This document has the following objectives: be a conversation starter during talks with existing as well as new clients provide the opportunity to ask many different different questions to clients so that we understand their context better give a clear picture what InSites Consulting is all about, covering all relevant aspects of our business (business units, spearheads, research, consulting, ...) add credibility to our beliefs and company (via client references, the books, global reach, ...) spread the word on InSites Consulting in a consistent way across our offices and sales people act as the basis for future discussions on specific methods or solutions, going more in-depth support our positioning and brand (as it is shared on Slideshare) Please only distribute in pdf format!
was part of everybody’s toolkit? What if ? Analytics RE!SET Analytics BAQMaR RE!SET █ December 16, 2010 █ Ghent
Would Adam have eaten the apple ? RE!SET Would Santa Claus need letters to know which presents to bring the kids? Would BP then have run into problems? Would we then all receive so much irrelevant information? NO! NO! NO! NO ! NO! NO! NO! NO! NO! NO ! NO! NO! NO! NO! BAQMaR RE!SET █ December 16, 2010 █ Ghent
from the start! RE!SET ANALYTICS BAQMaR RE!SET █ December 16, 2010 █ Ghent
Why run qualitative research online? “ 83% of online users, ages 18 – 54 use social media online. How do we understand what consumers think without using the channel they use most to communicate?” (Knowledge Networks 2009)
Don’t throw anything away! Re!Set Long-term discussions Video Concept testing Make your research more convenient for your respondents Co-Creation Adaptive/interactive discussions Iterative concept development Bigger, deeper insights In-depth interviews Low-incidence groups Simple Pre/Post tasking Mini-Communities Live online focus groups Diaries with probing/QA User Journeys International Research easy as pie! Ad Testing Usability studies Forum discussions New Product Development
Re!Set your boundaries Why not engage people in an environment that they feel most comfortable in – their home or office?
Let respondents reply or discuss any way they want either by replying by text , annotating a whiteboard or image, recording a webcam entry, uploading their own images, arranging a collage/moodboard and even uploading videos…
Re!Set and Empower your researchers Let them go wild!
Re!Set and empower Your respondents Let them express themselves the way they NEED to!
Use visual representations to illustrate and ask questions the way you want to! Re!Set your respondent engagement
Use visuals for deeper responses Collages and mood boards keep respondents engaged!
The time to Re!Set is NOW… “ Online Focus Groups - 40% growth in 3 years - 76% of UK and US consumers have Internet connectivity” (Research May 2009, Esomar 2009) Get an edge now for future growth The online environment is now a part of almost everyone's ‘natural environment’ - so make sure your research is relevant!
Brands Re!Setting as we speak Website usability Marcomms Testing Concept Testing Ad Testing Co-Creation Online Collaboration *courtesy of Insites NV - using Visionslive.com online focus groups Focus Groups
UK/EU - firstname.lastname@example.org Ph. +44 (0)8453374484 26 York St London W1U 6PZ United Kingdom www.visionslive.com
High performance Predictive Analytics For EveryBody People usually see predictive analytics as something « inaccessible » and very expensive. We make Predictive Analytics as easy to use as simple OLAP reporting (and cheaper)! Forget other technique (OLAP/Segmentation) to analyze your data! Go straight to the best technique with garantueed ROI: Predictive Analytics!
1-click modelling & Extreme ROI 68000 € (the first week) (not using TIMi) KDD2009: (orange competition) 1% difference in the lift ≈ 250.000 € difference in ROI (This is a pessimistic estimation)
1-click modelling Best accuracy = Best ROI World-level predictive datamining competitions:
1-click modelling on any database My experience: To create a predictive model: Classical Tool: From 10 to 50 computing hours TIMi: Less than 5 minutes. KDNugget poll: What was the largest database or dataset you data-mined?
1-click modelling on any database « French Telecom » internal benchmark: 148 columns Using the « In-database scoring engine » of TIMi: 60 millions rows scored in a few minutes (compatible with teradata, sqlserver, mysql, oracle, etc.) Dataset Size Modelling: Scoring: (Millions rows) Computing Time Computing Time 0.5 1.8 minutes 5 sec. 1 3.5 minutes 10 sec. 3 10.8 minutes 30 sec. 5 18.5 minutes 50 sec. 10 19.9 minutes 100 sec.
99% of the time: you need to create a specialized dataset for each different model that you are building.
No need to clean the data (the rule « Garbage IN = Garbage OUT » does not apply anymore)
You can create ONE (very large) Repository to build ALL your models, because:
TIMi is unlimited in the number of columns.
The accuracy of TIMi does not degrade with the number of columns.
TIMi can model very small targets (less than 0.003% of the dataset is ok.)
CONCLUSION: No more Lengthy Data Preparation BUT TIMi requires a different way of working: . You should federate your data in a single very large Dataset (i.e. 50.000 columns is Ok for Anatella and TIMi).
TIMi delivers predictive models with around 10 to 15 variables.
These models have higher accuracy than any model built using any « old » predictive tools.
TIMi delivers clear Excel and Word reports that explain how these 10 to 15 variables contributes to the prediction.
Non-trained Business-Uers can read & understand the Excel and Word reports easily.
Accuracy = Incomprehensible model ? AUSDM2009 datamining competition: devoted to the study of « ensemble learning »: « Team UniQ » (1st place): a few hundreds of models combined as one « ensemble learner »: AUC= 69.72%. 1 model created with TIMi: AUC= 69.24%.
TIMi open new doors in predictive Busines-Intelligence
To Summarize, with the « TIMi suite »:
Very limited data preparation required (No data cleaning & One dataset for everything):
Predictive Datamining for Everybody.
Any Business-User can use TIMi to create 95% of the standard predictive models:
Extreme Accuracy ( top winners at KDD cups ), and thus Extreme ROI in a few mouse-clicks:
Accuracy and easily comprehensible models, at the same time:
« Near-Real-Time » datamining: (Computing time is divided by 100 to 1000 compared to other solutions: Instead of several days, we have a few minutes computing-time).
Unlimited data size: (for both the ETL tool: Anatella & the Predictive tool: TIMi )
« in-database » scoring.
Bargain price ( th of SAS/SPSS price for 4 time more licenses)
BAQMaR – RE!SET What works in market research and what needs change! Month Yr
The Catch-22 of Market Research January 2010 But bridging the gap between the two has always been a challenge… The market research industry has rested on two approaches to gathering intelligent insights: quantitative & qualitative.
Bridging the Gap Provides Insights January 2010 Capturing and utilizing both qualitative and quantitative data allows researchers to tap insights that could be lost in statistical analysis or limited by the scope and reach of focus groups. I want to re!set quali-quant research!
Personalized Targeting Through the most appropriate Marketing Channel By offering the right Product(s) Reaching the right Customer
Customized Offers Well known examples: Google
Customized Offers Well known examples: Amazon
Customized Offers Motivation: the paradox of choice 6 jams 24 jams 40% stops 60% stops 30% purchased 3% purchased S. Iyengar & M. Lepper, When Choice is Demotivating: Can One Desire Too Much of a Good Thing? Journal of Personality and Social Psychology, 2000, Vol. 79, No. 6, 995-1006 Source
Customized Offers Motivation: Overtoom facts All categories are purchased to a certain degree Most customers purchase in a limited number of categories