Improving Data Entry Accuracy

[rs11738]

Our company has a contract to do data entry. This happens at the customer's site (the customer provides all equipment and systems used. They also control the work environment entirely). They are not at all 'modernized' as far as using OCR, double data entry, data entry screen aligned to the forms we enter... they will not change any of this (we've asked multiple times). So all data entry is done by key operators without the aid of most modern techniques.

Once the key operators are done with the files that they've data entered, the files have a chance to be selected as part of a sample for review by our QC dept or the customer could select it for their separate QC review.

Our requirement is 100% accuracy each month for all work (this is very unrealistic). However, if we do not make at least 98% accuracy, according to our contract we pay a substantial fine. Of course, we also have production standards, varying volumes of forms to process with no set predictability to know incoming numbers from week to week in most cases. If we don't meet production standards, there are also fines which are higher over the month. This system favors production, but it can't be changed by us as it's in the contract.

When we started (inherited all our key operators from the prior contractor), all key operators were hired without administering a data entry accuracy or speed test. We found that our accuracy in data entry was in the low 90's.

For the past year, we have focused on accuracy being important as production, provided remedial training, done coaching with key operators who have the most difficulties, etc. However our quality is still fluctuating between 96 - 97.9%... we've never reached 98% according to the customer. Internally we find it is running slightly lower by a percentage point or so normally - so I think that the customer's measurements are fairly accurate. Their sampling methodology isn't the best, but I can't change this.

We started to break out our errors by procedural (did the wrong thing) vs simple typos (did the right thing but made a key entry error). It is now running about 70% typos to 30% procedural. In procedural cases it would seem to make sense to provide the key operator training. This has helped in many cases as we started with a lot more procedural errors than typos. However we've reached a plateau now - no matter how much training we seem to provide, this ratio isn't changing anymore.

What can we do to help key operators with typos? We have indicated that they are expected to key at a reasonable pace and verify their work at certain points in the day - but this hasn't helped enough.

We in the QA dept have suggested many ways to address this frustration, including data entry accuracy assessment and training for every key operator not meeting at least 98% accuracy. We have approximately 120 key operators and only 10% of them reliably make the 98% accuracy on a regular basis. Many others are in the 90's, but we have some that are in the 70-80's range consistently who are making it impossible to make it to 98%. Management seems to be opposed to spend money to do this training. Instead they wanted to put these people on performance improvement plans... however many of these are denied by HR. Now all this has come to the point where we've been told that QA must bring up data entry accuracy or else.

I am stuck on other ideas to help them as the QA manager as there is not a lot of follow through or support but results are still expected by upper management.

I realize that there is a much bigger problem here than just data entry quality... however my job is specifically related to improving data entry quality - I am stuck as to what else I can do in my position to help do this.

I'm hoping someone here has insight into what might help. Please let me know - I am on a very short deadline to do this from my boss.

Thanks & I appreciate any assistance you may be able to provide.

[Marc]

Any comments or suggestions, folks?

[Jim Wynne]

Originally Posted by rs11738

Our company has a contract to do data entry. This happens at the customer's site (the customer provides all equipment and systems used. They also control the work environment entirely). They are not at all 'modernized' as far as using OCR, double data entry, data entry screen aligned to the forms we enter... they will not change any of this (we've asked multiple times). So all data entry is done by key operators without the aid of most modern techniques.

Once the key operators are done with the files that they've data entered, the files have a chance to be selected as part of a sample for review by our QC dept or the customer could select it for their separate QC review.

Our requirement is 100% accuracy each month for all work (this is very unrealistic). However, if we do not make at least 98% accuracy, according to our contract we pay a substantial fine. Of course, we also have production standards, varying volumes of forms to process with no set predictability to know incoming numbers from week to week in most cases. If we don't meet production standards, there are also fines which are higher over the month. This system favors production, but it can't be changed by us as it's in the contract.

When we started (inherited all our key operators from the prior contractor), all key operators were hired without administering a data entry accuracy or speed test. We found that our accuracy in data entry was in the low 90's.

For the past year, we have focused on accuracy being important as production, provided remedial training, done coaching with key operators who have the most difficulties, etc. However our quality is still fluctuating between 96 - 97.9%... we've never reached 98% according to the customer. Internally we find it is running slightly lower by a percentage point or so normally - so I think that the customer's measurements are fairly accurate. Their sampling methodology isn't the best, but I can't change this.

We started to break out our errors by procedural (did the wrong thing) vs simple typos (did the right thing but made a key entry error). It is now running about 70% typos to 30% procedural. In procedural cases it would seem to make sense to provide the key operator training. This has helped in many cases as we started with a lot more procedural errors than typos. However we've reached a plateau now - no matter how much training we seem to provide, this ratio isn't changing anymore.

What can we do to help key operators with typos? We have indicated that they are expected to key at a reasonable pace and verify their work at certain points in the day - but this hasn't helped enough.

We in the QA dept have suggested many ways to address this frustration, including data entry accuracy assessment and training for every key operator not meeting at least 98% accuracy. We have approximately 120 key operators and only 10% of them reliably make the 98% accuracy on a regular basis. Many others are in the 90's, but we have some that are in the 70-80's range consistently who are making it impossible to make it to 98%. Management seems to be opposed to spend money to do this training. Instead they wanted to put these people on performance improvement plans... however many of these are denied by HR. Now all this has come to the point where we've been told that QA must bring up data entry accuracy or else.

I am stuck on other ideas to help them as the QA manager as there is not a lot of follow through or support but results are still expected by upper management.

I realize that there is a much bigger problem here than just data entry quality... however my job is specifically related to improving data entry quality - I am stuck as to what else I can do in my position to help do this.

I'm hoping someone here has insight into what might help. Please let me know - I am on a very short deadline to do this from my boss.

Thanks & I appreciate any assistance you may be able to provide.

Several things come to mind, but at the top it seems that you're dealing with a bad system and expecting operators to overcome its pitfalls. You should not expect that this problem will be solved quickly or painlessly. I know this doesn't help you, but you're not going to find an easy answer.

Some factors you need to think about:

• The first and most obvious problem is that the data entry screens don't jibe with the forms the operators are using. That needs to be changed.
• People need to know something about the data that they're entering. If it's all just abstract numbers and meaningless (to the operator) alpha fields, there will be trouble. If you have a field (for example) where the entry is related to a specific type of information (such as a social security number) if the fieldname is "SS" and calls for a nine-digit number, the operator might not know that (a) it's a social security number and (b) that a typo or transposition might result in someone not getting his paycheck on time. As operators become more familiar with the meaning of the data they're entering and why accuracy is important (beyond meeting an arbitrary requirement) the number of mistakes will be reduced.
• Talk to people. What's different about the operators who have consistently high accuracy rates? What do the poor performers say about their trouble with meeting expectations? Listen to what they say about the system and its deficiencies.

If your customer and your own management aren't willing to do their part in fixing the system, they will keep getting what they've always gotten. The only way you can objectively evaluate operators is by giving them a system that's not designed for failure. In order to illustrate the problem for management, you might want to present a statistical evaluation of the current process and demonstrate what the current system is capable of doing. Do your best to make them understand that if they want it to improve, they'll need to change it and stop blaming the operators for all of the problems.

[Sidney Vianna]

Originally Posted by rs11738

Management seems to be opposed to spend money to do this training. Instead they wanted to put these people on performance improvement plans... however many of these are denied by HR. Now all this has come to the point where we've been told that QA must bring up data entry accuracy or else.

The preceding excerpt from your original post shows how dysfunctional the organization is. Seems to me that you are in a position that you simply can not win. It is like customer and management are saying: The data accuracy has to improve, but nothing else can be changed to that effect. I am sure you heard the cliche' about keep doing the same thing over and over and expecting different results.

Until customer, top management, data entry clerks and HR realize that a collaborative approach is necessary to achieve the goals, this is a battle which can't be won.

[JaneB]

Originally Posted by rs11738

we've reached a plateau now - no matter how much training we seem to provide, this ratio isn't changing anymore.

What can we do to help key operators with typos? We have indicated that they are expected to key at a reasonable pace and verify their work at certain points in the day - but this hasn't helped enough.

We in the QA dept have suggested many ways to address this frustration, including data entry accuracy assessment and training for every key operator not meeting at least 98% accuracy. We have approximately 120 key operators and only 10% of them reliably make the 98% accuracy on a regular basis. Many others are in the 90's, but we have some that are in the 70-80's range consistently who are making it impossible to make it to 98%. Management seems to be opposed to spend money to do this training. Instead they wanted to put these people on performance improvement plans... however many of these are denied by HR. Now all this has come to the point where we've been told that QA must bring up data entry accuracy or else.

I am stuck on other ideas to help them as the QA manager as there is not a lot of follow through or support but results are still expected by upper management.

I am stuck on ideas to help! This sounds like a sheerly lunatic situation: 'It is your responsibility QA - you must make things improve! Oh, but nothing can change, and we do not accept any of your suggestions'.

It's an absolute nightmare trying to do accurate data entry when the forms don't match - that alone is enough to bring on errors. Add in inheriting people (no checking) and having no support or follow-through from upper management... and it sounds like a head in the sand 'don't bother us with any facts, we're busy Managing'. Yikes.

It sounds like a mission impossible - is there any way you can help them understand this? If the keying rate is brought way down, does that bring up the accuracy? Could that help communicating the choice to be made?

[BradM]

If I may.... one particular area of interest:

Originally Posted by rs11738

Our requirement is 100% accuracy each month for all work (this is very unrealistic). However, if we do not make at least 98% accuracy, according to our contract we pay a substantial fine. Of course, we also have production standards, varying volumes of forms to process with no set predictability to know incoming numbers from week to week in most cases. If we don't meet production standards, there are also fines which are higher over the month. This system favors production, but it can't be changed by us as it's in the contract.

I can't remember my first name 100% of the time. That part needs to be renegotiated... Immediately. It needs to be 95% (my opinion). Are there any other contractors who can provide any higher...?

Originally Posted by rs11738

....However our quality is still fluctuating between 96 - 97.9%... we've never reached 98% according to the customer. Internally we find it is running slightly lower by a percentage point or so normally - so I think that the customer's measurements are fairly accurate. Their sampling methodology isn't the best, but I can't change this.

I'm not familiar with accuracy rates for your industry, but that sounds like a phenomenally high accuracy rate. So if you have 95%, and the workers produce 96-97%, then you get paid a bonus. The workers can get pizza for lunch every now and then.

Also, can you rotate them to different positions and tasks? That would seem awfully mind-numbing to stay in one position that long.

Also, maybe you can examine the nature of the typos. There might be a pattern that can be remedied.

[harry]

Sitting down there, staring at the flickering screen while keying in data day in and day out is no joke.

Have you look at ergonomics - the chair, table and the keyboard? Note that everybody is different and choices should be available to them to choose the furniture and keyboard that suits them most.

How about the environment? How about monotony? Some good tips are found in this thread: Breaking the Monotony in Data Entry

[jkuil]

As a former data manager I am amazed by the low error rate of your data entry typists. The entry speed will most likely also be a performance indicated. I think you have reduced it as far as possible by the use of behavioral methods. Any further efforts in this area will not be very cost effective and most likely reduce entry speed, getting you in trouble on the next aspect. If want to reduce the error rate further, your management has to accept investments in technological solutions: double data entry, OCR, etc.
You might implement data validation: program edit rules into your system that compare the entered values against expected data limits, checks for missing data (If yes... data are expected, if no... no data are expected), consistency checks, etc. The data entry typist can be alarmed on-line of unexpected data and be asked to confirm or correct the entry, or you can do this off line.
You might also look at the forms you have to enter. Open question raise more entry errors, than multiple choise questions. The lay-out should be simple and entry fields should be easy to identify, so they are not missed during entry. The entry forms should enforce the required format of the entry (fixed date / time formats, numerical values (e.g. entry should be 0 not "none"), required accuracy (decimals), required unit (cm or m or inch), etc). The fontsize should not be too small, with entries becomming illegible and the forms too crowded. Bottom line, whatever you can do to facilitate the entry of the form. Of course your entry screens should match the form.