Top 30 Easy Analytics & Metrics Interview Questions and Answers for Freshers

30 questions. Pick an answer, or open “Show the answer” to read it.

Answers are graded in your browser. Nothing is saved, and no XP or streak is earned here. The app keeps score.

1. Question 1 of 30

   What should a well-structured metric hierarchy include when tracking engagement for a new feature?

   • aOne north star metric, two to three supporting KPIs, and at least one guardrail metric
   • bA comprehensive list of every possible user action to ensure complete data coverage
   • cVanity metrics such as total page views and total clicks to show overall activity
   • dOnly a north star metric to keep the plan simple and avoid metric overload
   Show the answer

   Answer: a · One north star metric, two to three supporting KPIs, and at least one guardrail metric

   A strong measurement plan uses a tight hierarchy with one north star, supporting KPIs, and guardrails to detect unintended harm. Option B is wrong because tracking every possible action creates instrumentation sprawl without prioritization, and Option C is wrong because vanity metrics do not tie directly to the feature's success.

   Read the full bite: How do you translate increase user engagement into a technical measurement plan?

2. Question 2 of 30

   A product manager asks you to help "increase engagement" with a new feature. What is the most effective first step to take?

   • aSelect an analytics tool and draft a technical spec for instrumenting user interaction events.
   • bPropose A/B testing several different UI designs to see which one users prefer.
   • cAsk clarifying questions to define what specific user action constitutes 'engagement' for this feature.
   • dBegin tracking generic metrics like Daily Active Users (DAU) and session duration to establish a baseline.
   Show the answer

   Answer: c · Ask clarifying questions to define what specific user action constitutes 'engagement' for this feature.

   The first step is always to collaborate with the PM to clarify the vague business goal into a specific, measurable user action. Jumping to generic metrics like DAU is a common mistake as it doesn't provide specific insight into the new feature's performance.

   Read the full bite: Translate 'increase engagement' into a technical measurement plan

3. Question 3 of 30

   What is the most crucial first step when translating a vague business goal like "increase engagement" into a technical measurement plan?

   • aIdentify a single, high-level metric, such as Daily Active Users (DAU), to track progress.
   • bDesign the database schema to capture all possible user interactions.
   • cDefine "engagement" with the Product Manager, specifying frequency, depth, or breadth.
   • dSelect an analytics platform and begin instrumenting basic click events.
   Show the answer

   Answer: c · Define "engagement" with the Product Manager, specifying frequency, depth, or breadth.

   The card emphasizes that the first and most crucial step is to collaborate with the Product Manager to define what 'engagement' specifically means for the feature. Without this clarification, any technical measurement plan would be based on assumptions and might not align with the actual business goal. Identifying a high-level metric (A) is part of the next step, but only after 'engagement' itself is clearly defined.

   Read the full bite: Translate 'increase engagement' into a technical measurement plan

4. Question 4 of 30

   For a new data ingestion pipeline into a cloud data warehouse, which scenario is the strongest argument for choosing an ELT approach?

   • aRaw, high-volume data needs to be retained for future, unpredictable analysis.
   • bThe target system has limited compute power for complex transformations.
   • cData contains sensitive information that must be anonymized before storage.
   • dData must conform to a rigid, pre-defined schema before loading.
   Show the answer

   Answer: a · Raw, high-volume data needs to be retained for future, unpredictable analysis.

   ELT is ideal for loading large volumes of raw data directly into a scalable cloud data warehouse, enabling flexible in-place transformations for diverse and evolving analytical needs. Masking sensitive data or conforming to rigid schemas before loading are typical use cases for ETL, which transforms data on a separate server.

   Read the full bite: ETL vs. ELT: Key differences and when to use each?

5. Question 5 of 30

   When is ELT generally the preferred approach over ETL?

   • aWhen the pipeline goal is to reduce costs by avoiding all transformation workloads inside the data warehouse
   • bWhen semi-structured JSON clickstream data with evolving fields is loaded into a cloud warehouse for later transformation
   • cWhen sensitive data must be cleansed and masked before it enters the target system to meet compliance rules
   • dWhen structured relational data requires strict schema enforcement on a secondary processing server before loading
   Show the answer

   Answer: b · When semi-structured JSON clickstream data with evolving fields is loaded into a cloud warehouse for later transformation

   ELT is ideal for semi-structured data like evolving JSON because the schema is applied after loading in the warehouse, whereas the compliance-focused option describes ETL, since sensitive data often requires cleansing and masking before it reaches the target system.

   Read the full bite: Differences between ETL and ELT, and when to choose each

6. Question 6 of 30

   Why has the ELT pattern become more prevalent than ETL in modern data architectures?

   • aThe rise of powerful cloud data warehouses allows for efficient in-database transformations.
   • bELT completely eliminates the need for data transformation, simplifying the entire process.
   • cData extraction and loading tools for ELT are fundamentally faster than those for ETL.
   • dELT pipelines are inherently more secure and compliant by default.
   Show the answer

   Answer: a · The rise of powerful cloud data warehouses allows for efficient in-database transformations.

   ELT's popularity is driven by powerful cloud data warehouses that can efficiently handle transformations on raw data, offering greater flexibility. ETL, not ELT, is often chosen for strict compliance because it allows data to be cleaned or anonymized before loading.

   Read the full bite: ETL vs. ELT: Key Differences and When to Use Each

7. Question 7 of 30

   For an automated pipeline loading daily CSVs from cloud storage to a database, which architecture is most efficient and scalable?

   • aA developer manually uploads the CSV to a processing server that loads the entire file into memory before a bulk insert into the database.
   • bAn S3 event triggers a Lambda function that downloads the entire CSV into memory, parses it, and inserts each row individually into RDS.
   • cAn S3 event triggers a Lambda function to stream the CSV, batch records, and execute a single batch INSERT into RDS.
   • dAn EC2 instance with a cron job polls S3, downloads the CSV, and inserts each row individually into an RDS database.
   Show the answer

   Answer: c · An S3 event triggers a Lambda function to stream the CSV, batch records, and execute a single batch INSERT into RDS.

   The correct architecture leverages event-driven serverless functions to stream and batch data, avoiding memory limits and inefficient row-by-row database operations. Distractor D, while event-driven and serverless, fails by loading the entire file into memory and performing individual inserts, which are both anti-patterns for scalability.

   Read the full bite: Build a pipeline to load CSVs into a database

8. Question 8 of 30

   Which design best distinguishes a production-ready daily CSV pipeline from a simple demo script?

   • aUsing an event trigger, validating the file, loading into a staging table, merging idempotently, and alerting on failures
   • bOrchestrating the load with a complex Airflow DAG using long-lived database credentials
   • cRunning a cron job on a VM that polls for the file and appends it directly to the final table
   • dUsing an S3 event to trigger a direct load into the final table and skipping validation to keep the architecture serverless
   Show the answer

   Answer: a · Using an event trigger, validating the file, loading into a staging table, merging idempotently, and alerting on failures

   Option A captures the card's production requirements: event-driven triggers, pre-load validation, idempotent staging-to-final loading, and observability. Option C is the most tempting distractor because a cron job on a VM is a common beginner choice, but it introduces latency, unnecessary cost, and duplicate-row risks.

   Read the full bite: Outline an automated pipeline to load a daily CSV into a database

9. Question 9 of 30

   What is the most efficient, event-driven design for a pipeline that loads a daily CSV file from an S3 bucket into a database?

   • aA scheduled cron job on an EC2 instance that checks the S3 bucket for new files once per day.
   • bA CloudWatch Event rule on a fixed schedule that invokes a Lambda function to check for the file.
   • cAn AWS Glue job scheduled to run daily, which reads from S3 and writes to the database.
   • dAn S3 Event Notification that triggers a Lambda function to parse the file and load the data.
   Show the answer

   Answer: d · An S3 Event Notification that triggers a Lambda function to parse the file and load the data.

   The correct approach is event-driven, using an S3 Event Notification to trigger a Lambda function instantly upon file creation. A scheduled job on EC2 or via CloudWatch is a polling mechanism, which is less efficient and introduces unnecessary delay.

   Read the full bite: Build a pipeline to load a daily CSV into a database

10. Question 10 of 30

    An analyst needs to calculate total sales revenue by region for the last five years. Which system is best suited for this type of complex, read-heavy query?

    • aA nightly backup of the transactional database, to avoid impacting production performance while using the same schema.
    • bA data warehouse, because its denormalized schema is optimized for fast analytical queries across large historical datasets.
    • cThe transactional (OLTP) database, because it is designed to handle a high volume of concurrent queries.
    • dThe transactional (OLTP) database, because it contains the most current and granular transaction data.
    Show the answer

    Answer: b · A data warehouse, because its denormalized schema is optimized for fast analytical queries across large historical datasets.

    A data warehouse is correct because its purpose is to support complex analytical queries on historical data. Running this query on a transactional (OLTP) database would be slow and could negatively impact real-time business operations.

    Read the full bite: What is a data warehouse vs. a transactional database?

11. Question 11 of 30

    A retail company wants to analyze ten years of sales trends without slowing online checkout. Why is a data warehouse more suitable than its production OLTP database?

    • aIt uses a denormalized schema optimized for large aggregations while OLTP is optimized for fast row-level transactions.
    • bIt is essentially a larger version of the OLTP database with more memory and CPU for bigger datasets.
    • cIt normalizes historical data into third-normal form to reduce storage costs and eliminate redundancy.
    • dComplex queries could run directly on the OLTP database if more indexes were added, avoiding the need for a separate warehouse.
    Show the answer

    Answer: a · It uses a denormalized schema optimized for large aggregations while OLTP is optimized for fast row-level transactions.

    The card explains that warehouses use denormalized schemas to maximize read performance for analytical aggregations, while OLTP databases rely on normalized schemas for fast, consistent row-level writes. Option B is tempting but wrong because the card explicitly flags the belief that a warehouse is simply a bigger OLTP database as a major red flag, since the two systems are architected for fundamentally different workloads.

    Read the full bite: Explain data warehouse purpose and how it differs from OLTP

12. Question 12 of 30

    When a business needs to analyze historical sales trends over multiple years, which database system is typically preferred and why?

    • aA data warehouse, because it provides real-time data updates and immediate operational insights.
    • bA data warehouse, because its architecture is optimized for complex, read-heavy analytical queries on large, historical datasets.
    • cA transactional database, because its normalized structure ensures data integrity for all historical records.
    • dA transactional database, because it efficiently handles a high volume of small, individual data retrievals for reporting.
    Show the answer

    Answer: b · A data warehouse, because its architecture is optimized for complex, read-heavy analytical queries on large, historical datasets.

    Data warehouses are designed for understanding the business through historical data analysis, handling complex queries across large datasets. Transactional databases, while ensuring data integrity, are optimized for real-time operational transactions and high volumes of small, fast operations, not complex historical trend analysis.

    Read the full bite: What is a data warehouse vs. a transactional database?

13. Question 13 of 30

    When encountering 10% nulls in a key numerical column, what is the most critical first action before choosing an imputation method?

    • aAnalyze the distribution of the non-null values to determine the best imputation statistic.
    • bInvestigate the root cause and characteristics of the null values.
    • cImpute the nulls using a standard method like the mean or median.
    • dDrop all rows containing nulls to simplify the dataset for analysis.
    Show the answer

    Answer: b · Investigate the root cause and characteristics of the null values.

    The card emphasizes diagnosing the problem before prescribing a solution. Understanding the root cause and characteristics of nulls is crucial to determine the most appropriate and least misleading handling strategy, rather than immediately applying an imputation method.

    Read the full bite: How would you handle 10% nulls in a key column?

14. Question 14 of 30

    A key numerical column in your dataset has 10% null values. What is the best initial step to take?

    • aDrop all rows with null values to create a clean, complete dataset for analysis.
    • bInvestigate the cause and distribution of the nulls to inform your handling strategy.
    • cImmediately fill the nulls with the column's median, as it is robust to outliers.
    • dImmediately fill the nulls with the column's mean, as it is a standard default.
    Show the answer

    Answer: b · Investigate the cause and distribution of the nulls to inform your handling strategy.

    The correct first step is always to investigate why data is missing before applying a solution. Dropping 10% of rows is a significant data loss, and imputing with the mean or median without understanding the data's distribution is premature.

    Read the full bite: How would you handle 10% null values in a key column?

15. Question 15 of 30

    You are building a dashboard and find 10% nulls in a key numerical column. Which approach best balances accuracy and transparency?

    • aUse a deep-learning model to predict each missing value precisely
    • bFill all nulls with zero to preserve row count
    • cImpute using mean or median based on distribution and add an indicator column for missingness
    • dDrop the affected rows to keep the dataset clean
    Show the answer

    Answer: c · Impute using mean or median based on distribution and add an indicator column for missingness

    The card recommends choosing mean or median imputation matched to the distribution and adding an is_missing flag so downstream users can filter imputed records. Dropping rows (D) is a common red flag because losing 10% of data can silently shift averages and distort dashboard KPIs.

    Read the full bite: How do you handle 10% nulls in a key numerical column?

16. Question 16 of 30

    What is the most reliable method for confirming an email address is both deliverable and actually controlled by a user?

    • aImplementing strict real-time validation checks exclusively on the frontend.
    • bPerforming a synchronous DNS lookup to check for the domain's MX records.
    • cValidating the format using a comprehensive, RFC 5322-compliant regular expression.
    • dSending a verification email with a unique link for the user to click.
    Show the answer

    Answer: d · Sending a verification email with a unique link for the user to click.

    Sending a verification email is the only method that confirms both deliverability and user control. A complex regex is a common but flawed approach, as it can reject valid emails and cannot prove ownership.

    Read the full bite: What validation checks would you implement for an email field?

17. Question 17 of 30

    Which validation method provides the most definitive proof that a user-submitted email address is both valid and owned by the user?

    • aSending a verification email with a unique link that the user must click to confirm.
    • bEmploying a complex regex to ensure full compliance with RFC 5322 standards.
    • cChecking for the presence of an '@' symbol and non-empty local and domain parts.
    • dPerforming a DNS lookup for MX records to confirm the domain can receive mail.
    Show the answer

    Answer: a · Sending a verification email with a unique link that the user must click to confirm.

    The card states that the "ultimate validation is an asynchronous confirmation loop" where the user clicks a unique link, providing "final proof of validity" and ownership. While MX record lookups confirm domain deliverability, they don't verify the specific inbox or user ownership, and complex regex is explicitly discouraged as impractical.

    Read the full bite: How would you validate user-submitted email addresses at ingestion?

18. Question 18 of 30

    When designing ingestion guards for a user-submitted email field, which practice best aligns with data-quality engineering principles?

    • aParse with an RFC 5322-aware validator, check DNS MX records, normalize whitespace and domain casing, and deduplicate before loading.
    • bReject plus signs and quoted local-parts as invalid, lowercase the entire address, and use an SMTP handshake to confirm the mailbox exists.
    • cUse one strict regex to validate syntax, then send a verification email and consider that part of ingestion validation.
    • dValidate syntax against RFC 5322 and conclude that successful validation guarantees the mailbox exists and can receive mail.
    Show the answer

    Answer: a · Parse with an RFC 5322-aware validator, check DNS MX records, normalize whitespace and domain casing, and deduplicate before loading.

    Option A reflects the multi-layer approach the card describes: RFC-aware syntax parsing, DNS MX verification, normalization, and deduplication. Option D is tempting because RFC validation sounds comprehensive, but the card explicitly warns that syntax validation alone does not prove a mailbox exists.

    Read the full bite: What validation checks would you implement for a user-submitted email field?

19. Question 19 of 30

    A team finds their analytics are unreliable due to inconsistent data formats. What is the most effective strategy to proactively improve data quality at the source?

    • aEnforce a schema on ingestion, rejecting data that doesn't match the required format.
    • bRely on data consumers to write more defensive queries that can handle unexpected data types.
    • cImplement complex cleaning logic in the final analytics pipeline to fix inconsistencies.
    • dManually inspect and correct the data in the data lake each day before analysis.
    Show the answer

    Answer: a · Enforce a schema on ingestion, rejecting data that doesn't match the required format.

    Enforcing a schema on ingestion acts as a proactive quality gate, preventing bad data from ever entering the system. Implementing downstream cleaning is a reactive, more expensive approach that doesn't fix the root cause.

    Read the full bite: What is a data schema and why enforce it on ingestion?

20. Question 20 of 30

    What is the primary reason for enforcing a data schema at the point of data ingestion?

    • aTo prevent malformed or invalid data from entering the system, thereby ensuring the reliability and trustworthiness of downstream analytics.
    • bTo reduce the computational resources required for data processing by pre-validating data.
    • cTo standardize data formats, making it easier for different systems to share and integrate data.
    • dTo enable faster data retrieval and querying by optimizing the physical storage layout.
    Show the answer

    Answer: a · To prevent malformed or invalid data from entering the system, thereby ensuring the reliability and trustworthiness of downstream analytics.

    The card highlights that schema enforcement at ingestion acts as a "critical quality gate" to prevent "garbage in, garbage out," ensuring data is consistent and reliable for analytics. While schemas do aid standardization, the core benefit of enforcement at ingestion is proactive quality control, not just standardization or resource optimization.

    Read the full bite: What is a data schema and why enforce it on ingestion?

21. Question 21 of 30

    Why should a data schema be enforced at ingestion rather than treated as passive documentation?

    • aIt pushes validation downstream so analysts can decide which records to keep.
    • bIt acts as a fail-fast contract that rejects bad records before they reach the lake or warehouse.
    • cIt prevents schema evolution from ever being needed when adding new columns.
    • dIt ensures all data is stored in a relational database for easier SQL access.
    Show the answer

    Answer: b · It acts as a fail-fast contract that rejects bad records before they reach the lake or warehouse.

    Enforcing a schema at ingestion establishes a fail-fast contract that rejects bad records before they reach downstream systems, preventing silent corruption and preserving trust. Option A is wrong because pushing validation to query time makes data quality reactive and forces analysts to debug type errors and nulls instead of preventing them.

    Read the full bite: What is a data schema and why enforce it at ingestion?

22. Question 22 of 30

    When setting SLOs for API latency, why can the mean alone give a misleading view of user experience?

    • aA few extreme tail values can pull the mean up and hide suffering at the p95
    • bThe mean and standard deviation fully describe latency because request times are normally distributed
    • cOutliers are usually measurement noise, so the mean smooths them away to show true performance
    • dPercentiles are only useful for large-scale systems and are too noisy for smaller services
    Show the answer

    Answer: a · A few extreme tail values can pull the mean up and hide suffering at the p95

    Latency data is right-skewed, so a handful of multi-second outliers can inflate the mean and mask terrible tail experiences. Distractor B is tempting but wrong because standard deviation assumes a normal distribution, whereas latency distributions have a long tail and a hard floor near zero.

    Read the full bite: Why prefer median and p95 over mean for API latency?

23. Question 23 of 30

    An API's latency dashboard shows a mean of 500ms but a median of 150ms. What does this discrepancy most likely indicate?

    • aThe median is misleading because it ignores outliers; the true typical experience is closer to the 500ms mean.
    • bThe system is inconsistent, with roughly half of requests taking 150ms and the other half taking 500ms.
    • cMost requests complete around 150ms, but a few very slow requests are skewing the mean upward.
    • dThe latency follows a normal (bell-curve) distribution, so this difference points to a data collection error.
    Show the answer

    Answer: c · Most requests complete around 150ms, but a few very slow requests are skewing the mean upward.

    The median represents the typical experience. A mean that is much higher than the median indicates a right-skewed distribution with a 'long tail' of a few very slow requests, which pull the average up without affecting the midpoint.

    Read the full bite: Why use median/p95 for API latency instead of the mean?

24. Question 24 of 30

    When analyzing API response times, why are percentiles (e.g., p50, p95) generally considered more informative than the arithmetic mean?

    • aPercentiles are easier for non-technical stakeholders to understand and interpret.
    • bThe mean is primarily useful for metrics that follow a normal distribution, which API latencies typically do not.
    • cPercentiles better represent the typical and worst-case user experiences because the mean can be heavily skewed by a small number of very slow requests.
    • dPercentiles offer a more precise measure of the absolute fastest and slowest requests.
    Show the answer

    Answer: c · Percentiles better represent the typical and worst-case user experiences because the mean can be heavily skewed by a small number of very slow requests.

    The card emphasizes that API response times often have a long-tail distribution where a few slow requests can dramatically skew the mean, making it a poor representation of most users' experiences. Percentiles like p50 (median) accurately reflect the typical user's experience, while p95/p99 capture the worst-case for the majority, directly linking to user satisfaction. Option B, while true that mean is better for normal distributions, doesn't fully capture the user-centric reason for preferring percentiles for latency, which is the core of the card's argument.

    Read the full bite: Why use p50/p95 over mean for API response times?

25. Question 25 of 30

    Which scenario best justifies using a pie or donut chart for an executive dashboard?

    • aDisplaying three customer acquisition channels where organic search accounts for about 55% of the total
    • bComparing monthly website traffic from five sources to identify the exact rank order of each
    • cShowing how a quarterly budget divides among ten departments with roughly equal allocations
    • dPlotting year-over-year revenue changes for four product lines to highlight upward trends
    Show the answer

    Answer: a · Displaying three customer acquisition channels where organic search accounts for about 55% of the total

    Pie charts are appropriate only for simple part-to-whole relationships with very few categories and a recognizable majority, making the three-channel scenario ideal. The ten-department budget is a tempting distractor because it is genuinely part-to-whole, but similar angles across ten slices force viewers into error-prone judgments that undermine the chart’s purpose.

    Read the full bite: When is a pie or donut chart appropriate?

26. Question 26 of 30

    When is a pie chart generally considered an appropriate visualization choice?

    • aTo compare the growth trends of three different product lines over a decade.
    • bTo display the percentage breakdown of a single total into 3 to 5 distinct components.
    • cTo illustrate the distribution of 10-12 categories where each has a slightly different value.
    • dTo show the relative performance of individual sales representatives against a team average.
    Show the answer

    Answer: b · To display the percentage breakdown of a single total into 3 to 5 distinct components.

    Pie charts are best suited for representing a part-to-whole relationship with a small, manageable number of categories (ideally 2-5). Using them for comparisons over time or with too many similar categories makes them difficult to interpret accurately.

    Read the full bite: When is a pie chart an appropriate choice for visualization?

27. Question 27 of 30

    In which of these situations would a pie chart be the most appropriate and effective visualization?

    • aIllustrating the composition of a single team's budget, broken down into 5 spending categories.
    • bComparing a company's revenue sources this year versus last year.
    • cShowing the change in market share among 4 competitors over the last 5 years.
    • dComparing the performance of 15 different products to see which sold the most.
    Show the answer

    Answer: a · Illustrating the composition of a single team's budget, broken down into 5 spending categories.

    A pie chart excels at showing the proportional composition of a static whole with a small number of categories. Comparing data over time (A, D) is better done with line or bar charts, as judging changing wedge sizes is difficult.

    Read the full bite: When is a pie chart an appropriate visualization?

28. Question 28 of 30

    Why is a line chart a poor choice for comparing market share across five separate companies in a single quarter?

    • aIt falsely implies a sequence or trend between discrete company categories.
    • bIt compresses the y-axis and distorts the percentage scale.
    • cIt demands a color legend that overwhelms the five data points.
    • dIt makes ranking harder because viewers must judge point positions instead of bar lengths.
    Show the answer

    Answer: a · It falsely implies a sequence or trend between discrete company categories.

    A line chart encodes continuity over time, so connecting discrete companies falsely implies a nonexistent sequence or trend. While bar lengths do make ranking easier, the most tempting distractor misattributes the core issue to ranking rather than the semantic mismatch that actively misleads viewers.

    Read the full bite: Compare five companies' market share: bar or line chart?

29. Question 29 of 30

    For comparing the market share of five distinct companies at a specific moment, which chart type is most suitable?

    • aA stacked area chart, as it shows how each company's share accumulates to the total market.
    • bA line chart, as it visually connects each company's share to show their relative positions.
    • cA pie chart, because it clearly illustrates each company's contribution to the total market.
    • dA bar chart, because it effectively compares discrete values across different categories.
    Show the answer

    Answer: d · A bar chart, because it effectively compares discrete values across different categories.

    A bar chart is ideal for comparing discrete values across distinct categories, such as the market share of different companies at a single point in time. A line chart is incorrect because it implies a sequence or trend that does not exist between unrelated companies.

    Read the full bite: Bar Chart vs. Line Chart for Market Share Comparison?

30. Question 30 of 30

    You need to visualize the market share for five competing companies in a single quarter. Which chart is most appropriate for comparing their individual performance?

    • aA pie chart
    • bA bar chart
    • cA scatter plot
    • dA line chart
    Show the answer

    Answer: b · A bar chart

    A bar chart is correct because it is used to compare values across discrete categories (the companies). A line chart is incorrect as it would misleadingly imply a continuous relationship or trend between the distinct companies.

    Read the full bite: Bar Chart vs. Line Chart for Market Share Comparison?