Implementation is THE Challenge 1 IT Project Success

Implementation is THE Challenge! 1. IT Project Success vs. Cost & Size 2. Data – A Perennial Problem 3. Gaps in MIS Data: Kraft Foods Case Example 4. Data Management Issues: Nestle Case Example 5. Data Cleansing vs. Data Hygiene 6. Other Big Hurdle: People Issues 7. How to Manage Organizational Resistance 8. What Works? What Does Not Work? L. Mohan 1

IT Project Cost vs. Success “Once you hit $10 million, the chances of coming in on time and on budget are statistically zero” - Jim Johnson (Standish Group) L. Mohan 2

More Recent Statistics …. üIn Budget, On Schedule and Met Expectations – 28% üOver Budget, Over Schedule and with Less Functionality – 49% üCancelled or Failed Prior to Completion - Source: Standish Group, 2000 – 23% L. Mohan 3

IT Project Size vs. Outcome L. Mohan 4

IT Project Size vs. Duration L. Mohan 5

IT Projects & Bridge Building Ø Bridges are normally built: § On-time. . . § On-budget. . . § And don’t fall down Ø Software § Never on-time. . . § Never on-budget. . . § Always break down L. Mohan 6

Why are Bridges Successful? Ø Bridges: § Extreme detail of design § Design is frozen at beginning § Contractor has little flexibility on changing the specifications Ø Hard model to use in today’s fast moving business environment L. Mohan 7

Other Differences Ø When a bridge falls down: § It is investigated § A report is written on the cause of failure Ø When a IT project fails: § Covered up § Ignored § Rationalized We don’t seem to learn from our mistakes! L. Mohan 8

Implementation of Enterprise Systems - Two Big “Non-IT” Hurdles: 1. Data Problems – Cleansing the “dirty” data in legacy systems – Integrating the data from the silos 2. Organizational Resistance to Change – Getting Buy-in: Unfreezing Moving Refreezing (Lewin-Schein Model of Change) – Performance Measurement and Reward Systems L. Mohan 9

Murphy’s Law for Data The Data You HAVE Is NOT The Data You WANT Is NOT The Data You NEED Data problems are more difficult to solve than hardware and software problems. L. Mohan 10

The Data Problem for Analytical Systems Ø Needed data is not collected Ø Useful data, e. g. , time spent by sales people, are not captured in the computer Ø Available data is not usable because it is poorly organized, not timely, poorly analyzed, difficult to access and internally inconsistent Ø The focus of information systems has been on hard-data-oriented applications L. Mohan 11

Gaps in MIS Data - An Old Problem Citibank “We found … that management … didn’t even have a good profile of its market and customers. It didn’t really know in summary form what (its position was) with respect to discrete market segments … There was very little account profitability and not even market segment profitability information. ” General Electric “Information on orders, sales and margins … are of maximum value when tied to … meaningful market segments. And segment-based data are of limited use to finance, hence the common misalignment problem between finance and marketing. ” L. Mohan 12

A More Serious Problem. . . Data That is NOT Available a“Soft” Data relating to. . . Ø Customer’s Buying Process Ø Reasons for Infrequent Purchase Ø Reasons for Defection Ø Quality of Customer Support a Who Should Collect This Data ? Ø People at the Customer’s Touch-Points . . . Field Sales, Telesales, Service, Call Centers, Storefront, . . . L. Mohan 13

Kraft Foods: Missing Data for a BI System üBI Objective: To evaluate sales impact of trade promotions üAvailable data on promotions: How much was spent? When the bills were paid? üMissing key data: When were the promotions run? . . . to correlate with sales data ü Data problem is solvable in principle But. . . Is it worth the effort and cost? L. Mohan 14

How to Assess Cost-Effectiveness of Missing Data - A Pragmatic Approach Design a Prototype scaled to the barest minimum Collect data for the Prototype - Lowest data cost Develop Prototype using real data Users evaluate benefits of system Stop “No Go” Value vs Cost? L. Mohan “Go” Full-blown System 15

The Low-Cost Prototype - To Assess Value of Data Ø Model limited to the core variables Øsales, promotion expenditures and dates, margins Ø Detailed data still needed for useful analysis Øby packs for each brand by markets Øweekly data for capturing sales fluctuations Øtwo years of data to compare pre- with post-deal sales levels Ø Cost of data ØManual effort to extract dates of promotions from logbooks L. Mohan 16

Quick & Dirty Prototype - The Barest-Minimum ü Limited the System to: - 2 brands, a major brand a new brand - 8 markets (out of 50), 3 large, 3 medium and 2 small ü User saw value of collecting the missing data ü Led to the development of a promotion-event calendar system L. Mohan 17

The Data Integration Problem Required data resides in different incompatible systems Compounded by discrepancies in data definitions and coding schemes Building an Integrated Data Base Takes Significant Time and Cost BUT, Data for Data’s Sake is a Worthless Luxury -- Users should get value from the Data L. Mohan 18

Data Management Issues Ø Develop Relevant and Clear Data Definitions Coding Standards and Ø Streamline Procedures for Data Collection and Flow : Ø Eliminate unnecessary paperwork Ø Ensure timeliness of data Ø Assign responsibility and authority to a specific individual: The Data Administrator Ø A demanding job for which appropriate rewards must be given L. Mohan 19

A Common Source of Data Problems - Autonomous Divisions: Nestle Case ü Before 1991: A collection of independently operating brands such as Carnation and Stouffer’s, owned by the Swiss-based parent ü After 1991: Brands were unified and reorganised into Nestle USA ü Still, Not a Single Entity: New company continued to function more like a holding corporation ü Autonomous Operation: An Insurmountable Hurdle Divisions located in dispersed headquarters offices made their own business decisions ü Bottom-Line: – 9 Different Ledgers, 28 Points of Customer Entry – Multiple Purchasing Systems: No clue of total business with a particular vendor – Every factory had set up their own vendor masters and purchased on their own. L. Mohan 20

Lack of Common Data Standards - 29 Brands of Vanilla ü 1997: Took stock of various systems across the company – Found troubling redundancies ü An Example: 29 Brands of Vanilla ! – Different brands were paying 29 different prices for Vanilla, a synonym for bland – To the SAME VENDOR ! “Every plant would buy vanilla from the vendor, and the vendor would just get whatever it thought it could get. And the reason we couldn’t even check is because every division and every factory got to name vanilla whatever they wanted to. So you could call it 1234, and it might have a whole specification behind it, and I might call it 7778. We had no way of comparing. ” “My team could name the standards but the implementation roll-out was at the whim of the business. ” CIO of Nestle USA Source: CIO Magazine, May 15, 2002 L. Mohan 21

LOCKHEED EXAMPLE What is “Sales”? “The importance of standard definitions can be illustrated by the use of the word “sign-up. ” In general, the term refers to a customer’s agreement to buy an aircraft. However, prior to the establishment of a standard definition, it was used differently by various organizational units. To marketing people, a sign-up was when a letter of intent to buy was received. Legal services considered it to be when a contract was received. Finance interpreted it as when a down payment was made. The standard definition of a sign-up now used is a signed contract with a non-refundable down payment. ” L. Mohan 22

Get the Dirt on Your Data – Dirty Data: Incorrect Data; Missing Data; Misplaced Data … It’s Everywhere ! 2006 Poll of 1, 160 knowledge workers by Harris Interactive Inc. … 75% of respondents “made critical decisions based on faulty data” – A Simple Check: Look at Trends … “If you’re looking at data that seems too good, too bad, or just too strange to be true, it probably is. ” Financial reporting systems manager at ABB Inc. Source: Computerworld, Sept. 11, 2006 L. Mohan 23 -

“Cleaning House” – An Action Plan 1. Determine which types of information must be captured Form a data mapping committee – but keep it small or risk never reaching agreement 2. Find mapping software that can harvest data from different sources such as legacy applications, PC files, HTML files, unstructured data sources and enterprise-wide systems (ERP) 3. Start with a high payoff project inside a business unit that is a big revenue generator for the company - you will get the cost justification for the buy-in from the top 4. Create an ongoing process for data hygiene - to keep the data clean. L. Mohan 24

Process Change Is Another Hurdle Employees, especially touch-point personnel, have to change the way they work Ø how they collect data from customers Ø quality of the data collected Ø how to customize the products and services offered to the customers Big roadblock for CRM implementation Ø Why should a sales rep record details of his customer contacts because it doesn’t help him sell more? He sees it as filling out forms just for the sake of filling out forms. L. Mohan 25

No. 1 Implementation Problem: Resistance to Change Ø Change in Process Typically Results in: Ø Changes in peoples’ jobs Ø Changes in required skills Ø Most Important: Ø Must consider what people think, what they believe is important and what motivates them Ø Align these with the new processes Ø May require changes in measurement and reward systems L. Mohan 26

A Reengineering Failure - Ignored People Issues ØLevi Strauss Ø Time to fill orders too long Ø Embarked on BPR project 1. Reduced time from 3 weeks to 3 days ØBUT…. Ø Created extreme turmoil by demanding that 4, 000 workers re-apply for their jobs as part of a reorganization into process groups Ø BPR project timetable stretched to 2 -years Ø Had to make repeated promises for “no layoffs” Ø Spent an extra $14 million for a 2 - year “education” effort to “calm” employees Wall Street Journal, Nov. 26, 1996 L. Mohan 27

Hammer* Acknowledges…. • . . . Reengineering forgot about people. I wasn’t smart enough about that. I was reflecting my engineering background and was insufficiently appreciative of the human dimension. I’ve learned this is critical. Ø Expanded BPR three-day “basic training class” to five Ø Two more days for “people issues” * Michael Hammer, the Reengineering Guru L. Mohan 28

One Key to Sell Change Ø Get key people involved ØBuy-in from opinion leaders would persuade others ØBetter to get criticism from the inside, than resistance from the outside ØLet them take some ownership of the project Ø Participation creates a feeling of control - Instead of “them” doing it to “you”, “we” are all doing it together L. Mohan 29

Managing Organizational Resistance 1. People whose data the BI requires - Include the data owners in the project team - Incorporate capabilities in BI to benefit data feeder 2. People who do not have access to the BI - BI is becoming Everyone's Information System - Access to data pertaining to knowledge worker's responsibility 3. People who fear too much visibility into their operations - One solution: vertical security - BI is for Everyone 4. People who see no need to change things L. Mohan - Educate and persuade resisters 30

A "Good" BI Sponsor 1. AUTHORITY and INFLUENCE - Because an BI system has far-reaching organizational impacts. 2. VISION and COMMITMENT - To position the BI where it has maximum leverage - To provide adequate support during implementation. 3. "RIGHT" ATTITUDE to COMPUTERS - As a tool for improving managerial efficiency and effectiveness. L. Mohan 31

Problems With The BI Sponsor Ø No Time to work with BI team Ø System drifts along steered by MIS people Solution: Designate Operating Sponsor - Time and knowledge to direct the BI - Attend to the political hurdles L. Mohan 32

The BIG Implementation Traps 1. Data Quality Matters “During the next two years, more and more global business will be conducted via automated decisions and automated processes – all relying on the data residing in corporate databases. But, given the poor state of data quality today, that’s a scary thought, according to a study of PWC in New York… Study found that 75% of the 599 companies surveyed experienced financial pain from defective data. ” 2. People, Process & Politics - More daunting than the technical challenges of building BI system Quaker Chemical CIO: “It was 9 to 10 -year process for implementing BI… not a simple matter of installing the applications and giving workers access to them… because BI required employees to collaborate and share information on a global basis, the company decided to tie employee’s pay to their level of cooperation on the project. ” L. Mohan 33

Technology is NOT the Solution a Place more emphasis on : Business Process Execution vs. IT Infrastructure a Do Your Homework ! Ø Identify High Value-add Applications Ø Pilot Test Analytical System. . . Measure returns. . . Proof-of-concept. . . Top Management buy-in a Pay Heed to Non-IT Hurdles Ø Getting Right Data Ø Organizational Barriers Ø Performance Measurement & Reward Systems L. Mohan 34

Connect All the Dots… 1. Identify the business problem 2. Set specific goals to address the problem 3. Develop the strategy to achieve these goals, incorporating all relevant groups and processes 4. Implement in a series of small steps to get value We derive more long-term value by trying to hit singles instead of home-runs. L. Mohan 35

Low Risk Approach To Build Analytical System - Works Especially When IT Budgets Get Slashed Ø Break it into smaller modules with a clearly defined benefit for each Ø Pick the first module with significance for proof-of-concept and top management buy-in Ø Phased implementation will harvest benefits to provide incentives for continuing the BI project Best Way To Build a Big System - Not to Build It…. . Let It Evolve L. Mohan 36

Serious BI Mistakes 1. Doesn’t do the Job 2. Time and Cost Overruns 3. Can’t Trust the Data or Outputs 4. Users Hate It – Won’t Use It L. Mohan 37

1. BI Doesn’t Do The Job! WHY? Ø The project-scope was not clearly defined. Ø The software lacked the functions or was forced to fit the job. Ø Software modifications were not fully tested. Ø Users were not properly trained on how to use the new tools in their jobs to replace existing tools. ACTION ITEMS: Ø Define the project requirements with the help of key users who are hungry for a BI to do their jobs better. Ø Train the trainers, the early champions, who can train, and provide ongoing support, to the other users Ø Training is needed not only to navigate the BIS but, more important, on how to use it in their daily tasks Ø Pilot test of BI BEFORE rollout is a MUST! L. Mohan 38

2. Users Hate It – Won’t Use It! WHY? Ø System is not user-friendly - The system interface is unfamiliar and not easy to user Ø Technical considerations took precedence over users Ø Users are not motivated to use the system due to lack of attention to change management ACTION ITEMS: Ø BI projects should NOT be led IT Ø Prototype the system with hungry users - They can “test-drive” it and provide constructive feedback to improve the usability and usefulness of the system. Ø Critical to get “buy-in” from user group by making them recognize the need for change and demonstrating “What’s in it for me” L. Mohan 39

3. Can’t Trust the Data or Outputs! WHY? Ø Data in legacy systems is not clean. - problems created when data is transferred “as-is” to HRIS Ø Not enough validation edits in the data entry screens to stop entry of patently incorrect data Ø No incentive for data feeders to enter correct data Ø System tested with only dummy data before rollout ACTION ITEMS: Ø Must include a Data Quality Plan for which significant time and money should be budgeted in the BI project Ø Test the quality of legacy data and cleanse it BEFORE moving it to BIS Ø Implement an ongoing process for data hygiene - to keep the data clean Ø Data feeders must benefit from the system Ø System test with REAL data is a MUST before rollout L. Mohan 40

4. Time and Cost Overruns WHY? Ø Unrealistic estimates for both time and cost Ø Hidden costs in project implementation Ø Changes in project scope ACTION ITEMS: Ø SBUs must take ownership of the BI project to get payoff from IT investment Ø Project oversight must be at TWO levels - Executive Steering Committee, including top management of vendors, for monitoring progress and approving any changes in project scope and the cost/budget implications - Project Management Office with SBU and IT managers jointly overseeing the project work - Must budget specifically for hidden costs in: Training, Integration of BI with other systems, Conversion of Legacy Data & Data Analysis L. Mohan 41