One of my daily activities involves the discovery and managing the initiations phase for suppliers of publishing services. I know that most people don't know what to do with the "request for work" emails that hit their emails but somebody has to do this kind of job also.
I have to deal in that sense with suppliers from China, Germany, India, Maroco, Philipines and Romania and with freelancers form wherewer you could imagine.
It is a business like any other, just that the entrance bariers are very low. That should normaly drive the evolution of the industry in the god sens but unfortunately does not.
When I first got in contact with the industry most of the people involved had a god ideea of the specific tehnologies and processes. Everybody knew what an AQL was. And everybody knew that XML is the next big thing (it was the SGML era).
7 year after a project manager from Springer told me that he did not found a single provider at a fair who knew what DITA was. Today if I make a typo in an email refering a specific quality level expected on a volume, 1/3 of the receivers where not aware if that fact and will ofer me the writen level (wrongly spelled) instead of something you would expect.
What hapened? It is a matter of small prices and knowing what to ask for. They are used to be client driven. If the client ask for the lowest price they will ofer it. But that comes with a "price" also. The quality is much lower than what you would expect. In order to give you a good price they will use somebody unprepared for the job and if you don't qualify your suplier you will have a big surprize when the deliveries start coming.
In the last week I had to do two "train by email" sesions for using an OCR with supliers who did not know that using such and application is not a "press a button job" if you want to do a god job. What should I expect from the Quality Assurance team of such a supplier. I'll tell you: more than 50 errors in a batch where only 10 were alowed.
What you should do then if you need to find a supplier?
From my experience there are some simple rules that will save you job and your wallet:
1. Do you homework ad contact enough posible suppliers to chose from. If you have to find a decent supplier expect ... or may be I should say force yourself to test at least 5 for the job. If the job is something complex most probably you should double that number. If the job is big and the timeframe short ... count that and search for much more providers.
2. Prepare an set of specifications for the pilot phase. Something clear, and easy to understand. Try to create a complete package with samples of input and output, expected procesing steps and so on.
3. Define the quality level.
For data capture that involves a number of errors per 10k character (betwen 1 and 5) The industry standard is 2 errors per 10k character.
For conversion to XML, you should expect no errors at the structure level.
4. Be very specific and define what an error means in your case. If for example your project requires data capture you have to avoid "mising characters and spaces, badly recovered characters, broken paragraphs, aditional spaces and cariage returns, missing or badly recovered punctuation".
5. Send a sample of input of few pages and ask for it to be procesed in order to see if they understood the job and if they respect the quality level. The test will not cost you a dime. Is a standard process in the industry. A test will help them also to understand the dificulty of the job and to give you an evaluated price not something from the stomach.
6. Analyze the test output carefully. Usualy a supplier will do his best for the test samples. If they are not capable to give you good sample for the test you should not expect for more during the project. Let them know what was the result of the test and what was your expectancy. They're humans and ned to know what hapened. May be next time will do better if they know what was wrong.
7. Adjust you're requirements if you discover unexpected types of error in the samples.
8. Chose a set of suppliers form the best performers and try to improve the results by letting them know what was wrong. Chose the best price and tell the others what's the best price. They will try to beat that price for sure.
9. Prepare an work order and write all the info in it: specifications, error definition, expected quality, etc. Mention there if you can allow multy level quality (different prices for different quality levels). Be sure to writhe that you will not pay for batches that were not conformant with the expected quality level defined in that document.
10. Broke the job in batches that you can manage and send them one by one to the suppliers you chose in the end (you should chose a single one only if the project is small enough). Make sure you check the quality level provided at least for the first 2 or 3 batches from each supplier. If you find errors, more that expected send feedback and send back the job for rework. They should do the rework for free. When it comes back analize again ... if you still find erors you should not pay for that batch.
I shall continue this series in a new post during the next days. Good luck!
No comments:
Post a Comment