Method note: this brief proposes a source-led review approach. It does not claim that private client or firm data has been collected.
Question and scope
This protocol asks where time-entry narratives fail the client's outside counsel guidelines. The unit is a narrative line that received a client edit, rejection, or clarifying question during the observation window. The purpose is to locate the recurring defect types so training and templates can target them. The result does not measure the quality of a timekeeper's overall work.
Scope the study to one client or a defined client group with a written narrative requirement. Include lines from matters that were billed in the window and exclude lines that were never submitted. State whether the study covers all timekeepers or a sample. A clear scope keeps the denominator consistent and makes the defect rates comparable across periods.
Defining a narrative failure
A failure is a client action that requires the narrative to change before the line is accepted. That includes a write-down tied to the narrative, a rejection with a narrative reason code, or a documented request for clarification. A purely internal preference from a partner is not a client failure and should be coded separately. The definition should name the evidence that qualifies.
Distinguish a missing component from an inaccurate statement. A missing outcome is a different defect from a narrative that describes work that did not occur. Inaccuracy is more serious and should be reported separately rather than folded into a general quality score. The coding scheme should preserve that distinction even when both lead to a write-down.
Population and unit of analysis
The population is every submitted narrative line for the in-scope matters during the window. The primary unit is the line, because a single invoice can contain both strong and weak narratives. A secondary unit is the matter, which shows whether failures concentrate in particular engagements. Report the line-level rate first and the matter-level pattern second.
Record the timekeeper, the matter type, the client guideline version, and the review outcome for each line. These fields allow the analysis to separate a template problem from a matter-specific issue. Do not aggregate lines before checking whether the mix of matter types explains the difference. A raw rate across dissimilar work can be misleading.
Evidence standard
Use the client's written edit, rejection notice, or documented clarification as the evidence of failure. If the client only discussed the issue verbally, record the date and the participants but mark the evidence as indirect. Do not rely on a reviewer's memory of a conversation. The standard should let a second analyst confirm the defect from the file.
Retain the original narrative and the revised narrative together. The difference between them is often the clearest evidence of the missing component. If the line was removed entirely, record the client's stated reason and the amount. A removed line without a reason should be coded as unexplained rather than assumed to be a narrative defect.
Sampling and collection
If the volume is manageable, review every line in the window. If not, sample by client, matter type, and timekeeper so the sample reflects the population. Draw the sample before reviewing outcomes to avoid selecting only the lines that were edited. Record the sampling frame and the inclusion rules in the study file. State the sampling frame and the date it was drawn in the study file.
Collect the guideline requirement that applies to each line before reading the client's edit. That order prevents the analyst from reverse engineering a rule to fit the outcome. Use the guideline version in force for the work date, not the version current at review. The version field is essential when a client changes its narrative requirements mid-year. Keep the sampling frame with the study file so the inclusion rule can be audited later, and record the date the frame was drawn.
Coding the defect
Use a short set of codes that match the guideline: missing task, missing subject, missing outcome, vague time description, block entry, duplication, and confidentiality concern. Add an other category with a required note. Each code should have a definition and an example. A code that cannot be applied consistently should be revised before the main review.
Code one primary defect per line, then list any secondary defects. The primary defect is the one the client cited or the one that most clearly prevented acceptance. Secondary defects are useful for training but should not inflate the count. Have a second reviewer code a sample and compare the results. Report the agreement rate and resolve disagreements with the guideline text.
Analysis and reporting
Report the defect rate per hundred lines and the share of each defect type. Include the amount of the associated write-down or adjustment when the data allows. A frequent defect with a small financial effect may be less urgent than a rare defect that removes a large charge. The report should show both frequency and consequence.
Break the results down by matter type, timekeeper group, and guideline version. Look for a defect that appears only after a guideline change or only in one practice area. Compare the pattern with the narrative standard and the training examples. The output should identify one or two specific fixes, not just a ranking of defects.
Limits and sources of error
Client review behavior can vary by reviewer, so an edit may reflect a preference rather than a documented requirement. Code only failures tied to a written rule or a documented clarification. Report the share of edits that could not be tied to a rule. That share is a useful signal about the clarity of the guideline and the review relationship.
The sample may overrepresent matters with unusual client attention. A client that reviews every line will produce more documented failures than a client that reviews a sample. Report the client and the review intensity beside the rate. Do not compare rates across clients without noting the difference in review practice. Report the number of lines that could not be matched to a guideline version and explain why.
Retention and privacy
Keep the original and revised narratives in the controlled billing environment. The study file should contain only the fields needed for coding and should remove client identifiers when possible. If the narratives contain privileged detail, limit access to the study team and follow the firm's confidentiality policy. The coding notes should not quote sensitive content unnecessarily.
When sharing results, aggregate at the matter type or practice group level. Do not include examples that could identify a client or reveal a legal strategy. Use redacted examples only after the supervising attorney approves them. The study is a process improvement tool, not a substitute for the client's own review. Use a redaction checklist before any example leaves the controlled environment.
Applying the result
If missing outcomes dominate, revise the narrative template and the coaching example rather than adding another general instruction. If block entries dominate, change the prebill check that allows a block to pass. If confidentiality concerns dominate, add a review step before the narrative is submitted. The fix should match the defect that the data actually shows.
Test the fix on a small sample in the next cycle and compare the defect rate with the baseline. A drop in one defect type with no increase in another is evidence that the change worked. If the rate does not move, examine whether the guideline or the client reviewer changed during the period. The result should guide the next test, not close the question.
Review cadence
Repeat the study quarterly while the narrative standard is being revised, then move to a semiannual cycle. Keep the coding guide and the guideline version history stable so the trend is comparable. Add new defect codes only when the guideline introduces a new requirement. Report the change in method beside the result. Write the baseline and the target beside the chosen fix so the next cycle has a comparison.
Review the findings with the billing manager and the timekeeper group that produced the sample. Ask which fix is practical in the current workflow and who will own it. Set a follow-up date before the next cycle. A measurement that does not lead to an owned action becomes another report that no one uses. Record the guideline version in force at the follow-up so a change in the source is visible.
Put the guidance to work
Choose one recurring billing workflow, identify its authoritative source records, and name the owner who reviews exceptions. We can help turn that boundary into a focused support role.
Discuss your workflow


