Introduction to Celebrus Technologies
Upcoming SlideShare
Loading in...5

Introduction to Celebrus Technologies



An introduction to Celebrus Technologies, example use cases, key concepts and deployment architectures

An introduction to Celebrus Technologies, example use cases, key concepts and deployment architectures



Total Views
Views on SlideShare
Embed Views



1 Embed 1 1



Upload Details

Uploaded via as Microsoft PowerPoint

Usage Rights

© All Rights Reserved

Report content

Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

  • Full Name Full Name Comment goes here.
    Are you sure you want to
    Your message goes here
Post Comment
Edit your comment

Introduction to Celebrus Technologies Introduction to Celebrus Technologies Presentation Transcript

  • © 2014 Celebrus Celebrus An Introduction to Celebrus Technologies Ant Phillips
  • © 2014 Celebrus Introduction 2 in order to increase revenue, maximise marketing effectiveness and enhance brand loyalty. Celebrus software enables organisations to understand their customers’ interactions with their digital channels in real-time. Celebrus clients use this individual-level streaming data to power their customer analytics and drive personalisation
  • © 2014 Celebrus ▪The core of Celebrus is a high performance event collection platform: ▪ Typically deployed in-domain and runs seamlessly as part of your applications ▪ Collects events from a wide range of applications (web, Adobe, kiosk, PoS etc) ▪ Loads the events into databases and warehouses at high speed and very low latency ▪ Is completely reliable, security tested and scalable to the most demanding needs ▪Celebrus creates business meaning from customer activity ▪ Event collection understands what your customers do when they interact with your site ▪ Customer activity is a stream of low level events (clicks, form submits etc) ▪ From this data Celebrus generates meaningful business events ▪ For example: transactions, baskets, product views, wishlists and searches ▪What kinds of question can the data answer? ▪ The questions depend greatly on who is asking! ▪ Operations team: sessions/pages, hot pages, page load time, dead ends, script errors ▪ Senior management: daily purchase value, average value, unique visitors, KPIs ▪ Marketing: customer intent, basket abandon rate, hot products, top campaigns, conversions ▪ Data scientists: transaction flow, path analysis, purchasing propensity It’s All About The Data!
  • © 2014 Celebrus  Audience segmentation  Channel performance  Spend management  Governance and audit  Identifying opportunities  Single customer view  Promotions and offers  Exception management Use Cases 4
  • © 2014 Celebrus ▪Understand the customers using your web site (the audience) ▪ Divides a broad target market into subsets of similar consumers ▪ Includes historical reporting (past), trends (current) and predictions (future) ▪ There are many questions which segmentation looks to answer – for example: ▪ What proportion of my audience is mobile? ▪ Does mobile deliver or is it just used more for product browsing? ▪ How much traffic is coming from mobile devices? ▪ Is mobile use increasing or decreasing? ▪ What kinds of mobile device are being used? ▪ What are the demographics of my mobile users? Audience Segmentation
  • © 2014 Celebrus ▪Key operational statistics on channel performance: ▪ Number of unique visitors ▪ Drop off points ▪ Time on site ▪ Sales revenue ▪ Abandoned baskets ▪ Channel shift targets ▪ Hot pages and products ▪Data typically required in two form factors: ▪ Standard set of reports (on demand or emailed) ▪ Real-time business monitoring (KPIs/dashboard) Channel Performance
  • © 2014 Celebrus ▪Understand the impact of your marketing spend: ▪ Enables revenue to be correctly attributed to marketing endeavours! ▪ Quantifies the effectiveness of search engine optimisation (SEO) ▪ Calculates return on investment (ROI) for content/SEO/campaigns ▪ Discriminates between different marketing campaigns (email/web etc) ▪ Campaign attribution is a broad subject with many subjective measures: ▪ For example, how to attribute sales to a particular campaign: First campaign that brought a customer to your web site? Most recent campaign through which the customer arrived? Frequency of campaigns for a returning customer? Spend Management
  • © 2014 Celebrus ▪What has the channel delivered to the audience, and when? ▪ Were individuals shown correct or appropriate content? ▪ What recommendations were shown and for what reason? ▪ May also extend into details of the content, for example the price offered ▪ Often requires long term archiving depending on regulatory requirements ▪ Common requirement in many industries (for example, financial services) ▪ Needs the browser to confirm what has been displayed and when! Governance and Audit Identify Opportunities ▪Seeks to deliver on specific business goals and strategies ▪ For example, increase revenue by identifying cross and up sell opportunities: ▪ Is a customer a brand lover with specific product affinity? ▪ People who exhibit certain behaviours on site might lead to… ▪ How to sell more, or attract more people who might buy... ▪ Leads directly to marketing activity, individual level personalisation
  • © 2014 Celebrus ▪Communicates with individuals across channels consistently ▪ For example, don’t present something on a page if already done with an email ▪ Joins the customer journey up wherever and however they communicate ▪ Improves customer experience and ultimately retention ▪ Requires individual level understanding not broad statistics Single Customer View Promotions and Offers ▪Customises the content displayed based on the individual ▪ Promotions and offers chosen and presented in the web page in real-time ▪ Described using rules which segment customers (who) and then action (when) ▪ Out of the box integration with decision management products like Teradata RTIM ▪ Prioritised list of offers can be presented and feedback collected (click through tracking)
  • © 2014 Celebrus ▪Identify exceptional behaviours because they have business value: ▪ Spotting potential churn (for example, a customer searching for PUCs, or T&Cs) ▪ Identifying fraudulent activity on a web site (such as repeated check out attempts) ▪ Often implemented by accumulating a metric score for the individual ▪ Growing importance because retailers have less time before dispatch ▪ Catching exceptional behaviour earlier enables more options for the enterprise ▪ Path analysis for offline processing to see the journey an individual has taken: Exception Management
  • © 2014 Celebrus Concepts
  • © 2014 Celebrus Overview 12
  • © 2014 Celebrus13 Touch Points
  • © 2014 Celebrus14 Creating Business Value
  • © 2014 Celebrus Events 15 ▪A customer interacts with a web site or application: ▪ A series of things occur: page navigation, clicks, text entered, forms etc ▪ Many people use the web site at the same time ▪ This series of things from everyone online is an event stream ▪Events are categorised by their event source: Web The largest source of events covering all the events which come from the client applications – click, form submit, page loaded, key pressed, mouse over etc Action Specialist source of events involved with displaying content in a customer’s client application or web page – for example when an offer is displayed Facebook Facebook events such as adding a comment, like, unlike etc Twitter Twitter events such as follow, tweet, retweet and user status Linked In Linked In events such as share and user status Google Plus Google Plus events such as share and user status Computed System generated events, for example computing a list of products in an abandoned basket when a browsing session ends Scenario Includes all the business events which the system can generate such as transaction, basket, wishlist, and search – user defined events can also be configured which are included in this event source
  • © 2014 Celebrus Business Events 16 ▪An event stream can be enriched ▪ Higher level business events can be described by configuring Celebrus ▪ When the configuration is matched, a new event is added to the event stream ▪ The table below shows the main types of business event in Celebrus: Basket Events covering basket activities such as checkout, add, amend, failed add, remove, abandoned products, and view Behaviour A general purpose event which captures a customer who has exhibited a well defined behaviour - for example a customer who has asked for more product information or who has searched for terms and conditions Search Search, search results, search no results – captures search related activity on the web site, the search terms which customers used etc Profile Identifies individuals who use the web site, either when they explicitly identify themselves (login) or implicitly through cookies Promotion Family of events related to selecting, displaying and tracking promotions Transaction Everything to do with purchases and transactions including abandonment – a checkout process can further be refined into a series of steps Wishlist Similar to baskets and caters for web sites offering wishlists User defined Extends Celebrus to include any kind of user-defined business event
  • © 2014 Celebrus Attributes 17 ▪Events have attributes which provide detailed information ▪ Every attribute has a name and an associated value ▪ Some attributes are mandatory, others are optional and may not be present ▪ Shown below are some of the primary attributes in the BasketProductAdd event ▪Attributes are classified as either primary or context ▪ Primary attributes are the key pieces of information for that event ▪ For example, an account number in a bank balance enquiry ▪ Context attributes fill in details about the customer and their browsing session ▪ For example, the current marketing campaign attributed to the customer
  • © 2014 Celebrus Event Documentation 18 ▪The system provides HTML documentation for every event and attribute ▪ At the moment there are about 75 events and several hundred attributes ▪ The documentation includes any custom events which you have configured
  • © 2014 Celebrus Deployment
  • © 2014 Celebrus Introduction 20 ▪A Celebrus deployment cluster is typically split into two parts: ▪ One or more Collection Servers which are located in the DMZ ▪ Traffic should be load balanced between the Collection Servers ▪ Sessions should be sticky to keep traffic on the same Collection Server ▪ Configuration, Analytics and Real-Time Servers deployed behind the inner firewall ▪ Real-Time Servers can be used to provide personalised content to browsers ▪ Alternatively a separate web server or content management system can do this ▪Security is an important consideration for any deployment ▪ All network connections can be secured using SSL ▪ Port numbers for all communications are configurable ▪ Includes connections between the Celebrus applications ▪ Complete flexibility on what is collected and what isn’t: ▪ Collection rules flow all the way back to the CSA ▪ Addresses privacy and regulatory concerns ▪Servers can be added and removed to clusters as required ▪ Cluster Management Console provides complete administrative application ▪ Management status alerts easily plugged into enterprise monitoring tools
  • © 2014 Celebrus Who Does What 21 ▪Collection Servers are the point of contact for client applications ▪ Collect events as they arrive from the browsers and applications ▪ Writes the events into the events database and sends them to Real-Time Servers ▪The Configuration Server stores versioned system configuration ▪ Business event configuration stored along with security information ▪ Role based security provides authentication and authorisation ▪Analytics Servers perform batch processing of event data ▪ Results are written into a separate database ▪ Analytics processing currently supports MySQL, Oracle and Teradata ▪Real-Time Servers process the event stream in real-time ▪ Responsible for handling personalisation and real-time rules execution ▪ Real-Time export of events in a variety of formats (CSV, XML and JSON) ▪ Interfaces to external decision management systems (RTIM, Pega etc) ▪ General purpose plugins call out to HTTP, SOAP and JMS applications ▪The ABI Server stores information about each individual who visits ▪ Stateless server which manages historical information for each visitor
  • © 2014 Celebrus The Insert 22 ▪Data collection is enabled through a single static insert on your site: <script type=“text/javascript” src=“insert.js”></script> ▪Designed to be simple to use with a minimal client footprint ▪ Typically added at the top of the web page for best performance ▪ The JavaScript which runs in the browser is called a Client Side Adapter (CSA) ▪Unified collection from multiple channels: ▪ Set top boxes ▪ Games consoles ▪ Web applications ▪ Mobile devices ▪ Bespoke applications ▪Some applications have their own native CSA: ▪ iOS ▪ Flash/Flex ▪ Microsoft .NET ▪ Java
  • © 2014 Celebrus Batch Processing ▪Data collected into one or more events databases (MySQL or Oracle) ▪ Analytics Servers process the data into results databases ▪ Results databases are an ideal data source for reporting
  • © 2014 Celebrus Real-Time Processing ▪Data streams directly from Collection into Real-Time Servers ▪ Integrates where required with content and decision management systems ▪ Out-of-the-box integration with Teradata RTIM for offers and promotions ▪ Export capability loads processed events into customer database (within 60 seconds)
  • © 2014 Celebrus Cloud Collection ▪Separates data collection from the warehousing and data processing ▪ Data often required on customer premises (regulatory, security, performance) ▪ Files are transferred and loaded at customer premises with very low latency
  • © 2014 Celebrus Configuration
  • © 2014 Celebrus Scenarios 27 ▪Celebrus is configured to identify and create business events ▪ Often this is something simple like a customer arriving on a particular page ▪ It may be a more sophisticated business process described by several steps ▪The configuration for an event is described using a scenario ▪ A scenario is essentially a graph which advances as steps are matched ▪ In the example below, the scenario has a single step which matches FormSubmit events ▪ When this scenario reaches the end node it creates a new InternalSearch event ▪ Matches are often constrained on one or more attributes: ▪ For example, match only forms submitted from this page location
  • © 2014 Celebrus Advanced Scenarios 28 ▪Scenarios have a partition which says when they are created / discarded ▪ The partition is either the current page, the current window or the whole session ▪ Some scenarios only make sense when the partition is a session or window: ▪ For example, a scenario which matches on several page load events ▪Scenarios also say how many times they can repeat (in a partition): ▪ The default for a scenario is to partition on session and allow multiple repetitions
  • © 2014 Celebrus Rules 29 ▪Rules allow customised promotions and offers to be presented: ▪ Triggers segment the audience into smaller sets of customers ▪ These triggers are specified using easy to understand phrases ▪ Options allow for control groups, testing (by IP address) and date ranges ▪Built-in Content Library where images and other content can be stored ▪ Content can also be picked up from a separate Content Management System (CMS)
  • © 2014 Celebrus Advanced Rules 30 ▪Rules can also be used to send data to third party systems ▪ The third party system can optionally choose the content to be presented ▪ Third party systems can be interfaced using HTTP(s), SOAP and JMS ▪The Real-Time Server provides a SOAP interface for querying ▪ Primarily used by web applications which want to build customised web pages ▪Rules can even run when customers are not currently browsing ▪ Offline rules periodically sweep through all known customer details ▪ For example, an offline rule can be used to send follow up emails ▪Rules can limit how many are times they are allowed to happen ▪ All actions applied to a customer are stored as part of the individual’s history
  • © 2014 Celebrus Data Loading
  • © 2014 Celebrus Introduction 32 ▪The process by which event data is delivered to target data systems ▪ The target can be a traditional data warehouse for reporting and analytics ▪ Equally, the target data system might be a discovery platform like Aster ▪ More recently, distributed processing clusters like Hadoop have become popular ▪ Celebrus can load many different types of data system out-of-the-box ▪Real-Time Servers let you capture and act on data within a second ▪ The data loading however works more slowly to maximise throughput ▪ This means we deliver the data within 60 seconds of it happening ▪ Data can be batched up into larger and less frequent loads if required
  • © 2014 Celebrus Data Loading 33 ▪Data loading is actually two separate processing steps which interact: ▪ Events are written to files as they occur in the Real-Time Server ▪ As and when the files arrive, they are loaded into the target data system ▪File export processing is therefore decoupled from data loading ▪ Makes the system resilient to problems, bottlenecks and taking systems offline! ▪ The export and load processing steps run as part of the Real-Time Server ▪ But not necessarily in the same Real-Time Server instance ▪ Makes it easy to geographically separate the two processing steps ▪ For example, cloud collection with on premise data loading ▪ In this example, the files would be securely transferred through SFTP ▪Files must match the format expected by the target data system ▪ Databases have many quirks on how files should be formatted ▪ For example: line terminator, field separator, escape characters, character set etc ▪ Pre-configured formats for supported databases (MySQL, SQL Server, etc)
  • © 2014 Celebrus File Formats 34 ▪Variety of file formats are available to support a range of target systems: ▪ Real-Time Server can write the event stream in multiple formats simultaneously ▪ Each file format has pros-and-cons which are discussed in the following sections CSV The most common interchange format for database and warehouses. A general standard for the CSV file format does not exist but RFC4180 provides a de facto standard for some aspects of it. JSON Language independent text-based format popularised as an interchange format for JavaScript (web) applications. XML Widely used mark-up language for encoding documents in a format which is both human and machine readable. The specification (and a plethora of related standards) is maintained by the W3C.
  • © 2014 Celebrus Further Considerations 35  There are several considerations related to data loading: Target database The export options are often constrained by the target database (for example, CSV on SQL Server, and JSON for MongoDB). Storage requirements CSV has relatively little overhead given that each row contains just the data values – contrast with XML or JSON where each value repeats the field name. This has considerable impact on the storage requirements. Schema driven CSV is primarily used for loading traditional database tables (Teradata, Oracle, MySQL, SQL Server etc) where the database structure is well defined and (mostly) normalised – document databases generally do not enforce a schema and are therefore more flexible in what they store. Document oriented An event is conceptually one document and therefore fits well with a document database approach. Equally, the data in an event is straightforward to shred into a more normalised set of tables. How will the data be used? Last but certainly not least – it is very important to consider what the data will be used for. Is everything in an event required or are there key attributes which will be required for a report? This guides a range of considerations such as sharding, clustering, indexing and more.