A negative quantity, a price 100 times higher than usual, an order dated in 1900: unusual values are often typos, sometimes fraud, and occasionally the most important thing in your data. Either way, you want to see them before they distort an average.
Plumb is built on a simple idea: the data your business runs on should be trustworthy before anyone draws conclusions from it. Cleaning, checking and reshaping data should not need macros, scripts or a weekend. Plumb keeps your files private, shows every change it proposes, and lets you approve fixes instead of hunting for problems by hand.
How Plumb helps: Anomaly detection
Plumb's anomaly detection looks at each column in context. It flags statistical outliers, impossible values (negative ages, future birth dates), and rows that are unusual as a combination, like a small customer with a very large order, and explains why each was flagged.
How It Works
Use simple rules for impossible values and statistics for unusual ones.
- Sort each numeric column ascending and descending. Extreme values at either end are your first suspects.
- Highlight outliers with Conditional Formatting > Top/Bottom Rules or Color Scales.
- Z-score method: =ABS((A2-AVERAGE(A:A))/STDEV(A:A))>3 flags values more than 3 standard deviations from the mean.
- IQR method: values below Q1 - 1.5*IQR or above Q3 + 1.5*IQR, using =QUARTILE(A:A,1) and =QUARTILE(A:A,3). It works better on skewed data.
- Add rule checks for impossible values, such as negative quantities or dates outside a valid range.
- In Plumb: upload the dataset and review flagged rows with a reason for each, then mark them as errors or valid.
Where This Helps
Expense reports
Spot claims that are far above an employee's normal pattern.
Sensor or ops data
Catch spikes and drops that point to faulty readings.
Pricing
Find products priced far outside their category's range.
Try Plumb in early access
Plumb is in early access. Join the waitlist on tryplumb.pro to get an invite. Early access covers Excel (.xlsx), CSV and JSON files, and we are adding more sources as people ask for them. Your files are processed privately and are never used to train models.
See the odd ones out
Now you have a repeatable way to do it, by hand or with Plumb. Join the waitlist and tell us what your data looks like; we read every reply.
Join the waitlist