and then some more) for "solving" the HWs.

Mind you, these are not "solutions" in themselves, merely intermediate results.

Your task of thinking through the HW problem, interpreting the results and putting it all together to highlight insights obtained remains critical in solving the HW.

Pls find on LMS a folder titled "R analysis results for the HW" or something like that.

It should contain the results from my R runs for:

  • Session 5 HW (individual part)
  • Session 6 HW
  • Session 8 HW part 1.

Pls let me know if anything else is required.

Sudhir

P.S. When we first wrote and checked the code, we assumed students' web access rights would be much the same as ours. Seems we were mistaken.

Then, new hope was shinyapps would solve the problem. Mistaken again.

Saturday, September 27, 2014

Session 8 Homework

Hi all,

Pls find uploaded in LMS in 'session 8 folder' two R code files and one data file.

The code and Homework is self-explanatory, for the most part.

Pls follow instructions and use the blog comments or email aashish_pandey@isb.edu in case of any queries.

You HW has two parts. Both individual.

Part 1: Mapping Brand communities based on Affiliations

Run the code in 'session 8 HW code 1.txt'.

Generate the brand association graph with weighted edges.

Run the community detection algorithm on it.

Answer the following Qs in a few slides.

  • Which brands seem most 'central' to the network?
  • Which 'space' is this? What might brand locations in this space signifiy?
  • How many brand communities arise?
  • Pick any one brand cluster/ community. Interpret what might be the most important affiliations or attributes driving clustering in it.
  • How many singletons (or single member clusters) are there? Speculate on why these brands seem isolated.
  • If you're the AXN channel, which brand community do you think is best aligned or affiliated with you? Why?

Part 2: Mapping your Facebook friends communities

Run the code in 'session 8 HW code 2.txt'.

Get access token for FB API as was demo-ed in class.

Pull and save the data - a named friends' list, an anonymized friends list and a listof pages liked.

In a few slides, answer the following Qs:

  • Do you see natural groupings arise? Which ones are they? ID them like I had done for my FB example in class and paste that on a Slide.
  • List some common network properties associated with your FB graph (transitivity, Avg path length)
  • How many communities arise? Average size?
  • How many singletons (or single member communities) are there?
  • Pick a community to analyze. Use anaonymized graph for this. What are the network peoperties for this community?
  • Put your marketing hat on. Write two Qs that you as a Marketer might be interested in after studying the graph.
Deliverable: Three pieces to submit - (i) A PPT that answers Qs posed above. Break PPT into two sections for each part.

Title slide should contain your name PGID.

(ii) the anonymized FB friends' list and (iii) the liked pages list.

Pls zip these into a folder titled yourName.zip and submit into appropriate dropbox.

Deadline: Midnight Tuesday 30-sept.

Tight deadline, I know. But the code runs smoothly, will be done in minutes.

My suggestion - do the HWs in a group. If any one person in your group has solved their HW by running the R code, pls borrow and share the results with peers.

After all, its the interpretation of the analysis and not running the analysis that this HW hinges on.

Part 1 is common to everyone anyway.

Any queries, contact aashish or me.

Sudhir

Friday, September 26, 2014

MKTR Project

Hi all,

Pls find below details regarding your MKTR group project.

Your Project Task:

1. Pick any substantive business problem which can in turn be mapped into 2-3 R.O.s

2. Ensure that at least one R.O. is exploratory in nature and at least one, confirmatory.

3. Among the repertoire of tools we've covered (think of the survey method, FGDs, perceptual maps, text analytics, STP, experiments etc.)...

4. ... ensure that your R.O.s map onto one tool each.

5. Collect data as required (don't aim for large samples given the paucity of time etc) for your R.O.s

6. Think of the frameworks we've covered (constructs, habit patterns, network-perspective, hypotheses testing)... 7. ... and apply them to solve the R.O.s.

Deliverable Format:

One PPT, no more than 20 slides long (excluding annexures).

First slide should contain group name and membership and project title.

Second slide explains the business problem context and lists the R.O.s

Next few slides constitute the "Methods Section", outline the steps taken to address the R.O.s and the tools employed.

Next few slides are the "Data Section", summarize the type and amount of data collected.

Then follows your "Results and Discussion" section, wherein you lay out your results, interpretation etc.

Finally, end with a "Insights and recommendations" section.

Name the PPT as yourGroup.pptx and submit into teh appropriate dropbox.

Deadline:

The last midnight before Term 5 starts. (12-Oct midnite, I think).

Pls find below a list of Grading criteria likely to apply in project evaluation:

1. Quality of the Problem context chosen - Creativity; alignment with the rest of the project; How well can it be resolved given the data at hand. Etc.

2. Quality of the R.O.s - How well defined and specific the R.O.s are in general; How well the R.O.s cover and address the business problem; How well they map onto specific analysis tools; How well they lead to specific recomendations made to the client in the end. Etc.

3. Clarity, focus and purpose in the Methodology - Flows from the D.P. and the R.O.s. Why you chose this particular series of analysis steps in your methodology and not some alternative. The methodlogy section would be a subset of a full fledged research design, essentially. The emphasis should be on simplicity, brevity and logical flow.

4. Quality of Assumptions made - Assumptions should be reasonable and clearly stated in different steps. Was there opportunity for any validation of assumptions downstream, any reality checks done to see if things are fine?

5. Quality of results obtained - the actual analysis performed and the results obtained. What problems were encountered and how did you circumvent them. How useful are the results? If they're not very useful, how did you transform them post-analysis into something more relevant and useable.

6. Quality of insight obtained, recommendations made - How all that you did so far is finally integrated into a coherent whole to yield data-backed recommendations that are clear, actionable, specific to the problem at hand and likely to significantly impact the decisions downstream. How well the original problem is now 'resolved'.

7. Quality of learnings noted - Post-facto, what generic learnings and take-aways from the project emerged. More specifically, "what would you do differently in questionnaire design, in data collection and in data analysis to get a better outcome?".

9. Creativity, story and flow - Was the submission reader-friendly? Does a 'story' come through in an interconnection between one slide and the next? Were important points highlighted, cluttered slides animated in sequence, callouts and other tools used to emphasize important points in particular slides and so on.

Any queries etc, pls contact me.

Tuesday, September 23, 2014

Pre-reads for Sessions 8 and 9

Hi all,

A few quick announcements.

There's been a change in session schedule. I'm swapping sessions 8 and 9.

So, next we'll have Network Analytics, which was originally session 9 (and now will be session 8).

--------------------------------------

Pre-reads for session 8 network analytics:

There are two, both from McKinsey Quarterly.

The first is an article called "Demystifying Social Media" from April 2012. Pls free-register with the site to download a PDF.

Optional: The second is a 12-minute video, again from McKinsey on "Making sense of Social Media".

More McKinsey speak on their Social Media Marketing framework.

Worth a quick watch, if you want to further explore the themes that came up in the first pre-read above.

--------------------------------------

Pre-read for Session 9: Hypothesis-testing and Experimentation

Reading No. 9 "How to design smart business experiments" in the coursepack.

Special Reading for Today's regression piece.

I did get the sense that folks weren't entirely comfortable with the basic regression part in today's class.

As a refresher to your stats core, pls consider reading the first few pages of reading no. 12 in your coursepack "Practical regression - regression basics".

I hope that helps.

--------------------------------------

Announcement: R tutorial

Will be held in AC4 7-8 pm on Thursday, 25-Sept. (pending venue confirmation).

Plan is to walk you through basic input-output and help functions in R.

Any and all Qs and doubts and installation and implementation and execution issues are fair game.

--------------------------------------

Running the tm.plugin.webmining package for session 5 homework.

Several folks have contacted me with issues in running this.

Pls ensure you have tm and tm.plugin.webmining installed.

Further, run the code on base R and not on Rstudio.

I've updated the code in session 5 HW blog post.

Sudhir

Saturday, September 20, 2014

Keeping Track of the various submissions due

Hi all,

One of you, SS, wrote the following to me:

Dear Prof,

We have received quite a few mailers on the MKTR assignments over the last week and we are finding it hard to track these separately and fish out for all of these in the blog.

Request you to kindly create a spreadsheet or a word doc in your blog (with a dedicated section for HWs) and put all the HW details and deadlines on it.

We usually have such a document floated for other subjects so that it becomes a one-stop shop for looking up the deadlines for a course.

Thanks for your help.

My response is as below: (click for a larger view)

Sudhir

Session 9 Network Analytics Homework Survey (Time sensitive)

Hi all,

I just realized that due to oct-2 Gandhi Jayanti holiday, session 9 is pushed forward from 30-sept to 26-sept.

That gives me precious little time. Need to collect two sets of data from you for session 9:

(a) One on self-reported snowball-sampled social interaction data (see survey below), and

(b) one where you pull your data from facebook for analysis (friend-lists and pages Liked)

I'll go over the how to of the facebook data pull in class in the first few minutes of session 7.

Meanwhile, pls take the one last survey filling exercise remaining at the link below:

Social network analysis Co2015 Data collection survey

Deadline: The deadline for the survey is Monday 22-sept Morning 6 a.m.

This is time-sensitive, Pls don't keep it till the last minute.

Shouldn't take more than 15 minutes on average. Qs take you through both objective and hypothetical scenarios.

Pls think through before you answer. Full credit *only* for complete and timely submissions.

This homework will have more credit than the other surveys you filled up because of its slightly higher length and complexity.

Any Qs etc, pls use the blog comments section or contact me.

Sudhir