This section of our website focuses on solutions offered specifically to internet businesses.

Market Intelligence

• Price Elasticity
  We use log-linear regression models to determine optimum pricing strategies. These models incorporate competitors' prices and help identify the main sales drivers. Previous client: Buy.com.

• Competitive Intelligence
  We help pay-per-click search engines identify their best affiliates and partners in terms of click-to-lead ratio or conversion rates. Our tools also allowed us to identify several major cases of click fraud. In addition, we provide intelligence on competitors' performance. We also have access to a large amount of logfile data. These logfiles provide highly valuable information that we share with our clients.

• Reach and Frequency
  How many Ad impressions should one purchase to achieve a specified reach on a website? We have investigated the problem and found a simple mathematical solution:

  Online Advertising Mathematical Formula Reach   =   Unique Users   -   Σ   Uk   ( 1 - P )k the sum is over all positive integers k = 1, 2, etc. Uk is the number of users turning exactly k pages P is the ratio of purchased impressions by total pages

  Unique users who see the Ad n times: Σ   Uk   C(k,n)   Pn ( 1 - P )k - n the sum is over all positive integers greater or equal to n   C(k,n) = k! / [ n! (k-n)! ] are the binomial coefficients

  The formula relies on the distribution of pages per unique user during the targeted period of time. It helps determine the optimum number of ad impressions to purchase.

• Assessing Advertising Effectiveness
  When several advertising programs are running concurrently, it is difficult to assess the impact of each individual campaign. For instance, you have one spot running every week on a few selected TV shows, to promote your product. You want to know which shows work best for your product. Our approach consists of comparing weekly show mixes with traffic growth. We focus on pairs of mixes that are similar in terms of shows, but dissimilar in terms of traffic impact. We then perform a contrast analysis on these pairs of shows. This efficient methodology applies to many practical situations. This study was successfully implemented for NBC Internet.

Web Site Auditing

• Spam Elimination
  We create decoy e-mail addresses to identify spammers. In addition we will install software to verify which return e-mail addresses exist. Messages that fail the return address test are flagged as spam and deleted before you read them. These technologies identify nearly 100% of the spam with very few false positives. In some situations, another simple solution consists of filtering out all E-mail that do not contain a pre-specified password in the subject line.

• Inventory Management
  Data Shaping Solutions has developed ad inventory management techniques. In particular:
  • we have designed an Ad remnant optimization algorithm for a major internet company
  • We have helped NBCi save $1,000,000 per month, by improving keyword and ad impression inventory forecasting, in collaboration with GE's auditing team

• Quality Assurance - Alarm Module
  Have an alarm program implemented to detect traffic abnormalities based on various metrics (new users, page views, pages per visit, unique users, traffic per hour). Our program successfully handles traffic variations due to weekend fluctuations. It automatically sends an e-mail with red or blue flags when low or high numbers are present.

• Logfile compression
  By discarding useless information and then applying a compression algorithm specifically designed to process logfile data, we were able to increase the compression factor from 10 to more than 30, resulting in substantial cost savings in data storage, particularly for websites with large traffic.

• Click Fraud Detection
  This service is offered to pay-per-click (PPC) search engines, pay-per-lead programs and PPC advertisers. Previous client: InfoSpace.

Customer Profiling

• Path Analysis, User Behaviour
  We investigate the most common types of visits on your site. For each aggregated directory, we track incoming and outgoing traffic, to optimize your site map. You can perform this analysis to place internal links on the most appropriate pages. We also track entry and exit points to your site, as well as time spent per page to identify bottlenecks. This kind of analysis is underlined by Markov models of traffic flow and mathematical graph theory. It reveals the internal traffic chart of your site and allows you to uncover unusual paths. In addition, we create a keyword directory of your website by gathering kewords from all your URLs. The dictionary is then used to assign a type of action to each page request, to eventually analyse user behaviour. We have performed this analysis for Wells Fargo.

• User Retention and Traffic Prediction
  Do you attract more new users than you keep old ones? Autoregressive processes and Markov chains with new, returning and ex-users are the base models for traffic prediction. Our analysis will also reveal characteristics of faithful users, identify user segments that are too costly to acquire, and estimate the average return of a user over her lifetime. As a result of our analysis, Snap.com was sold to GE.

• User Segmentation
  What are the traffic patterns of loyal users? Segmentation is a classical analysis to detect the main user groups. We also track geographic data and will help you capture user behaviour, through cutting edge database technologies to handle large datasets. As a result of our analysis, NBC Internet did not renew a muti-million dollar deal with Netscape.

• Web Surveys
  We help you design or review your web surveys, avoiding the common pitfalls. For example: sampling pages instead of sampling users, or setting up the survey on your front door. We specialize in detecting associations or anti-associations between factors (age, sex, income or zip code, versus user interests or visit patterns), using dissimilarity metrics such as the Chi-square.