| Course | BU 351 Introduction to Business Analytics |
|---|---|
| Unit | Unit 2 |
| Paper type | Data quality assessment |
| Length | About 960 words, 6 pages |
| Format | APA 7 student paper |
| School | Herzing University |
| Program | BS in Business Analytics |
| Updated | September 2026 |
Free sample paper for BU 351 Unit 2
Thirty-Eight Duplicates, Fourteen Gym Codes and Five Members Who Joined After They Quit: A Data Quality Assessment of a Climbing Gym Membership File
Student Name
Business, Herzing University
BU 351: Introduction to Business Analytics
Instructor Name
February 4, 2026
Thirty-Eight Duplicates, Fourteen Gym Codes and Five Members Who Joined After They Quit: A Data Quality Assessment of a Climbing Gym Membership File
Purpose and Method
Before any question from the Unit 1 framing can be answered, Basecamp Climbing's membership file has to be checked; the chain is the invented seven-gym business this course follows. This assessment profiles the file field by field, counts what is missing, duplicated, inconsistently coded or out of range, and reports each finding with what it would do to the analysis downstream. The file is a comma-separated export of 4,238 rows and ten fields covering members who joined from January 2023 through November 2025. The profile was produced in Python by counting filled values, distinct values and ranges for each field. Wang and Strong (1996) grouped data quality into intrinsic qualities, such as accuracy, contextual qualities, such as completeness for the task, representational qualities, such as consistent coding, and accessibility, and the findings below touch the first three.
Each finding below follows the same pattern: a count, an example of the problem, a judgment about its likely cause and a statement of which later stage of the analysis it would damage if left alone. That last element matters most. A data quality assessment is not a list of imperfections; it is a judgment about which imperfections would change an answer. A field that is untidy but never used can be left as it is, while a small problem in a field that drives a model can mislead every result built on it.
The Profile
The table shows one row per field. Distinct values are counted among filled records.
| Field | Type | Filled | Distinct | Range or values |
|---|---|---|---|---|
| member_id | Identifier | 4,238 | 4,200 | 100001 to 104200 |
| gym | Categorical | 4,238 | 14 | Seven three-letter codes and seven full names |
| plan | Categorical | 4,238 | 3 | monthly, student, family |
| join_date | Date | 4,238 | 1,042 | January 1, 2023 to November 30, 2025 |
| cancel_date | Date | 2,636 | 946 | Blank means still active |
| age_band | Categorical | 3,988 | 4 | 18-24, 25-34, 35-44, 45+ |
| visits_last_90 | Count | 4,238 | 34 | minus 5 to 31 |
| autopay_failures_12m | Count | 4,238 | 5 | 0 to 4 |
| intro_class | Binary | 4,238 | 2 | 0, 1 |
| referred | Binary | 4,238 | 2 | 0, 1 |
Finding 1: Duplicate Members
The file holds 4,238 rows but only 4,200 distinct member identifiers, so 38 rows are duplicates. Of those, 31 are exact copies of another row. The other seven share an identifier with a second row but differ in one field, such as a blank age band in one copy or a gym written as a full name in the other. The likely cause is an export that appended a partial second batch. Left in place, the duplicates would overcount members and cancellations by about 1 percent and give duplicated members double weight in any model. The fix is to keep one row per identifier, preferring the copy with more complete values. Redman (2013) argued that data errors of this kind usually surface only after a decision has been made on them, which is why the check belongs before analysis rather than after.
Finding 2: Inconsistent Gym Codes
The gym field should contain seven codes but contains 14 distinct values: the seven three-letter codes, such as BLD and AUR, and 61 rows where the gym is written as a full name, such as Boulder or Aurora. Every full name maps unambiguously to one code. The cause is probably staff entering members by hand at the front desk during a system outage. Left uncorrected, any count by gym would split each location into two groups and understate every gym's figures. The fix is a lookup that maps each name to its code.
Finding 3: Missing Age Bands
Age band is blank in 250 rows, about 5.9 percent. Nothing else in those rows looks unusual, which suggests the field was optional on the sign-up form. For the questions framed in Unit 1, age band is not required, so the rows should be kept and age band treated as unknown where it is missing. Dropping those rows would discard members for no benefit and could bias results if members who skip optional fields differ from those who complete them.
If a later question did depend on age, the missing values would need a second look. Members who skip optional fields may be younger, busier or less engaged, and removing them could shift any comparison by age. For now the rows stay in, with age marked unknown.
Finding 4: Impossible Values
Three rows report negative visits in the last 90 days, one as low as minus 5. Visits cannot be negative, so these are entry or calculation errors. Because only three rows are affected, the value should be set to missing rather than guessed. Separately, five rows show a join date later than the cancellation date, which is impossible. These five cannot be repaired from the file alone, since either date could be wrong, and should be removed from analyses that depend on tenure while the gym manager checks them in the billing system.
Finding 5: A Field Measured at the Wrong Time
One problem does not appear as an error in the profile. The visits field records visits in the 90 days before each member's last date: the cancellation date for members who left and December 31, 2025 for those still active. It is therefore not measured at a common point in time. Canceled members averaged 6.0 visits and active members 13.7, but that comparison mixes the months before leaving with a recent quarter for current members. Used as an input to a model predicting cancellation, the field would leak information about the outcome, since low visits partly reflect a member who has already decided to leave. Provost and Fawcett (2013) warn that this kind of leakage makes a model look accurate in testing and fail in use. The field should be excluded from the predictive exercise.
The Cleaned File
After removing the 38 duplicate rows, mapping gym names to codes, setting three negative visit counts to missing and setting aside the five impossible date records, the analysis file holds 4,195 members. Every later unit uses this cleaned file. The changes and their counts are recorded here so that anyone repeating the analysis can reproduce it. The file, the chain and every count here are fictional teaching material (Wang & Strong, 1996).
References
Provost, F., & Fawcett, T. (2013). Data science for business: What you need to know about data mining and data-analytic thinking. O'Reilly Media.
Redman, T. C. (2013). Data's credibility problem. Harvard Business Review, 91(12), 84-88.
Wang, R. Y., & Strong, D. M. (1996). Beyond accuracy: What data quality means to data consumers. Journal of Management Information Systems, 12(4), 5-33. https://doi.org/10.1080/07421222.1996.11518099
BU 351 Unit 2 instructions, in plain terms
The Unit 2 assignment asks students to assess the quality of a supplied dataset before analysis. Most versions want a field-by-field profile showing type, completeness, distinct values and ranges, followed by findings on missing values, duplicates, inconsistent coding and impossible values. For each finding, rubrics usually ask for a count, an example, a likely cause and a statement of how it would affect later analysis. Many instructors also want recommended fixes and a description of the cleaned dataset. The profile can be built in Excel, Python or another tool if the method is described. Expect a table and a few pages of findings, citing data quality sources in APA 7.
How this BU 351 Unit 2 example is built
The sample states its purpose and method and ties them to published dimensions of data quality. A ten-row profile table follows, one row per field. Five findings are then written up in turn, each with a count and an example: duplicates, inconsistent gym codes, missing age bands, impossible visit and date values, and a timing problem in the visits field that the profile alone would not reveal. Each finding explains its likely cause and which later stage it would harm, and proposes a fix proportionate to the problem. The final section lists every change made and gives the size of the cleaned file, so the analysis can be reproduced.
Reading the BU 351 Unit 2 rubric
The rubric for data quality typically assesses the completeness of the profile, the identification of problems, the analysis of their impact and the proposed fixes. The profile earns full points when every field is described with type, fill count, distinct values and range. Identification is judged by specific counts and examples rather than general statements that the data are messy. Impact analysis carries heavy weight and rewards findings tied to the stage they would damage, such as double-counting or model leakage. Fixes score highest when they are proportionate, for example keeping rows with a missing optional field. Instructors often reward spotting a problem the profile cannot show, like the timing issue in the sample.
BU 351 Unit 2 help before you submit
A frequent Unit 2 mistake is deleting every row with a missing value, which can throw away useful records and bias results. Another is reporting problems without counts or examples. Students often fix issues without recording what they changed, making the analysis impossible to reproduce. Some profiles list fields but skip ranges and distinct counts, which is where many errors show up. Duplicates are commonly missed when copies differ slightly. Finally, think about timing; a field measured at different moments for different records can look perfectly clean in a profile and still ruin a predictive model built on it. Check duplicates on the identifier field, not only on whole rows, since near-copies hide there.
Get a BU 351 Unit 2 example written to your instructions
Post us the Unit 2 instructions, the BU 351 rubric your classroom carries, and the dataset the assignment supplies. We write a custom example that profiles every field, reports findings with counts and example records, separates patterned gaps from random ones and documents each treatment decision. First custom sample free, returned in 24 to 48 hours.
More BU 351 papers and business analytics samples
- BU 351 Unit 1: Analytics types and framing
- BU 351 Unit 3: Descriptive analysis
- BU 351 Unit 4: Exploratory analysis
- BU 351 Unit 5: Predictive exercise
- BU 351 Unit 6: Prescriptive or optimization exercise
- BU 351 Unit 7: Reporting to a business audience
- BU 351 Unit 8: End-to-end analytics case
- BU 451 Unit 2: Data import and inspection script
- BU 453 Unit 2: Entity relationship diagram
- BU 454 Unit 2: Chart critique
- BU 455 Unit 2: Conventional alternative analysis
BU 351 Unit 2 questions, answered
What if the supplied dataset is already clean?
Say so and show the evidence, since a clean file is a real result and not a dead end. Report the profile, note that coverage is complete and the coded fields hold no strays, then look harder at the things a clean file still hides: values that are present but implausible, categories that mean the same thing under two labels, and dates that fall outside the period the file claims to cover.
Should I remove records with missing values?
Only after saying what removing them costs. Dropping incomplete records is easy and quietly changes who is left in the file, which matters if the gaps concentrate in one group. Where the instructions leave the choice open, show the count you would lose, say whether the remainder still represents the population, and record the decision either way rather than doing it silently.
Which tool should the profile be built in?
Whatever your section requires, and where nothing is specified a spreadsheet handles a file of this size comfortably. Pivot counts, a distinct value list per field and a few conditional checks produce everything the profile needs. If you use a statistical package, paste the output into the document rather than referring to it, since the marker is reading a report and not opening your working file.
Where can I find a free BU 351 Unit 2 sample paper?
The complete BU 351 Unit 2 assessment, profile table and all five findings, is published above for free. For an assessment of a different dataset, share the file description and the prompt, and a free custom assessment will be written, commonly inside 48 hours.
Should I remove records with missing values in BU 351 Unit 2?
Only when the missing field is needed for the question. The sample keeps 250 records with missing age bands because no question depends on age, and dropping them would discard members for no benefit. Explain your choice for each field.