Splitting Time-Signal Data for AutoML Tables

PMLE By islever Editorial Last updated

Exploring and preprocessing data for ML. Training models.

Short answer

D Manually split the data using the time-signal columns, placing validation 30 days after training and testing 30 days after validation, so the newest rows are held out.

Question

You work for a large hotel chain and have been asked to assist the marketing team in gathering predictions for a targeted marketing strategy. You need to make
predictions about user lifetime value (LTV) over the next 20 days so that marketing can be adjusted accordingly. The customer dataset is in BigQuery, and you are
preparing the tabular data for training with AutoML Tables. This data has a time signal that is spread across multiple columns. How should you ensure that
AutoML fits the best model to your data?
  1. Manually combine all columns that contain a time signal into an array. AIlow AutoML to interpret this array appropriately. Choose an automatic data split across the training, validation, and testing sets.
  2. Submit the data for training without performing any manual transformations. AIlow AutoML to handle the appropriate transformations. Choose an automatic data split across the training, validation, and testing sets.
  3. Submit the data for training without performing any manual transformations, and indicate an appropriate column as the Time column. AIlow AutoML to split your data based on the time signal provided, and reserve the more recent data for the validation and testing sets.
  4. Submit the data for training without performing any manual transformations. Use the columns that have a time signal to manually split your data. Ensure that the data in your validation set is from 30 days after the data in your training set and that the data in your testing sets from 30 days after your validation set. Correct Answer

Key Takeaways

It tests whether you know that AutoML's automatic Time column split only works when a single Timestamp column holds the entire time signal — and that a fragmented signal requires a manual split instead.

This PMLE question tests how AutoML Tables should partition tabular data when the time signal is spread across multiple columns. The correct answer is D: perform a manual chronological split using those columns, holding out the newest data for validation and testing.

Common mistake: Option C is the tempting pick: designate an appropriate Time column and let AutoML split automatically. It is wrong here because the time signal is spread across multiple columns, so no single column can drive a trustworthy automatic chronological split.

Expert Analysis

The core issue

The question is not whether AutoML Tables can handle time — it can. It is about where the time signal lives. When one Timestamp column carries the whole signal, you mark it as the Time column during schema review, and AutoML performs a chronological split: the earliest 80% of rows train, the next 10% validate, the latest 10% test. Here the premise is that the time signal is spread across multiple columns, so no single column can act as that Time column. Google's tabular best-practices guidance states that if the time information is not contained in a single column, you should use a manual data split so the most recent data is reserved for testing.

Why D is correct

Option D follows that guidance: build the split yourself from the time-bearing columns, holding out progressively newer rows for validation and testing. Because the target is 20-day LTV, the held-out partitions must be strictly newer than training; the 30-day offsets in D encode exactly that chronological, gap-separated discipline.

Why the others fail

Option C presupposes an "appropriate column" as the Time column, which contradicts the stated premise — with the signal fragmented, any single timestamp is only a proxy for true order, so AutoML's automatic split would be built on a partial signal. Option A's array trick does not create a usable Time column; AutoML treats rows as independent examples and cannot infer ordering from a constructed array. Option B simply ignores the time structure and invites temporal leakage, where future information informs training.

The adjacent decision

If the dataset contained one clean event_timestamp column, the answer would flip to C: no manual work, just designate the Time column and let AutoML split 80/10/10. The scenario's wording — signal spread across columns — is the single fact that forces D.

Operational consequence

Getting this wrong yields inflated offline metrics: random or proxy-ordered splits put future rows in training, and the deployed model then mis-ranks customers on the next 20 days, so marketing spend is targeted at the wrong guests.

Official Reference

https://cloud.google.com/vertex-ai/docs/tabular-data/bp-tabular

This link opens the official vendor documentation in a new tab.

Exam Strategy

When a question states that the time signal is spread across multiple columns, immediately distrust any option relying on AutoML's automatic Time column split or an automatic random split — both presuppose a single Timestamp column. Instead, choose the option that builds the training/validation/test partitions manually in chronological order with the newest rows held out. Treat fixed gap values (like 30 days) as supporting detail rather than the deciding factor; the deciding factor is who performs the split and whether future rows can leak into training.

Frequently Asked Questions

When should I use a Time column instead of a manual split in AutoML Tables?

Use the Time column only when a single Timestamp column contains the full time signal; AutoML then assigns the earliest 80% of rows to training, the next 10% to validation, and the latest 10% to testing. If the signal spans several columns, split manually instead.

Why does an automatic data split hurt an LTV prediction model?

AutoML treats rows as independent examples, so a random or proxy-ordered split can place future rows in training. That leaks information, inflates validation metrics, and produces 20-day LTV forecasts that mis-rank customers for the marketing campaign.

Does specifying a Time column override a manual data split?

No. The Time column selection takes effect only if you have not specified a data-split column. Once you define a manual split, that split governs how rows are assigned to training, validation, and testing.

More PMLE FAQ →

Community Votes

D
81%
C
19%

81% of anonymous learners picked answer D. Votes are pick records left by other test-takers — they are not the verified answer.

Discussion 53 contributions

Compiled from public community threads and reader submissions, lightly edited for clarity. Handles and vote tallies are shown as recorded, and comments reflect the views of individual contributors.

From public community threads 30 of 53

  1. kkd14 👍 26
    Should be D. As time signal that is spread across multiple columns so manual split is required.
  2. DucLee3110 👍 15
    C You use the Time column to tell AutoML Tables that time matters for your data; it is not randomly distributed over time. When you specify the Time column, AutoML Tables use the earliest 80% of the rows for training, the next 10% of rows for validation, and the latest 10% of rows for testing. AutoML Tables treats each row as an independent and identically distributed training example; setting the Time column does not change this. The Time column is used only to split the data set. You must include a value for the Time column for every row in your dataset. Make sure that the Time column has enough distinct values, so that the evaluation and test sets are non-empty. Usually, having at least 20 distinct values should be sufficient. https://cloud.google.com/automl-tables/docs/prepare#time
  3. george_ognyanov 👍 10
    Correct answer is D. It clearly says the time signal data is spread across different columns. If it weren't then C would be correct and your point would be valid. However, in this case the answer is D 100%. https://cloud.google.com/automl-tables/docs/data-best-practices#time
  4. GogoG 👍 8
    Correct answer is C - AutoML handles training, validation, test splits automatically for you when you specify a Time column. There is no requirement to do this manually.
  5. Werner123 👍 7
    I think the answer is C. In this case I am interpreting time signal as the features that hold predictive power as a function of time i.e. time signal. There is no indication to how much data is available so using the 30 days after mark is not wise. You only have 30 days worth of data for validation set. If you have a few years worth of data this seems like a unnecessary small validation set.
Show 25 more comments
  1. gcp2021go 👍 4
    it should be C, since Automl tables give you the option of indicate the time column. But in practice, D could work.
  2. sensev 👍 4
    Also think it is D, since it mentioned that the time signal is spread across multiple columns.
  3. atlas_lyon 👍 3 Answered D
    A Wrong, Even if columns are combines into a 1D-array(column), the time signal should be noticed to autoML anyway. Automatic split cannot work since we need more than 20 days history B Wrong, Without indicating time signal to AutoML, data would leak in (time leakage) in training/validation/test sets C Wrong, but might be possible if time signal wouldn't have bee spread across multiple columns D True, because time signal is spread accross multiple columns require to manually split the data. Since we want to predict LTV over the next 20 days, it is necessary to have at least 20 days history between the splits (30 seems okay: 10 days predictions) Validating and testing on the last 2 months seems reasonable for marketing purpose (usually seasonal).
  4. guilhermebutzke 👍 3 Answered D
    thinking that "spread across multiple columns" seems like "columns with redundant information," and considering how AutoML can deal with correlated columns, I think option C is the best choice, with no need for a manual split. However, "time information is not contained in a single column" is the same thing as "time signal that is spread across multiple columns." I agree that D could be the best option. Then, I tend to think that D is the best choice because the text could be more clearly expressed in redundant options.
  5. salsabilsf 👍 3
    From the link you provided, I think it's A : The Time column must have a data type of Timestamp. During schema review, you select this column as the Time column. (In the API, you use the timeColumnSpecId field.) This selection takes effect only if you have not specified the data split column. If you have a time-related column that you do not want to use to split your data, set the data type for that column to Timestamp but do not set it as the Time column.
  6. ramen_lover 👍 3 Answered C
    You may have multiple columns that have a data type of Timestamp. However, you can only choose one "Time column". This Time column is used only for splitting the dataset. Therefore, you don't need to manually split the dataset even if you have multiple timestamp columns. Also, "30 days" should be "20 days" in D; AutoML does it for you if you correctly specify "Time column". https://cloud.google.com/automl-tables/docs/prepare#time
  7. pddddd 👍 3
    A. you need timestamp column, not array B. You need timestamp column, so this wont work - AutoML will not automatically create a timestamp colument C. You do not have a timestamp column, so this won't work D. Seems the only one viable
  8. rgrand8 👍 3 Answered D
    CF this link : https://cloud.google.com/automl-tables/docs/data-best-practices#time Time is spread over multiple column thus leading to manual split
  9. PhilipKoku 👍 2 Answered D
    D)D is correct, as this would satisfy the days criteria mentioned in the question. 30 days is more than 20 days, and the prediction model can be used on a validation dataset to validate the results for the next 20 days.
  10. Sum_Sum 👍 2 Answered D
    "data has a time signal that is spread across multiple columns" - I interpret as having > 1 timeseries column. AutoML knows how to deal with a single column but not multiple hence answer is D
  11. datawizard 👍 2 Answered D
    Shoud be D, since time signal is in multiple columns. See https://cloud.google.com/automl-tables/docs/data-best-practices#time
  12. hiromi 👍 2 Answered D
    D https://cloud.google.com/automl-tables/docs/data-best-practices
  13. Mickey321 👍 2 Answered C
    Either C or D but leaning towards C as not get the 30 days in D
  14. adydigi 👍 2
    Anyone wants to discuss all the questions in slack?
  15. shahriar096 👍 2 Answered C
    C is correct answer
  16. hghdh5454 👍 1
    C. Submit the data for training without performing any manual transformations, and indicate an appropriate column as the Time column. Allow AutoML to split your data based on the time signal provided, and reserve the more recent data for the validation and testing sets. This approach ensures that AutoML can handle the time-based nature of the data properly. By providing the Time column, AutoML can automatically split the data in a way that respects the time-based structure, using more recent data for validation and testing. This approach is especially important for time-series data, as it helps prevent leakage of future information into the training set, ensuring a more accurate and reliable model.
  17. testtnjhsd 👍 1 Answered C
    Why Option C is correct: AutoML is specifically designed to handle time-series forecasting and temporal data. By designating a specific column as the Time column, the service automatically understands the chronological order of the records. It will perform a time-based split—ensuring that the training set consists of older data, and the validation and testing sets consist of more recent data. This mimics real-world scenarios where you use historical data to predict future outcomes, avoiding the mistake of using "future" information to predict the "past."
  18. b7ad1d9 👍 1 Answered D
    Option D because AutoML needs clear TIME column to autosplit the training, validation ,test sets. Since that is not provided, you have to manually split the records. Not a very well-formed question. From: https://cloud.google.com/vertex-ai/docs/tabular-data/bp-tabular Provide a time signal If the time information is not contained in a single column, you can use a manual data split to use the most recent data as the test data, and the earliest data as the training data. AutoML can
  19. Krish6488 👍 1 Answered C
    Since AutoML is good enough to perform the splits, C appears to be the right answer. Moreover, time information across multiple columns which requires manual split as per option D is different from the question's scenario where the time signal is spread across multiple columns which can be hours, months, days, etc. if we can define in AutoML the right time signal column, its enojugh to split the data and pick most recent data as test data and earliest data as test data
  20. lance_yu 👍 1 Answered C
    This is a time-dependent prediction problem. For AutoML Tables (Vertex AI Tabular), Google's recommended practice is: Provide a Time column that represents the temporal ordering of examples. Let AutoML perform its built-in time-aware data split. Use older data for training and newer data for validation/testing. This prevents data leakage where future information accidentally influences training.
  21. giaZ 👍 1
    Looking at link: https://cloud.google.com/automl-tables/docs/data-best-practices#time I'm really undecided between C and D. The latter seems the right one but it should say 20 days rather than 30..So, maybe C is the only option.. But, is it ok to set one column as the Time column in AutoML Tables even if the time signal is not contained in a single column?
  22. Shilpika98 👍 1
    Should be D, for multiple time columns, a manual split is required. "f the time information is not contained in a single column, you can use a manual data split to use the most recent data as the test data, and the earliest data as the training data." source: https://cloud.google.com/automl-tables/docs/data-best-practices#time
  23. John_Pongthorn 👍 1 Answered D
    I go with D: https://cloud.google.com/automl-tables/docs/data-best-practices#time Read it carefully at the last paragraph of the topic: If the time information is not contained in a single column, you can use a manual data split to use the most recent data as the test data, and the earliest data as the training data.
  24. rajshiv 👍 1 Answered C
    C is the right answer as manually splitting data data based on time adds unnecessary complexity. AutoML Tables can handle the time-based splits for us automatically when we specify the time column. Option D requires more manual intervention and introduces the risk of making errors in the data splitting process.
  25. coupet 👍 1 Answered D
    D is Correct - A time signal spread across multiple columns in a spreadsheet or data table would typically represent a time-series data where each column corresponds to a specific time point or interval, and the values in each column represent the signal's value at that time

Reader comments & corrections

No reader corrections yet — spotted an error or have a note? Share it below.

Log in to comment, report an error, or add a note about this question.

How to Minimize Development Time for Weekly TensorFlow Retraining in Vertex AI?

S-grade

How to Handle Categorical Features in BigQuery ML for Fast Deployment?

S-grade

How to preprocess BigQuery data for TensorFlow?

S-grade

How to Track Model and Prediction Lineage in Vertex AI?

S-grade

Practice All PMLE Questions

Access 65 questions with complete answers and detailed explanations.

View Full PMLE Practice Test →

← Back to PMLE Study Guide