Tableau Quiz: Forecasting
10 questions · exam conditions
0:00
ForecastingQuestion 1 of 10

A monthly sales view has an exact forecast length of 44 periods. The analyst then changes the continuous date level from month to quarter and leaves the forecast length unchanged.

What should the analyst expect after Tableau recomputes the forecast?

The horizon remains four months because exact forecast length is stored independently of the date level.
The horizon becomes four quarters, and the model is fitted using the newly aggregated quarterly series.
The horizon becomes one quarter because four monthly periods are combined into a single quarterly mark.
The horizon remains four quarters, but Tableau continues fitting the model from the original monthly observations.
← Back to quizzes

Tableau Quiz

Tableau Quiz: Forecasting

Practice Forecasting in Tableau with focused quiz questions that help you check what you know, review explanations, and build confidence with test-style prompts.

What this quiz covers

This quiz focuses on Forecasting, giving you a quick way to practice the rules, question types, and explanations that matter most for Tableau.

How to use this quiz

Try each quiz question before looking at the correct answer. Use the explanations to review missed ideas, then come back to similar questions until the pattern feels familiar.

All questions

Question 1

A monthly sales view has an exact forecast length of 44 periods. The analyst then changes the continuous date level from month to quarter and leaves the forecast length unchanged.

What should the analyst expect after Tableau recomputes the forecast?

  1. The horizon remains four months because exact forecast length is stored independently of the date level.
  2. The horizon becomes four quarters, and the model is fitted using the newly aggregated quarterly series. (correct answer)
  3. The horizon becomes one quarter because four monthly periods are combined into a single quarterly mark.
  4. The horizon remains four quarters, but Tableau continues fitting the model from the original monthly observations.
Explanation: Whenever Tableau questions involve forecast settings and date level changes, you need to think about how the forecast length unit is tied to the current date granularity — not locked to whatever granularity was active when you first set it. In Tableau, an exact forecast length of 44 periods means "4 units of whatever the current date level is." When you switch from monthly to quarterly aggregation, Tableau re-reads that setting as 44 quarters. Simultaneously, the underlying time-series model is refitted from scratch using the newly aggregated quarterly data — because the source series has fundamentally changed. This makes B the correct answer: the horizon becomes four quarters and the model uses the quarterly series. A is wrong because it assumes the forecast length is stored as an absolute calendar duration (like "4 months") that survives a date-level change. Tableau doesn't work that way — the period count is reinterpreted against the new granularity, not frozen. D makes a similar error by claiming the model continues fitting from the original monthly observations; once you change the date level, Tableau aggregates the data and rebuilds the model entirely from the quarterly marks. C introduces a plausible-sounding but false conversion logic — the idea that 44 months "collapse" into 11 quarter. Tableau doesn't perform unit conversions on your forecast length; it simply applies the same integer (44) to the new date unit. Study tip: Think of Tableau's forecast length as a count of ticks on the current axis, not a fixed calendar span. Whenever the date level changes, that count stays the same but refers to the new, larger unit.

Question 2

A retailer uses Tableau to forecast weekly unit sales from historical sales alone. Management has already scheduled a major price reduction for the first forecasted week, but no comparable promotion appears in the historical series.

How should the analyst interpret the Tableau forecast for that week?

  1. The point estimate automatically includes the promotion because the event occurs within the selected forecast horizon.
  2. The upper interval bound represents the expected promotional lift, while the point estimate represents no promotion.
  3. The forecast extrapolates the historical time-series pattern and does not automatically estimate the causal effect of the planned promotion. (correct answer)
  4. The forecast excludes historical trend so that the planned promotion can be treated as the only source of future change.
Explanation: Whenever you see a question about Tableau's forecasting capabilities, anchor yourself to one core principle: Tableau's built-in forecast uses exponential smoothing on historical time-series data only. It identifies trend and seasonality patterns from past observations and projects them forward — nothing more. That's exactly why C is correct. The forecast for the promotional week is simply an extrapolation of the historical sales pattern. Because no comparable price reduction exists in the historical data, the model has no mechanism to "see" or account for the planned promotion. The point estimate reflects what sales would look like if historical patterns continued unchanged — not what will happen when the price drops. A is wrong because Tableau's forecast engine doesn't ingest planned business events. The fact that the promotion falls within the forecast horizon is irrelevant — the model only learns from past data, not from your calendar or pricing plans. B is a tempting distractor because confidence intervals feel like they might capture "best" and "worst" cases for different scenarios. But that's not what they represent. The upper and lower bounds reflect statistical uncertainty around the time-series extrapolation — not the differential impact of a promotion versus no promotion. D is incorrect and essentially inverts how forecasting works. Tableau's model absolutely incorporates historical trend; that's a foundational component of exponential smoothing. Excluding trend would make the forecast less accurate, not more useful for promotional planning. Study tip: On Tableau exam questions, if you see "causal," "external event," or "planned promotion" language, that's a signal that the answer involves the limits of time-series forecasting — which cannot capture drivers absent from historical data.

Question 3

For a stable monthly series, Tableau's prediction interval is relatively narrow for the first forecasted month but considerably wider for the twelfth forecasted month. The confidence level is unchanged.

What is the best conceptual explanation for the wider interval?

  1. Forecast uncertainty can accumulate with horizon because estimates farther from observed data depend on more unobserved future evolution. (correct answer)
  2. Tableau automatically raises the confidence level for later periods even when the selected level remains unchanged.
  3. A wider interval proves that the twelfth month has stronger seasonality than the first forecasted month.
  4. The model uses fewer historical observations for each successive period and eventually retains only the latest observation.
Explanation: When you see a question about forecast intervals in Tableau, the core concept to anchor on is forecast uncertainty and how it behaves over time. Prediction intervals aren't static — they reflect how much the model's confidence degrades as it projects further into the future. Forecasting models estimate future values by extrapolating patterns from historical data. Each step forward in time introduces additional uncertainty, because the model has never observed that future state. By the twelfth forecasted month, the model has had to "stack" twelve steps worth of unobserved evolution — errors and unknowns compound across the entire horizon. This is why A is correct: uncertainty accumulates with forecast horizon, widening the prediction interval even when the confidence level stays fixed. B is wrong because Tableau does not silently raise the confidence level for later periods. The confidence level is a user-controlled parameter; the widening interval reflects greater uncertainty, not a shifted level. Confusing the width of an interval with the confidence level is a classic trap. C is incorrect because interval width is not a measure of seasonality strength — a wider interval signals uncertainty, not a stronger seasonal pattern. D describes a fundamentally false mechanism. Tableau's exponential smoothing models use all available historical data; the model does not progressively discard observations for each successive forecast period. A useful study tip: on Tableau forecast questions, always distinguish between three separate concepts — confidence level (what you set), prediction interval width (what the model outputs), and sources of uncertainty (horizon, data volatility, model fit). Mixing these up is exactly what distractors B, C, and D are designed to exploit.

Question 4

Two regional forecasts have overlapping 95%95\% prediction intervals for the same future month. Region East has a higher point estimate than Region West.

Which conclusion is justified from this information alone?

  1. East will exceed West because a higher point estimate is definitive, regardless of interval overlap.
  2. West will exceed East because overlapping intervals indicate the ordering of point estimates is reversed.
  3. The regions have equal expected values because any interval overlap makes their point estimates statistically identical.
  4. East has the higher forecasted value, but the overlap alone does not establish a conclusive difference between the two regions. (correct answer)
Explanation: When comparing forecasts, you need to distinguish between a point estimate (the single most likely predicted value) and a prediction interval (the range of plausible outcomes). A 95%95\% prediction interval tells you where a future observation is likely to fall — but when two intervals overlap, it signals uncertainty about the relative ordering of the two regions, not a definitive ranking. Answer D is correct because both pieces of information it contains are accurate and appropriately qualified. East does have the higher forecasted value by point estimate, which is a real and meaningful piece of information. However, the overlapping prediction intervals mean you cannot rule out scenarios where West's actual outcome exceeds East's — the uncertainty ranges intersect, leaving the true difference statistically inconclusive. Answer A is wrong because a higher point estimate is not definitive when prediction intervals overlap. The overlap is precisely the evidence that uncertainty undermines a firm conclusion. Answer B contains a logical reversal — overlapping intervals do not imply the point estimate ordering is backwards. The point estimates remain what they are; overlap only introduces doubt about the outcome, not a contradiction of the estimates themselves. Answer C is wrong because overlapping intervals do not make point estimates equal or "statistically identical." Two forecasts can have clearly different point estimates and still produce overlapping intervals due to wide uncertainty ranges. A useful rule of thumb: point estimates tell you direction; intervals tell you confidence. When intervals overlap, direction is suggestive but not conclusive. On the Tableau exam, watch for answer choices that treat a higher point estimate as automatically decisive — that's a common trap.

Question 5

An analyst changes the prediction interval setting for an existing Tableau forecast from 90%90\% to 99%99\% without changing the data, date granularity, seasonality setting, or forecast horizon.

What result should the analyst generally expect?

  1. The point forecasts increase and the interval widths remain unchanged because higher confidence implies higher expected values.
  2. The point forecasts and intervals remain identical because confidence level affects only the wording in the forecast description.
  3. The point forecasts decrease and the intervals narrow because the model becomes more conservative at higher confidence.
  4. The point forecasts remain the same and the intervals widen to represent a higher requested coverage level. (correct answer)
Explanation: Whenever you see a question about forecast settings in Tableau, separate two distinct components in your mind: the point forecast (the model's best single estimate of future values) and the prediction interval (the band of uncertainty around that estimate). These are determined independently — the underlying statistical model drives the point forecast, while the confidence level controls how wide the uncertainty band must be to capture the true value at the requested rate. When you raise the prediction interval from 90%90\% to 99%99\%, you are telling Tableau: "I want a band wide enough that, if we repeated this process many times, the true value would fall inside 99%99\% of those bands." Because you changed nothing about the data or model itself, the point forecasts stay exactly the same. However, achieving higher coverage mathematically requires multiplying the standard error by a larger critical value — so the interval must widen. That makes D correct. A is wrong because confidence level has no influence on expected values; point forecasts come from the model's trend and seasonality components, not from the interval setting. B is wrong because the confidence level is a real mathematical parameter — changing it produces a measurably wider or narrower band, not just a label change. C gets the direction of the interval change backwards; higher confidence means wider intervals, not narrower ones, and point forecasts don't decrease either. A reliable rule of thumb: confidence level moves intervals only, never point forecasts. If a question changes only the confidence setting, the answer almost always involves unchanged point estimates and a shifted interval width.

Question 6

A Tableau forecast displays a 95%95\% prediction interval for each of the next 2020 months. After the actual values become available, one month falls outside its displayed interval.

Which interpretation of this result is most appropriate?

  1. The forecast failed because a 95%95\% interval must contain every actual value in the forecast horizon.
  2. The result can be consistent with a calibrated model because the interval does not guarantee coverage for every future value. (correct answer)
  3. The point forecast was accurate because no more than 5%5\% of the observed months fell outside the intervals.
  4. The confidence level should be read as a 95%95\% probability that all 2020 future values fall within their intervals.
Explanation: When you see a question about prediction intervals in Tableau forecasting, focus on what a confidence level actually guarantees — and what it doesn't. A 95%95\% prediction interval means that, for any single forecast point, there is a 95%95\% probability the actual value falls within that range. It does not mean every value across the entire forecast horizon will be captured. When you extend this across 2020 independent forecasts, you should actually expect some observations to fall outside their intervals — that's the model behaving correctly, not failing. Statistically, 5%5\% of 2020 points is 11 point, so seeing exactly one miss is a textbook-consistent result. This is why B is correct: one value falling outside its interval is entirely compatible with a well-calibrated model. A is wrong because it describes an impossible standard. A 95%95\% interval explicitly allows for misses — demanding 100%100\% coverage would require a 100%100\% interval, which would be infinitely wide and useless. C is subtly wrong in its reasoning: while the arithmetic (1/20=5%1/20 = 5\%) might seem to support accuracy, the answer frames this as evidence the point forecast was accurate, which conflates point estimates with interval coverage — two different concepts. D describes a common misconception. The 95%95\% applies independently to each interval, not jointly to all 2020 values simultaneously; the probability that all 2020 fall inside would actually be 0.9520≈36%0.95^{20} \approx 36\%. Remember: on Tableau exam questions about forecasting, always distinguish between per-interval confidence and joint coverage across multiple periods — they are not the same thing.

Question 7

A Tableau view initially shows total monthly revenue and one company-wide forecast. The analyst adds Region to the view, producing a separate time-series line for each region.

Which consequence is most important when interpreting the resulting forecasts?

  1. Tableau can fit forecasts separately by region, so each line may use a different model and have different intervals. (correct answer)
  2. Tableau must retain the company-wide model, so every region receives the same trend and seasonal parameters.
  3. Tableau allocates the original company-wide forecast to regions in proportion to each region's latest actual value.
  4. Tableau combines all regional observations into one longer time series before estimating a single regional forecast.
Explanation: When you add a dimension like Region to a Tableau view that already has a forecast, it's essential to understand how Tableau's forecasting engine responds — specifically, whether it produces one shared model or multiple independent models. Tableau's forecasting algorithm treats each distinct mark in the view as a separate time series. The moment you introduce Region, Tableau partitions the data by region and fits an independent exponential smoothing model for each partition. That means each regional line gets its own trend component, its own seasonal component, and its own prediction interval — all calibrated to that region's historical data alone. This is answer A, and it's the most important interpretive consequence because a region with volatile data may produce wide confidence bands, while a stable region may produce narrow ones, making direct visual comparison tricky if you treat all forecasts as equally reliable. B is wrong because Tableau does not lock in a company-wide model when you disaggregate the data. Each series is modeled fresh — parameters are not shared or inherited. C is wrong because Tableau never performs proportional allocation of an aggregate forecast. It doesn't take a company-wide number and divide it down; it re-estimates from the underlying regional data directly. D is wrong because Tableau does the opposite — it splits observations by region rather than pooling them together. Pooling would defeat the purpose of adding Region to the view entirely. As a study tip: whenever you see Tableau forecast questions involving added dimensions, ask yourself "does this partition or pool the data?" Tableau always partitions — and each partition gets its own model.

Question 8

An analyst has monthly sales through December. In Tableau's Forecast Options, the analyst sets Ignore Last to 33 periods before reviewing the forecast.

What is the most useful reason for applying this setting?

  1. It removes the first three months so Tableau can place greater weight on the most recent observations.
  2. It reserves the last three actual months for comparison with estimates generated without using those months to fit the model. (correct answer)
  3. It forces Tableau to use the last three actual months as the repeating seasonal pattern for all future estimates.
  4. It suppresses the last three displayed intervals while retaining those months when Tableau estimates the model parameters.
Explanation: When working with Tableau's forecasting tools, it helps to think about a fundamental modeling principle: the data used to build a model should be separate from the data used to evaluate it. The Ignore Last setting exists precisely for this purpose. Setting Ignore Last to 33 tells Tableau to hold back the final three months of actual data when fitting the forecast model. Tableau then generates predictions for those same three months, giving you a direct side-by-side comparison between what actually happened and what the model would have predicted — without the model having "seen" those months first. This is a classic holdout validation technique, and B captures it exactly. A is wrong because Ignore Last doesn't reweight observations toward recent data — that's a function of exponential smoothing parameters, not this setting. Excluding months from the end of the series would actually reduce recency emphasis, not increase it. C is wrong because Tableau does not use the excluded months as a template for the seasonal pattern. Seasonality is estimated from the broader series structure, not lifted directly from the most recent cycle. D contains a tempting half-truth — the last three months are visually suppressed from the fitting process — but it gets the key detail backwards. Those months are excluded from parameter estimation, not retained for it. The whole point is that they are withheld so you can benchmark against them independently. As a study tip, whenever you see Ignore Last in a Tableau question, immediately think holdout set for model validation, not data weighting or seasonal anchoring.

Question 9

Series A has a forecast of 100100 with a 95%95\% prediction interval from 9292 to 108108. Series B has a forecast of 102102 with a 95%95\% prediction interval from 7676 to 128128. Both forecasts cover the same future period.

Which interpretation best distinguishes forecast precision from forecast accuracy?

  1. Series A is necessarily more accurate because its narrower interval means every plausible outcome is closer to the actual future value.
  2. Series B is more precise because its point estimate is higher, and the wider interval confirms it captures a broader range of accurate values.
  3. Series A is more precise because its interval is narrower, but interval width alone does not prove its point forecast will be more accurate. (correct answer)
  4. Series B is necessarily more accurate because its wider interval covers more possible actual values, making the point forecast more reliable.
Explanation: Whenever Tableau asks you to distinguish precision from accuracy in forecasting, keep these definitions sharp: precision describes how narrow (certain) a prediction interval is, while accuracy describes how close a forecast ultimately lands to the real outcome. These are independent qualities — a forecast can be precise yet wrong, or imprecise yet correct. Series A has a point forecast of 100100 with an interval spanning only 1616 units (9292 to 108108), while Series B forecasts 102102 across a 5252-unit interval (7676 to 128128). Series A is clearly more precise — its model expresses much greater certainty. However, precision says nothing about whether 100100 will actually be closer to the true future value than 102102. The actual outcome could fall outside Series A's narrow band entirely, while Series B's wide interval comfortably contains it. Answer C captures this correctly: narrower interval = higher precision, but precision does not guarantee accuracy. Answer A is wrong because it conflates precision with accuracy. A narrow interval means fewer plausible outcomes, not that those outcomes are necessarily correct. Answer B contains two errors: precision is not determined by the magnitude of the point estimate, and a wider interval does not indicate broader "accuracy" — it indicates less certainty. Answer D similarly misreads the wider interval as a reliability advantage for the point forecast, which is backwards; a wider interval reflects more uncertainty, not more trustworthiness. Your study tip: on forecast questions, always ask yourself two separate questions — "How certain is this model?" (precision) and "How close did it get?" (accuracy). Tableau tests whether you can hold both concepts independently without letting one imply the other.

Question 10

A view contains 3636 months of demand data. Business operations create a known annual cycle, but Tableau's automatic model does not select seasonality because an unusual recent disruption weakens the detected pattern.

Which forecast configuration most directly tests the analyst's defensible business assumption about the cycle?

  1. Set seasonality to custom with a cycle length of 1212 monthly periods. (correct answer)
  2. Set the forecast length to exactly 1212 periods and retain automatic seasonality.
  3. Aggregate the dates to years and set seasonality to a cycle length of 1212 years.
  4. Increase the prediction interval to 99%99\% so the annual pattern enters the model.
Explanation: When Tableau's automatic seasonality detection fails — often because a recent disruption masks a regular pattern — you can override it by manually specifying the cycle length. This question tests whether you understand that custom seasonality is the mechanism for injecting domain knowledge into a forecast model when the algorithm won't cooperate. If your business runs on an annual cycle and your data is monthly, the correct move is to tell Tableau explicitly: "treat every 1212 periods as one complete cycle." That's exactly what answer A does — setting seasonality to custom with a cycle length of 1212 monthly periods. This directly encodes your defensible business assumption into the model, overriding the automatic detection that the disruption confused. Answer B is tempting but misses the point. Changing the forecast length to 1212 periods only controls how far ahead Tableau projects; it does nothing to the seasonality model itself. Automatic seasonality is still switched off, so the annual pattern still won't be captured. Answer C is a category error. If you aggregate dates to years, each data point is one year — you'd need decades of data for a 1212-year seasonal cycle to make sense. This destroys the monthly granularity that makes annual seasonality detectable. Answer D reflects a misconception about prediction intervals. Widening the interval to 99%99\% only expands the uncertainty band around the forecast; it has no effect on which seasonal model is selected. Study tip: On Tableau forecast questions, distinguish between three separate controls — seasonality model, forecast length, and prediction interval. They are independent. When a question asks about encoding a known cycle, go straight to seasonality settings.