Research question and scope
This protocol measures how much rework follows an incomplete intake record. The unit is one intake record that enters a work queue and later requires correction or repeat contact. The review counts the rework events that can be traced to a missing or wrong field. It does not attempt to measure every cause of rework, only the part linked to intake.
Set the scope to one queue and one intake form for a fixed window. Record the form version, the required fields, and the systems used at intake. If the form changes during the window, keep the earlier version for the full sample or split the period. A new required field can change the rework rate without changing the underlying work.
Keep the intake form version with the record because a later edit can make an old record look complete. A field added last month cannot be used to judge a record created before it existed. The version is part of the evidence, not a technical detail. Without it, the comparison across periods is invalid.
Record whether the intake was completed by a person or a form that the customer filled in directly. The two paths fail differently, and the fix may be different. A customer-completed form can omit a field that a trained specialist would have asked for. Report the path beside the rework rate.
Population and unit of analysis
Define the population as every record created during the window that reached a work queue. Exclude records that were cancelled before any work began, and count the exclusions. Include records that were corrected before the customer noticed, because hidden rework still consumes capacity. Include records that were reopened after closure if the reopening traces to an intake field.
The unit is the record, not the individual correction. One record can produce several rework events, and each should be counted with its own cause. A record with two missing fields is still one record but two field-level findings. The analysis should report both record-level and field-level results so the reader can see the spread.
Record whether the rework was caught before or after the customer was affected. Hidden rework still consumes time, but visible rework has a different cost. Report the two separately so the owner can see the full burden. A low visible count can hide a large internal correction load.
Count the repeat contacts that happened through a different channel from the original intake. A customer who called first and then emailed may create two records that look unrelated. Link them by the customer and the request. Unlinked channels are a common reason rework is undercounted.
Definitions of completeness and rework
Define completeness against the fields the next step needs, not against every field on the form. A field that no one reads should not count as required. Define rework as a correction, a repeat contact, or a manual lookup that would not have been needed if the intake record had been complete. Name the specific field and the specific event for each case.
Separate rework caused by intake from rework caused by a later customer change. A customer who moves a delivery date is not an intake failure. A missing phone number that forces a second call is an intake failure. Apply the same rule to every record and publish two boundary examples so a second reviewer can classify consistently.
Define a wrong field as clearly as a missing field. A phone number with a transposed digit is an intake defect even though the field is present. Include wrong values in the completeness check and name the specific error. A form that only counts blanks will miss the errors that cause the most repeat contact.
Define the minimum useful record for each work type rather than one universal standard. A billing record and a delivery record need different fields. Applying one standard to both will either over-collect or miss the field that matters. Publish the minimum for each type and review it with the owner.
Sampling and data collection
Draw the sample from the intake system and join it to the queue and the communication log. Record the export date, the join key, and the row counts before and after matching. If a record cannot be matched, put it in an exception list rather than dropping it. The match rate is part of the result because unmatched records may hide rework.
For each record, capture the form version, the missing or wrong field, the rework event, the person who did the rework, and the minutes spent. Add the customer impact if one is visible, such as a delayed reply or a repeat call. Store the evidence with the record so the classification can be checked. Do not estimate minutes from memory.
Match the intake record to the communication log using a stable identifier rather than the customer name. Names are duplicated, misspelled, and shared within households. A weak match can create false rework or hide real rework. Report the match rate and keep unmatched records in the exception list.
Sample records from the busiest hours as well as the quiet ones. Intake quality often drops when the queue is long, and a sample drawn only from quiet periods will look better than the real experience. Report the time of day beside the rework result. The pattern may point to a staffing or form design issue.
Measuring rework volume
Report the share of records with at least one intake-linked rework event and the median rework minutes per affected record. Also report the total rework hours for the sample and the most common missing field. A high share with low minutes is a different problem from a low share with high minutes. Show both so the owner can choose the right fix.
Break the result down by intake channel, form version, and work type. A channel with a high rework rate may need a different script or a shorter form. A work type with high rework may need a field that the form does not ask for. Compare only groups with similar work and state the sample size for each.
Report the most common missing field and the work type where it appears. A field that is missing across all work types is a form problem, while one that appears in a single type may be a training problem. The fix should follow the pattern. Do not add a field to the whole form when only one lane needs it.
Report the rework minutes as a range when the estimate comes from a small sample. A single average can imply a precision the data does not support. Show the median and the interquartile range, or the counts if the sample is very small. The owner can still make a decision with a range.
Limitations and next action
The record cannot capture rework that happened in a private chat or a hallway conversation. It also cannot show the customer's frustration if the repeat contact was handled well. Treat the result as a lower bound on intake-linked rework. Note any informal workaround the team reports.
Give the result to the intake owner and the queue owner and ask for one change. The change might be a required field, a validation rule, or a shorter form that asks only what the next step needs. Record the effective date and repeat the measurement with the same definitions. A form that is easier to complete is usually cheaper than a correction process.
Review the intake form with the people who complete it under time pressure. A field that requires a lookup may be skipped when the customer is waiting. Consider a short mandatory set and a follow-up set that can be completed later. The goal is a form that captures what the next step needs without slowing the call.
Check whether the proposed fix adds work at intake or removes it later. A new required field may reduce rework but slow every call. Estimate both effects before the change. The best fix often removes a field that nobody uses and adds one that prevents the common correction.
