I85PREP
⌘ K
Google Advanced Data Analytics study materials
Google Advanced Data Analytics study resources

Google Advanced Data Analytics practice questions

One question at a time. Build your Google Advanced Data Analytics understanding.

First 50 of 190 questions. Each one lists the choices, the correct answer, and an explanation. The checker at the top of this page saves progress in this browser when it is available.

Continue in the practice tool above to load more questions, 50 at a time.

Question 1Data projects, communication and accountable business value

What does the acronym PACE stand for in Google's data science project framework?

  1. Plan, Analyze, Construct, Execute Correct answer.
  2. Prepare, Analyze, Communicate, Evaluate
  3. Predict, Assess, Compare, Explain
  4. Process, Audit, Conclude, Export

Explanation

PACE is Google's project workflow for data science: Plan defines the problem and scope, Analyze explores the data, Construct builds the model or output, and Execute communicates results to stakeholders. It is the recurring framework used across all seven courses of the Advanced Data Analytics Certificate.

Question 2Data projects, communication and accountable business value

Which statement best distinguishes a data scientist from a data analyst?

  1. Data scientists only work with unstructured data; analysts only work with structured data
  2. Data scientists build predictive models and use ML techniques; analysts focus on descriptive analysis to answer current business questions Correct answer.
  3. Data scientists do not need to communicate with stakeholders
  4. Data analysts always earn more than data scientists

Explanation

The keyed option describes a common emphasis in the supplied career comparison, not an exclusive universal boundary. Analysts can build predictive models and data scientists also perform descriptive analysis; actual responsibilities vary by organization. Focus on the task, methods and decision rather than rely on a title as proof of competence.

Question 3Data projects, communication and accountable business value

A data engineer's primary responsibility is to:

  1. Design machine learning models
  2. Build and maintain data pipelines and infrastructure that move data into usable systems Correct answer.
  3. Write executive summaries for stakeholders
  4. Run A/B tests on product features

Explanation

Data engineers build and maintain pipelines, warehouses, and infrastructure that deliver clean, reliable data to analysts and scientists. They typically work with tools like Airflow, Spark, and cloud data warehouses. Analysts and scientists then consume that data downstream.

Question 4Data projects, communication and accountable business value

In the PACE framework, which stage involves cleaning, exploring, and understanding the data?

  1. Plan
  2. Analyze Correct answer.
  3. Construct
  4. Execute

Explanation

Analyze is the stage where EDA, cleaning, and statistical exploration happen. Plan defines the problem and stakeholders, Construct builds the model, and Execute communicates findings. Confusing Analyze with Plan is a common mistake - Plan is about scoping the question, not exploring data.

Question 5Data projects, communication and accountable business value

Which of the following is a fairness concern in a data science project?

  1. A model that uses 10 features instead of 5
  2. A hiring model trained on historical data that systematically scores candidates from one demographic group lower despite similar qualifications Correct answer.
  3. A regression model with R² of 0.65
  4. A pipeline that updates daily instead of hourly

Explanation

Fairness refers to whether a model treats individuals or groups equitably. Historical bias in training data can produce models that disadvantage protected groups. Detecting and mitigating this bias is a core ethical responsibility in data science.

Question 6Data projects, communication and accountable business value

What is the primary purpose of the Execute stage in PACE?

  1. Run the Python code in production
  2. Communicate insights and recommendations to stakeholders and drive action Correct answer.
  3. Execute SQL queries against the data warehouse
  4. Delete the working data after the project ends

Explanation

Execute in PACE is about communication and action - sharing findings with stakeholders, writing executive summaries, and ensuring the analysis leads to decisions. Despite the name, it is not about running code; it is about delivering value from the analysis.

Question 7Data projects, communication and accountable business value

Which of these is the BEST example of a transparency practice in data science?

  1. Hiding the model's logic from stakeholders to protect intellectual property
  2. Documenting the data sources, assumptions, model choices, and limitations in a project report Correct answer.
  3. Reporting only positive results
  4. Encrypting all training data

Explanation

Transparency means stakeholders can understand how a model works, what data it uses, and what its limitations are. Clear documentation of data lineage, modeling choices, and known weaknesses is a core ethical requirement, especially when decisions affect people.

Question 8Data projects, communication and accountable business value

A stakeholder asks for 'a quick analysis' with no defined success criteria. According to PACE, what should you do first?

  1. Start writing Python code immediately
  2. Use Plan to clarify the question, success metric, and decision being made before doing any analysis Correct answer.
  3. Send them a Tableau dashboard
  4. Train a machine learning model on whatever data is available

Explanation

Plan is the first PACE stage and is critical when scope is unclear. Clarifying the question, the decision the stakeholder will make, and the success criteria prevents wasted work and ensures the analysis is relevant. Jumping into code or modeling without Plan often results in rework.

Question 9Data projects, communication and accountable business value

Which data ethics issue is MOST closely related to GDPR and similar regulations?

  1. Model accuracy
  2. Privacy and consent for personal data Correct answer.
  3. Visualization aesthetics
  4. Pipeline latency

Explanation

GDPR (EU), CCPA (California), and similar laws govern how personal data is collected, stored, used, and shared. Privacy and informed consent are central. Data scientists must understand what personal data they may use and under what restrictions.

Question 10Data projects, communication and accountable business value

Predictive analytics typically attempts to answer which type of question?

  1. What happened?
  2. Why did it happen?
  3. What will happen? Correct answer.
  4. What should we do about it?

Explanation

Descriptive analytics asks what happened. Diagnostic asks why. Predictive analytics forecasts what will happen using models trained on historical data. Prescriptive analytics recommends what to do. These four are the standard analytics maturity tiers.

Question 11Data projects, communication and accountable business value

Prescriptive analytics:

  1. Reports past performance metrics
  2. Recommends specific actions to achieve desired outcomes, often combining models and optimization Correct answer.
  3. Cleans and standardizes raw data
  4. Visualizes data in dashboards

Explanation

Prescriptive analytics is the most advanced tier - it recommends actions, often by combining predictive models with optimization or simulation. Example: a model that not only predicts demand but recommends inventory levels by store.

Question 12Data projects, communication and accountable business value

When presenting a model to a non-technical executive, you should:

  1. Focus on technical details such as the loss function and hyperparameters
  2. Lead with the business impact, the recommendation, and confidence, and keep technical depth for an appendix Correct answer.
  3. Skip explaining limitations to avoid undermining confidence
  4. Show only the raw model output without interpretation

Explanation

Executive communication should lead with the so-what - the business impact and recommended decision. Technical depth belongs in an appendix or technical review. Hiding limitations is unethical; stakeholders need to know when results are uncertain.

Question 13Data projects, communication and accountable business value

Sampling bias occurs when:

  1. The sample is too large
  2. The data used to train a model does not represent the population the model will be applied to Correct answer.
  3. The model has too many parameters
  4. The data is stored in multiple files

Explanation

Sampling bias arises when the training data systematically excludes or under-represents parts of the target population. A model trained on data from only one region or demographic will produce biased predictions when applied to others. Detecting and correcting sampling bias is a fairness concern.

Question 14Data projects, communication and accountable business value

Which workflow correctly orders the typical data science project?

  1. Build a model, then ask the stakeholder what they want
  2. Plan the question, gather and explore data, build the model, communicate results Correct answer.
  3. Communicate results, then collect data
  4. Train the model first, then clean the data

Explanation

A well-structured data science project starts with understanding the stakeholder question (Plan), then gathering and exploring data (Analyze), then modeling (Construct), then communicating findings (Execute). This is the PACE workflow.

Question 15Data projects, communication and accountable business value

Which is the MOST appropriate response when a stakeholder asks for a model prediction but your training data is six years old and the domain has changed significantly?

  1. Build the model anyway and report results without caveats
  2. Refuse to do any analysis
  3. Flag the data freshness issue to the stakeholder, discuss whether new data is available, and document the limitation in the final report Correct answer.
  4. Generate synthetic data to replace the old data

Explanation

Data freshness is a legitimate concern. Ethical practice is to raise the issue early, explore whether better data is available, and clearly document the limitation if you proceed. Refusing all work or generating fake data are both inappropriate.

Question 16Python and dataframe reasoning for applied analytics

In Pandas, what does the .iloc accessor use for indexing?

  1. Column names only
  2. Integer-based positional indexing Correct answer.
  3. Boolean masks only
  4. Regex patterns

Explanation

.iloc is integer-location based. df.iloc[0] gets the first row by position. .loc is label-based, so df.loc['a'] gets the row with index label 'a'. Mixing them up is one of the most common Pandas mistakes.

Question 17Python and dataframe reasoning for applied analytics

Which Pandas method is BEST for combining two DataFrames on a common key column similar to a SQL JOIN?

  1. pd.concat
  2. df.merge Correct answer.
  3. df.append
  4. df.stack

Explanation

df.merge (or pd.merge) performs SQL-style joins on key columns, supporting inner, outer, left, and right joins via the how parameter. pd.concat stacks DataFrames along an axis. df.append is deprecated. df.stack reshapes wide to long.

Question 18Python and dataframe reasoning for applied analytics

What does NumPy broadcasting do?

  1. Sends arrays across a network
  2. Allows arithmetic between arrays of different but compatible shapes by virtually replicating the smaller array Correct answer.
  3. Compresses arrays for storage
  4. Converts arrays to lists

Explanation

Broadcasting lets NumPy perform element-wise operations between arrays of different shapes by virtually expanding the smaller array to match the larger one - without actually copying data. For example, adding a 1D array of length 3 to a 2D array of shape (4, 3) works without an explicit loop.

Question 19Python and dataframe reasoning for applied analytics

Which line correctly reads a CSV file into a Pandas DataFrame?

  1. df = pd.load_csv('data.csv')
  2. df = pd.read_csv('data.csv') Correct answer.
  3. df = pd.csv('data.csv')
  4. df = pd.import('data.csv')

Explanation

pd.read_csv is the standard Pandas function for reading CSV files. Related functions include pd.read_excel, pd.read_json, and pd.read_sql. There is no pd.load_csv, pd.csv, or pd.import.

Question 20Python and dataframe reasoning for applied analytics

What does df.groupby('region')['sales'].sum() return?

  1. A single number - total sales across all regions
  2. A Series with regions as the index and total sales per region as values Correct answer.
  3. A DataFrame with one row per sale
  4. An error

Explanation

groupby splits the DataFrame by the values in 'region', then ['sales'].sum() aggregates the sales column for each group. The result is a Series indexed by region with the per-region sum.

Question 21Python and dataframe reasoning for applied analytics

To find rows where the 'age' column has missing values in Pandas, use:

  1. df[df['age'] == None]
  2. df[df['age'].isnull()] Correct answer.
  3. df[df['age'] = NaN]
  4. df.find_missing('age')

Explanation

.isnull() (or its alias .isna()) returns a boolean Series flagging missing values. Comparing to None or NaN with == does not work because NaN is not equal to anything, including itself. Use .isnull() or .notnull().

Question 22Python and dataframe reasoning for applied analytics

Which of these correctly drops duplicate rows from a DataFrame, keeping the FIRST occurrence?

  1. df.drop_duplicates(keep='first') Correct answer.
  2. df.drop_duplicates(keep='last')
  3. df.unique()
  4. df.distinct()

Explanation

drop_duplicates(keep='first') is the default behavior and keeps the first occurrence of each duplicate row. keep='last' keeps the last. keep=False drops all duplicates entirely. .unique() works on Series, not DataFrames; .distinct() is not a Pandas method.

Question 23Python and dataframe reasoning for applied analytics

What is vectorization in NumPy?

  1. Drawing graphics with vectors
  2. Applying operations across entire arrays at once using low-level optimized code, avoiding explicit Python loops Correct answer.
  3. Converting scalars into vectors
  4. Reducing dimensionality

Explanation

Vectorization replaces slow Python for-loops with array operations implemented in compiled C under the hood. For example, arr * 2 multiplies every element at once - typically 10-100x faster than a loop. Vectorization is the main reason NumPy is used over plain Python lists for numerical work.

Question 24Python and dataframe reasoning for applied analytics

A Pandas Series is BEST described as:

  1. A 2-dimensional table
  2. A 1-dimensional labeled array, similar to a single column of a DataFrame Correct answer.
  3. A dictionary that can only hold strings
  4. A type of database connection

Explanation

A Series is a 1D labeled array - values plus an index. A DataFrame is a 2D structure built from multiple Series sharing an index. Each column of a DataFrame is itself a Series.

Question 25Python and dataframe reasoning for applied analytics

Which method converts a column of strings like '$1,234.56' into floats after cleaning?

  1. df['col'].astype(int) directly
  2. df['col'].str.replace('[$,]', '', regex=True).astype(float) Correct answer.
  3. df['col'].sum()
  4. df['col'].head()

Explanation

Use .str.replace with regex=True to remove $ and , characters, then .astype(float) to convert. Calling .astype(int) directly on '$1,234.56' would raise a ValueError because the string contains non-numeric characters.

Question 26Python and dataframe reasoning for applied analytics

For a Pandas DataFrame with the ordered integer index[0,1,2,3,4], what is the difference between df.loc[0:3]and df.iloc[0:3]?

  1. They always return the same result
  2. df.loc[0:3] is inclusive on the right (rows with labels 0,1,2,3), df.iloc[0:3] is exclusive (positions 0,1,2) Correct answer.
  3. df.iloc is for columns and df.loc is for rows
  4. df.loc is slower

Explanation

With the explicitly supplied ordered labels, loc includes the ending label3, selecting labels0through3. iloc uses positional slicing with an excluded endpoint, selecting positions0through2. Other index labels/order can change label-based behavior; do not universally equate every integer label with its physical position.

Question 27Python and dataframe reasoning for applied analytics

Which Pandas method creates a pivot table summarizing average sales by region and product?

  1. df.pivot_table(values='sales', index='region', columns='product', aggfunc='mean') Correct answer.
  2. df.merge('sales')
  3. df.sort_values('sales')
  4. df.drop_duplicates()

Explanation

pivot_table reshapes long data into a wide summary. values is what to aggregate, index is the row grouping, columns is the column grouping, and aggfunc is the aggregation function (mean, sum, count, etc.).

Question 28Python and dataframe reasoning for applied analytics

A NumPy array's shape attribute returns:

  1. The memory size in bytes
  2. A tuple giving the size along each dimension Correct answer.
  3. The data type
  4. The total number of elements as a single integer

Explanation

shape returns a tuple - e.g., (3, 4) for a 2D array with 3 rows and 4 columns. The single integer for total elements is .size. The data type is .dtype. Memory size is .nbytes.

Question 29Python and dataframe reasoning for applied analytics

Which line correctly applies a function to every element of a Pandas Series?

  1. s.apply(my_function) Correct answer.
  2. s.execute(my_function)
  3. s.call(my_function)
  4. s.run(my_function)

Explanation

.apply runs a function element-wise on a Series. On a DataFrame, .apply works along axis=0 (columns) by default. For simple math, vectorized operations (s * 2, np.log(s)) are faster than .apply because they avoid the Python-level function call.

Question 30Python and dataframe reasoning for applied analytics

Why are Jupyter notebooks particularly useful in data science work?

  1. They compile Python to machine code for speed
  2. They mix executable code, narrative text, equations, and visualizations in one document, supporting reproducible analysis Correct answer.
  3. They can only run one cell at a time
  4. They replace the need for version control

Explanation

Jupyter notebooks let analysts interleave code, markdown explanations, equations, and inline plots. This supports literate programming and reproducible analyses. They do not replace version control - notebook files (.ipynb) should still be tracked in Git, ideally with output stripping to keep diffs clean.

Question 31Statistical inference, regression and model assumptions

EDA stands for:

  1. Extended Data Analytics
  2. Exploratory Data Analysis Correct answer.
  3. Effective Data Architecture
  4. Estimated Data Aggregation

Explanation

EDA - Exploratory Data Analysis - is the practice of summarizing and visualizing a dataset to understand its structure, distributions, relationships, and anomalies before modeling. EDA is part of the Analyze stage in PACE.

Question 32Statistical inference, regression and model assumptions

Which chart is BEST for showing the relationship between two continuous variables?

  1. Pie chart
  2. Scatter plot Correct answer.
  3. Bar chart
  4. Box plot

Explanation

A scatter plot displays paired values of two continuous variables and reveals correlation, clusters, and outliers. Pie charts are for parts of a whole. Bar charts compare categories. Box plots show distribution summaries.

Question 33Statistical inference, regression and model assumptions

A histogram is BEST suited for visualizing:

  1. A categorical variable's frequency
  2. The distribution of a single continuous variable Correct answer.
  3. The relationship between two variables
  4. A time series trend

Explanation

Histograms bin a continuous variable into ranges and show the frequency of values within each bin, revealing the shape of the distribution (skew, modality, outliers). For categorical frequency use a bar chart; for two-variable relationships use a scatter plot; for trends over time use a line chart.

Question 34Statistical inference, regression and model assumptions

Which chart type is MOST appropriate for displaying a correlation matrix across many numeric variables?

  1. Pie chart
  2. Heatmap Correct answer.
  3. Line chart
  4. Stacked bar chart

Explanation

A heatmap uses color intensity to encode values in a 2D grid, making it ideal for correlation matrices, confusion matrices, and other dense numeric tables. The eye quickly spots strong positive or negative cells.

Question 35Statistical inference, regression and model assumptions

A box plot directly shows all of the following EXCEPT:

  1. Median
  2. Interquartile range (Q1 to Q3)
  3. Potential outliers (points beyond the whiskers)
  4. Mean Correct answer.

Explanation

A standard box plot shows the median (the line in the box), Q1 and Q3 (box edges), the whiskers (typically 1.5 × IQR), and outliers beyond the whiskers. The mean is not shown by default - some implementations add it as a separate marker.

Question 36Statistical inference, regression and model assumptions

Which is the BEST first step in EDA when you load a new dataset?

  1. Train a machine learning model immediately
  2. Inspect shape, dtypes, head/tail, summary statistics, and counts of missing values Correct answer.
  3. Drop all rows with any missing value
  4. Publish a Tableau dashboard

Explanation

EDA begins with structural inspection: df.shape, df.dtypes, df.head(), df.describe(), df.isnull().sum(). Understanding what is in the data - types, scales, missingness - must come before transformations, modeling, or visualization.

Question 37Statistical inference, regression and model assumptions

Which design choice would MOST mislead a viewer of a bar chart?

  1. Sorting bars from largest to smallest
  2. Truncating the y-axis to start at a non-zero value, exaggerating small differences Correct answer.
  3. Labeling each axis
  4. Using a colorblind-friendly palette

Explanation

Truncating the y-axis to start above zero on a bar chart visually inflates small differences between categories - a classic deceptive visualization. Bar charts should generally start at zero. Line charts may use truncated axes when zero is uninformative, but they should be labeled clearly.

Question 38Statistical inference, regression and model assumptions

An outlier in EDA is best described as:

  1. Any duplicate row
  2. A data point that lies far from the bulk of the distribution and may indicate error, rare event, or genuine extreme Correct answer.
  3. The value in the middle of a sorted list
  4. A null value

Explanation

Outliers are values that deviate substantially from other observations. They may be data errors, rare but real events, or evidence of a heavy-tailed distribution. The correct response depends on the cause - never automatically drop outliers without investigating.

Question 39Statistical inference, regression and model assumptions

Which is a recommended principle of clear data visualization?

  1. Use 3D effects to add visual interest
  2. Use chartjunk and decoration to engage viewers
  3. Maximize the data-ink ratio by removing unnecessary visual elements Correct answer.
  4. Show as many variables as possible on one chart

Explanation

Edward Tufte's principle is to maximize the data-ink ratio - devote the maximum proportion of ink to data, minimize chartjunk (decorative non-informational elements). 3D effects distort perception. Cluttering with too many variables defeats the purpose.

Question 40Statistical inference, regression and model assumptions

When choosing between a line chart and a bar chart for time-series data, line charts are preferred when:

  1. You have only two time points
  2. You want to show a continuous trend over many time points Correct answer.
  3. You want to show parts of a whole
  4. The data is categorical

Explanation

Line charts excel at showing trends across many time points because the connecting line emphasizes continuity. Bar charts work for a small number of discrete time periods or when emphasizing exact values at specific times.

Question 41Statistical inference, regression and model assumptions

Which Python library is MOST commonly used for statistical visualizations like violin plots and pair plots?

  1. NumPy
  2. Pandas
  3. seaborn Correct answer.
  4. requests

Explanation

seaborn is built on top of matplotlib and provides higher-level statistical plots - violin plots, pair plots, distribution plots, and heatmaps - with sensible defaults. matplotlib is more general-purpose. NumPy and Pandas focus on numerical and tabular data, respectively.

Question 42Statistical inference, regression and model assumptions

A treemap is MOST appropriate for showing:

  1. A time series trend
  2. Hierarchical part-to-whole relationships using nested rectangles sized by value Correct answer.
  3. The correlation between two variables
  4. The distribution of a single variable

Explanation

Treemaps represent hierarchical data through nested rectangles, with each rectangle's area proportional to a value. They are good for showing how a total breaks down across categories and subcategories. They are not ideal for precise comparisons or trends.

Question 43Statistical inference, regression and model assumptions

For a usual Seaborn violin plot configured with its inner box summary, which listed description identifies the displayed combination?

  1. It only shows the median
  2. It combines a box plot summary with a kernel density estimate of the distribution shape Correct answer.
  3. It is the same thing as a bar chart
  4. It only works for categorical data

Explanation

Violin plots show the distribution shape via a mirrored kernel density estimate and typically embed a box plot or median marker. They reveal multimodality and skew that a box plot can hide - useful when comparing distributions across groups.

Question 44Statistical inference, regression and model assumptions

When using color in data visualization, which choice is MOST inclusive for colorblind viewers?

  1. Red and green for comparison
  2. A perceptually uniform palette such as viridis, with redundant encoding via shape or labels Correct answer.
  3. Pure red and pure green with no labels
  4. Rainbow palette (jet)

Explanation

Red-green palettes confuse the most common form of color blindness (deuteranopia). Perceptually uniform palettes like viridis are colorblind-safe and print-safe. Adding shape, pattern, or text labels provides redundancy. The rainbow palette is widely discouraged because it is not perceptually uniform.

Question 45Statistical inference, regression and model assumptions

An EDA reveals that one numeric column has 60% missing values and the missingness appears correlated with the target variable. The MOST appropriate next step is:

  1. Silently drop the column
  2. Investigate whether the missingness itself is informative; consider creating a missingness indicator and discussing the implications with stakeholders Correct answer.
  3. Fill all missing values with zero without comment
  4. Train the model anyway and ignore the issue

Explanation

When missingness correlates with the target, the fact of missingness may itself carry information (missing not at random, MNAR). Blind imputation can introduce bias. A missingness indicator column plus stakeholder discussion is more defensible than silent imputation or deletion.

Question 46Statistical inference, regression and model assumptions

Which measure of central tendency is MOST affected by extreme outliers?

  1. Median
  2. Mode
  3. Mean Correct answer.
  4. Interquartile range

Explanation

The mean is computed by summing all values, so even a single extreme outlier can shift it substantially. The median is the middle value of the sorted data and is robust to outliers. The mode is the most frequent value. The IQR is a measure of spread, not central tendency.

Question 47Statistical inference, regression and model assumptions

Standard deviation measures:

  1. The middle value of a dataset
  2. The typical distance of observations from the mean Correct answer.
  3. The most frequent value
  4. The total range of the data

Explanation

Standard deviation is the square root of variance - it summarizes how far observations typically fall from the mean, in the same units as the data. Higher standard deviation means more spread; lower means observations cluster close to the mean.

Question 48Statistical inference, regression and model assumptions

For independent identically distributed observations with finite nonzero variance, what is the usual central-limit-theorem conclusion about the properly standardized sample mean as sample size grows?

  1. All distributions are normal
  2. The sampling distribution of the sample mean approaches a normal distribution as sample size grows, regardless of the population distribution Correct answer.
  3. Larger samples always have lower variance than smaller samples
  4. A sample of 30 is sufficient for any analysis

Explanation

Under the stated independent identically distributed finite-variance conditions, the standardized sample mean converges in distribution to a normal law. This is not an unconditional claim for every population or dependence pattern, and a universal sample-size cutoff cannot be inferred. Actual sampling and tail behavior matter to a finite-sample approximation.

Question 49Statistical inference, regression and model assumptions

Which distribution models the number of successes in a fixed number of independent yes/no trials with equal success probability?

  1. Poisson
  2. Binomial Correct answer.
  3. Normal
  4. Exponential

Explanation

The Binomial distribution counts successes in n independent Bernoulli trials, each with success probability p. The Poisson models the count of events in a fixed interval. The Normal is continuous and symmetric. The Exponential models waiting times between events.

Question 50Statistical inference, regression and model assumptions

Which probability distribution is MOST commonly used to model the number of rare events occurring in a fixed time interval?

  1. Binomial
  2. Poisson Correct answer.
  3. Normal
  4. Uniform

Explanation

The Poisson distribution models the count of independent events over a fixed interval when events are rare and the rate is constant - e.g., customer arrivals per minute, defects per square meter. It is characterized by a single parameter, λ (the mean rate).