Development of New Product Introduction metrics and their application in forecasting reliability
New Product Introduction means different things to different companies and different people. The common thread is that managing introduction of a new product in line with key dates to achieve an on-time release, but can it be used to make a more realistic prediction of Reliability ?
NPI is always a major discussion point among engineers as they face a variety of challenges.
How to make the NPI process quicker and easier to manage
How to reduce costs and minimise sample size needs
How to ensure all key areas are covered in the Quality Plan
How to understand if the NPI process followed is actually effective and strong enough to prevent customer issues from new product mass production
Is it simply a tick sheet to keep teams focussed?
Although the NPI manager may achieve the target release date using the company NPI model, it is rare that NPI is rated / scored in a manner that can be used to predict reliability of the new product which directly influences success in the market?
Are you interested Now? You should be!!
A strong and detailed NPI matrix with objective measures can do exactly this, let me introduce how this is done and how Reliability Solutions has used this with clients to reduce reliance on simple reliability testing from small sample sizes which rarely finds defects.
Within the NPI model of any company, reliability testing will of course be a key item but if used on its own to make a reliability prediction, there is a high level of risk . If a strong list of key performance metrics are collected and used together with a calculation model to give them all a ‘voice’ we can create a reliability prediction model. Such a model will use a vast amount of information which is often ignored, although the key items may already have been completed and awarded a Yes or No to indicate completion.
Take for example a complex product which is managed through NPI with the key items below which must be completed and accepted before a product can be fully released into Mass production.
DFMEA, PFMEA
Design For Assembly
Guard band test results (often referred ‘Test to Fail’ to find where the product performance becomes unacceptable , basically an operational robustness test)
Early Life Reliability test (short duration high stress testing)
Accelerated Reliability Test (to assess longer term reliability)
Key component / module supplier performance capabilities
New Product Pilot Production yield levels
Product Design Quality Maturity
Software Design Quality Maturity
Manufacturing Process Readiness
Etc
We can see all of above items will have some influence on the final product reliability, so why not use them to help predict reliability in a more meaningful manner ??
Although this seems complex it is not difficult to measure all of the above with suitable models and compare against targets. When comparing against meeting targets for each key item we can then product a score for each item. An example is shown below for several of the key items above
| Ref No. | Key Factor Measurements | Scoring Method | Scoring Guide |
|---|---|---|---|
| 1 | DFMEA - Reliability | Scoring based on collective RPN's | Achieving RPN target levels |
| 2 | DFA | % DFA score | <60% (0), 60-70% (0.5), 70-75% (0.75), >75% (1) |
| 3 | PFMEA | Scoring based on collective RPN's | Achieving RPN target levels |
| 4 | Design Quality Maturity from DQA Testing (P1) | % Design Maturity | <70% (0), 70-80% (0.5), 80-85% (0.75), >85% (1) |
| 5 | Process Readiness score | % Score | <80% (0), 80-85% (0.5), 85-90% (0.75), >90% (1) |
| 6 | Early Life Reliability Test | Score based on no. def found | 0 fails (1), <5% fail (0.75), 5-10% (0.5), >10% (0) |
| 7 | Accelerated Life Test | Score based on no. def found | 0 fails (1), <2% fail (0.75), 2-4% (0.5), >4% (0) |
| 8 | Product Guardband Testing (temp, RH, Voltage, pwr cyc) | Score based on Design Margin width | Meets spec conditions (0.5), +20% margin outwith (0.75), +30% margin (1) |
| 9 | Pilot Production Process Yield Measurement | % Rolled Yield | <50% (0), 50-65% (0.25), 65-70% (0.5), 70-75% (0.75), >75% (1) |
| 10 | Key Component / Module Supplier Capability Assessment | % scoring | <70% (0), 70-75% (0.5), 75-85% (0.75), >85% (1) |
| 11 | SW Design Maturity level | % Design Maturity | <70% (0), 70-80% (0.5), 80-85% (0.75), >85% (1) |
| 12 | Product Functional Test Coverage | % Func Test / Burn-In Coverage | <80% (0), 80-90% (0.5), 90-95% (0.75), >95% (1) |
As NPI scoring is collected reference to targets we produce an NPI table like below
| Key Factor Measurements | Scoring Method | Scoring Guide | Release | 6 mths after |
|---|---|---|---|---|
| DFMEA - Reliability | Scoring based on collective RPN's | Achieving RPN target levels | 0.75 | 1 |
| DFA | % DFA score | <60% (0), 60-70% (0.5), 70-75% (0.75), >75% (1) | 0.75 | 0.75 |
| PFMEA | Scoring based on collective RPN's | Achieving RPN target levels | 0.75 | 1 |
| Design Quality Maturity from DQA Testing (P1) | % Design Maturity | <70% (0), 70-80% (0.5), 80-85% (0.75), >85% (1) | 0.75 | 0.75 |
| Process Readiness score | % Score | <80% (0), 80-85% (0.5), 85-90% (0.75), >90% (1) | 0.75 | 1 |
| Early Life Reliability Test | Score based on no. def found | 0 fails (1), <5% fail (0.75), 5-10% (0.5), >10% (0) | 0.75 | 0.75 |
| Accelerated Life Test | Score based on no. def found | 0 fails (1), <2% fail (0.75), 2-4% (0.5), >4% (0) | 0.5 | 0.75 |
| Product Guardband Testing (temp, RH, Voltage, pwr cyc) | Score based on Design Margin width | Meets spec conditions (50%), +20% margin outwith (75%), +30% Margin (100%) | 0.75 | 0.75 |
| Pilot Production Process Yield Measurement | % Rolled Yield | <50% (0), 50-65% (0.25), 65-70% (0.5), 70-75% (0.75), >75% (1) | 0.5 | 0.75 |
| Key Component / Module Supplier Capability Assessment | % scoring | <70% (0), 70-75% (0.5), 75-85% (0.75), >85% (1) | 0.5 | 0.75 |
| SW Design Maturity level | % Design Maturity | <70% (0), 70-80% (0.5), 80-85% (0.75), >85% (1) | 0.5 | 0.75 |
| Product Functional Test Coverage | % Func Test / Burn-In Coverage | <80% (0), 80-90% (0.5), 90-95% (0.75), >95% (1) | 0.5 | 0.75 |
| Average Score | 0.65 | 0.81 | ||
| Correlation Factor | 0.7 | 0.7 | ||
| 17% | 9% | |||
To convert the NPI score into a predicted failure level requires a correlation factor. This is calculated doing numerous simulations of past products and how engineers feel the listed NPI items would have been scored. We build up knowledge in this way and use it to produce a range of historical NPI scores for a range of products then compare to measured field failure rates. Using all this data it will be possible to develop a correlation factor to use with the NPI scoring table.
This may all seem complex to many engineers but it actually works!! It makes New Product Risk Management Objective and Realistic which provides a MAJOR advantage to any company focussed on achieving World Class Reliability.
Where Next?
To find out more about this approach simply contact Martin Shaw on Reliabilitysolutions@yahoo.co.uk or Linkedin