| Course | MT438 Analytics in the Digital Supply Chain |
|---|---|
| Unit | Unit 2 |
| Paper type | Data quality audit |
| Length | About 1,050 words, 6 pages double-spaced |
| Format | APA 7 student paper |
| School | Purdue University Global |
| Program | BS in Business Administration |
| Updated | October 2026 |
Free sample paper for MT438 Unit 2
Two Errors Hiding Each Other: An MT438 Unit 2 Data Quality Audit of 21,684 Order Lines
[Student Name]
Purdue University Global
MT438: Analytics in the Digital Supply Chain
Unit 2 Data Quality Audit
[Instructor Name]
[Date]
Oakhurst Lawn and Garden, its plants and the order extract are composite; the sources cited are real.
The Extract
The extract comes from Oakhurst's planning system and covers every order line shipped from February 2 through May 29, 2026: 21,684 lines across both plants, Decatur and Macon. Eight fields were examined: order number, line number, item number, ordered quantity, shipped quantity, order date, requested delivery date and ship timestamp. The headline figures the extract produces as delivered are a unit fill rate of 99.5 percent, 2,400,000 bags shipped against 2,412,060 ordered, and an on-time rate that cannot be computed at all without deciding what to do with blank timestamps.
Tests Anyone Could Repeat
Each defect below was found by a rule another analyst could run on the same file. Duplicates are lines matching on order number, line number, item and quantity. Unit-of-measure errors are ordered quantities under ten on items that are sold only by the bag. Missing timestamps are blank ship-time fields. Impossible sequences are lines whose ship timestamp precedes their order date. Split items are product descriptions that appear under two item numbers. Default dates are requested delivery dates exactly seven days after the order date. Wang and Strong (1996) argue that data quality has to be judged by fitness for the use a consumer makes of the data, so each test below is reported with the metric it affects rather than as a bare count.
Defect 1: Pallets Recorded as Bags
On 1,146 lines, 5.3 percent of the file, the ordered quantity is in pallets while the shipped quantity is in bags, at fifty bags a pallet. A home-improvement buyer who orders 3 pallets appears to have ordered 3 bags and received 150. Across those lines, 3,668 pallets are recorded as 3,668 bags, understating demand by 179,732 bags. Correcting this defect alone drops unit fill from 99.5 to 92.6 percent.
Defect 2: Exact Duplicates
Electronic order resends left 388 exact duplicate lines, 1.8 percent of the file, carrying 42,860 bags of ordered quantity that was never really ordered twice. Removing duplicates alone pushes unit fill to 101.3 percent, a figure no shipment record can produce.
Removing the duplicates alone yields a fill rate above 100 percent, which is how the pallet error gives itself away.
Why the Two Must Be Fixed Together
These two defects offset each other in the reported figure. The pallet error deflates the ordered quantity, so fill rate looks too high; the duplicates inflate it, so fill rate looks lower than it should. Corrected together, the extract shows 2,400,000 bags shipped against 2,548,932 truly ordered, a unit fill rate of 94.2 percent, more than five points under the figure in the report. An analyst who fixed only the duplicates would have produced an impossible number; one who fixed only the pallets would have reported a fill rate 1.6 points too low.
Defect 3: Blank Ship Timestamps
Ship timestamps are missing on 1,927 lines. About 83 percent of them, roughly 1,600, come from the Macon plant, and they begin on March 16, the day Macon replaced its dock scanners. The blanks are not random. Macon's recorded lines run late more often than Decatur's: 83.5 percent on time at Macon against 94.4 percent at Decatur. Dropping the blank lines would therefore remove mostly Macon lines and flatter the on-time rate.
A Range Instead of a Guess
After removing duplicates, 21,296 lines remain, 19,369 of them with a timestamp, and 17,940 of those shipped on time. Following the bounds approach Little and Rubin (2019) describe for data that may be missing for a reason, the audit reports a range rather than one figure. If every blank shipped late, on-time delivery is 84.2 percent. If every blank shipped on time, it is 93.3 percent. If the blanks behave like Macon's recorded lines, the best estimate is about 91.8 percent. The audit recommends reporting the range until Macon's scanner fault is fixed, and assigns that fix to Macon's shipping supervisor.
Defect 4: Shipped Before Ordered
Sixty-four lines carry a ship timestamp earlier than the order date. All are standing orders that a customer service clerk re-entered after the truck had left, so the order date is the date of keying rather than the date of the request. The audit sets these lines' order dates to the earliest related call in the case log, which corrects 59 of them; the other five are excluded and listed by order number.
Defect 5: Seventeen Products, Two Histories
Seventeen products appear under two item numbers since the packaging redesign in January, when the bags changed but the product did not. Each half of each product's history looks like a product in decline, and any forecast built on either half would plan too little. The audit maps each old number to its new one and recommends that product management own the cross-reference table.
Defect 6: The Plus-Seven Default
On 6,830 lines, 31.5 percent of the file, the requested delivery date is exactly the order date plus seven days. That is the planning system's default when a clerk leaves the field blank. Any on-time rate built on those dates is scoring the software's assumption rather than a customer's wish. Until customers' real requests are captured, the audit recommends measuring on-time delivery against the date Oakhurst promised in its order confirmation.
Cleaning Rules and Owners
Each defect has a rule and an owner. Duplicates are deleted on the four-field match, owned by the EDI team. Pallet lines are multiplied by fifty, owned by the order entry supervisor, who should also lock the unit field on bag-only items. Blank timestamps are reported as a range, owned by Macon shipping. Out-of-sequence lines are corrected from the case log, owned by customer service. Split items are mapped through a cross-reference table, owned by product management. Default request dates are flagged and excluded from requested-date measures, owned by sales operations. Hazen et al. (2014) note that most supply chain data errors are created upstream by people and processes, so a cleaning rule without an owner simply repeats each quarter.
References
Hazen, B. T., Boone, C. A., Ezell, J. D., & Jones-Farmer, L. A. (2014). Data quality for data science, predictive analytics, and big data in supply chain management: An introduction to the problem and suggestions for research and applications. International Journal of Production Economics, 154, 72-80. https://doi.org/10.1016/j.ijpe.2014.04.018
Little, R. J. A., & Rubin, D. B. (2019). Statistical analysis with missing data (3rd ed.). Wiley.
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
What the MT438 Unit 2 assignment asks for
MT438 Unit 2 hands students an order extract, or asks them to obtain one, and expects a written audit of its quality. Describe the extract first: source, period, row count and the fields you examined. Then state each test as a rule another analyst could rerun on the same file. For every defect you find, count the rows affected and name the metric it distorts, with the figure before and after correction. Missing values need a deliberate treatment, because deleting them assumes the missing rows resemble the rest. A cleaning rule should follow each finding, and stronger audits name who owns the fix so the problem stops recurring. Order the findings by how much they move the metrics rather than by how many rows they touch. The course text's material on data quality dimensions should appear where it explains a choice, not as a standalone definition.
How the MT 438 Unit 2 example is put together
Oakhurst's audit starts by describing its extract, February 2 to May 29, both plants and eight fields, and notes that on-time delivery cannot even be computed until blanks are handled. The six tests are listed next as plain rules. The pallet error and the duplicates come first because together they move fill rate by more than five points; a highlighted line explains how the impossible 101.3 percent exposes the pallet problem. A short section shows why the two must be fixed in one pass. Blank timestamps follow, traced to Macon's March 16 scanner change, with a range from 84.2 to 93.3 and a best estimate near 91.8. Lines shipped before they were ordered, seventeen renumbered products and the plus-seven default each get a short section. A closing table of rules assigns every fix to an owner, citing Hazen and colleagues on errors created upstream.
MT438 Unit 2 rubric: what earns full points
Audits in this course earn their marks by tying each defect to a consequence. A list of problems with row counts and nothing else reads as housekeeping; showing that two errors hide each other across five points of fill rate reads as analysis. Tests need to be precise enough to repeat, so a rule such as matching on order, line, item and quantity is credited over a phrase like checking for duplicates. Graders look closely at how missing data is treated. Deleting blank rows without asking why they are blank tends to cost points, while reporting a range with a reasoned best estimate shows judgment. Ordering findings by impact signals priority. A cleaning rule and an owner for each defect show the audit is meant to change practice. Supporting sources should explain a decision in the audit, cited in APA format at Purdue Global.
MT438 Unit 2 help from the desk
The commonest weak Unit 2 audit deletes every incomplete or odd row and reports the cleaned figure without saying what was removed. A second frequent problem lists defects in the order they were found, so a trivial formatting issue sits above an error that moves a headline metric. Many students fix one error and stop, never checking whether a second error was masking the first. Units of measure are easy to miss; look at quantities that are implausible for how the item is sold. Default values hide in plain sight, so check whether a date field sits a fixed number of days after another. Some audits never state the before and after figures, leaving the grader to guess how much each defect matters. Before submitting, recompute your main metric after each correction separately and then all together. If any version exceeds what is physically possible, you have found a hidden defect.
Get an MT438 Unit 2 example written to your instructions
Forward the order extract or data description attached to your Unit 2 materials, the assignment brief and the rubric. The composite audit written from it states each test, counts affected rows, traces every defect to the metric it distorts, and handles missing values with bounds. Your opening custom sample is free of charge, within 24-48h. The paper above is an original model document written by our desk, not a submitted student paper and not an official Purdue University Global document.
More MT438 papers and BS in Business Administration samples
- MT438 Unit 1: Supply Chain Data Inventory Post
- MT438 Unit 3: Performance Metrics Exercise
- MT438 Unit 4: Dashboard Critique
- MT438 Unit 5: Forecast Accuracy Analysis
- MT438 Unit 6: Seminar Reflection
- MT438 Unit 7: Visibility Technology Brief
- MT438 Unit 8: Root Cause Analysis
- MT438 Unit 9: Predictive Model Evaluation
- MT438 Unit 10: Analytics Recommendation Memo
- MT437 Unit 2: Warehouse Flow Map
- MT436 Unit 2: Spend Analysis
- MT435 Unit 2: Process Flow Analysis
- MT434 Unit 2: Freight Mode Comparison
MT438 Unit 2 questions, answered
Should I delete bad records or fix them?
It depends on the defect. Exact duplicates can usually be removed. Unit errors can often be corrected with a conversion factor. Missing values are trickier, because deleting them assumes they resemble the rest, which is rarely true. Whatever you choose, state the rule, count the rows affected and show the metric before and after, so the grader can judge the choice.
How do I find unit-of-measure errors?
Look for quantities that are implausible for how an item is sold, such as a 3 on a product always ordered in hundreds, or ratios between ordered and shipped quantities that cluster at a pack size like 12 or 50. Checking the item master for its selling unit helps. Once found, a single conversion rule usually corrects most of them.
What if my extract has no obvious errors?
Look harder at fields that could be defaults: dates that sit a fixed number of days after another date, codes that repeat far too often, or quantities that are round numbers in suspicious volume. Also check whether totals reconcile to a known figure. A clean extract is possible, but saying which tests you ran and what they found is still the substance of the audit.
Where can I find a free MT438 Unit 2 sample paper?
The full MT438 Unit 2 audit is printed on this page, testing 21,684 order lines for six defects and recomputing fill rate and on-time delivery after each correction.
What on-time figure does the MT438 Unit 2 sample report?
A range of 84.2 to 93.3 percent, depending on how 1,927 blank ship timestamps are treated, with about 91.8 percent if the blanks behave like Macon's recorded lines.