Max Agent
Articles

Turn a question into an alert specification

Define the event, condition, interval and invalidation rule, then write an explicit policy for repeat notifications and test it.

MaxData EditorialPublished Updated

Start your alert exercise with a question you can inspect: “Has my selected price series fallen by the amount I specified over the period I chose?” Then write what you intend to check after receiving a message. For this guide, make that action a review of the data and your notes.

Everything described as a checklist, convention or workflow below is our editorial proposal. Numerical examples are invented for configuration practice. They express no view on an asset’s value or future direction. Product statements refer specifically to the supplied documentation; verify the settings you intend to use in your own interface.

1. Write a short observation brief

Our proposed brief has three sentences. The first states the question. The second identifies the exact data series. The third names the review you will perform when notified. Try this fictional example: “I want to inspect a decline in a selected asset pair. I will use the named provider’s closing values. After a notification, I will compare the timestamps and read the following four observations.”

Add an instrument identifier, quote currency, market type, provider, unit and definition link. For a metric you do not understand, leave the brief in draft and write the unanswered question beside it. Make “definition confirmed” an acceptance requirement for your exercise.

The MaxData Help Hub invites questions about MaxData and My workspace. That is the help scope described in the source. [en-c1] Use it as a place to frame a product question, such as “Which data series does this field select?”, without assuming an answer in advance.

Our proposed naming convention is plain: instrument, observation, interval and version. For example, use “PAIR decline review, 30-minute observations, v1”. Keep any intended interpretation in a separate note.

2. Specify the condition and the event separately

The Pro Alerts documentation describes an Alert as a single-condition notification and a Pro Alert as a combination of several conditions. It also describes activation when a condition begins, remains satisfied or ceases to be satisfied. [en-c2]

For our example specification, write the condition as “the selected change measure is less than minus 2%”. Treat that threshold as arbitrary. Then choose the event you want to inspect: entry into the condition, an observation while it remains true, or departure from it.

Under our proposed convention, these are three different test cases. Record your expected notification count for each. Include an observation exactly at the threshold and write how you want equality handled. Verify the interface’s behavior rather than assigning it an undocumented interpretation.

If you add another condition, give it its own definition and test. Specify whether both must be true together or whether their occurrences may fall within a chosen window. Mark that choice as an intended requirement until you have checked the implemented behavior. Do not silently substitute one interpretation for the other.

Our proposed specification method

Build a testable sentence

  1. 1

    Select

    Name the exact series and the value you intend to observe.

  2. 2

    Compare

    Write the operator, threshold and equality rule.

  3. 3

    Choose

    Specify entry, continued satisfaction or departure.

  4. 4

    Predict

    Write the expected message count for your test sequence.

Use this editorial sequence to express your intended behavior before configuring it. Four steps define the selected series, condition, notification event and expected test result.

3. Give every time setting a job

The documentation describes a time window for meeting the combined conditions and a choice of data values including open, close, minimum, maximum and average. [en-c3] Source: Pro Alerts.

Our proposed worksheet uses separate entries for observation interval, comparison lookback, combination window, cooldown, expiry and review horizon. For this exercise, choose a fictional 30-minute observation interval and a two-hour comparison lookback. Leave the combination window unused if your exercise has only one condition.

Next, specify whether your intended observation uses a completed interval. Write that intention explicitly, then check whether the selected configuration represents it. Add a timezone and your chosen convention for labeling interval boundaries. These are worksheet requirements, not assertions about the platform’s clock behavior.

Give latency its own field. Enter a verified value for the selected source or write “not established”. During testing, record the data timestamp and message receipt time. If you can also record the evaluation time, keep it in a separate column. Avoid filling any unknown timestamp with an estimate presented as a measurement.

Finally, write how long you intend to review the observation after notification. Keep that horizon distinct from how long the alert itself should remain enabled.

4. Choose a repeat policy before enabling messages

The Pro Alerts page describes sleep time after activation, one-time or repeated triggering, a repeated-trigger limit of no more than once per minute, and an expiration option. These are descriptions in the documentation, rather than confirmation of the settings available on a particular account. [en-c4] Source: trigger settings.

Our proposed first exercise uses one message per observation episode. Define an episode yourself. For example, start it when the fictional change measure first falls below minus 2%, and require a return above minus 1% before considering another episode. Neither threshold is a suggested trading parameter.

Treat cooldown and rearming as separate requirements in your notes. Write “wait this long” for cooldown and “require this observation” for rearming. If the interface cannot express your intended rule, mark it as a manual procedure or revise the exercise. Do not describe your worksheet convention as an available product feature.

Choose one delivery channel for the test. Assign each expected episode an identifier in your log. When reviewing repeated messages, record whether they belong to the same episode under your definition, a new episode or an unresolved case. Set your own acceptable message count before looking at the results.

Our worksheet convention

Separate waiting from rearming

Cooldown requirement
Write down
How long to wait after a message
Test boundary
Immediately before and after the wait ends
If unavailable
Document a manual waiting procedure
Rearming requirement
Write down
Which observation permits a new episode
Test boundary
Before and after the reset condition
If unavailable
Document a manual reset procedure
These are proposed specification requirements; check whether your interface can express them. Comparison of a time-based waiting requirement and an observation-based rearming requirement.

5. Define invalidation without changing the question

Our proposed specification distinguishes three decisions: stop monitoring, withdraw an interpretation and permit another alert episode. Give each a separate field. For example, stop this exercise at a specified time; suspend interpretation if the data source changes; consider rearming only after the recorded reset condition.

Also choose explicit reasons to leave the review unfinished. Our suggested checklist includes an unknown timestamp, an unconfirmed unit, a mismatch between the selected pair and the brief, or missing observations. Label the outcome “insufficient information” when that is your chosen acceptance rule.

Prepare alternative review notes before the message arrives. One can say, “The configured observation was recorded; continue the planned review.” Another can say, “The data check is incomplete; make no interpretation.” A third can say, “The observation ended before the review; archive this episode.” These are proposed responses, not explanations of market behavior.

If you later revise a threshold or invalidate the original question, create a new version. In our workflow, a new version requires a fresh statement of expected behavior rather than an undocumented edit to the earlier exercise.

6. Test behavior and inspect the historical record

The Backtests documentation says a backtest can report how many times an Alert or Pro Alert would have triggered historically. It describes average and median price-change lines before and after triggering, and testing either one condition or several together. [en-c5]

Our proposed review begins with configuration behavior. Prepare a sequence covering a false condition, first entry, continued satisfaction, exit and re-entry. For each step, write the expected message count, then compare it with the observed log. Include expiry and cooldown boundaries in a separate test case.

For historical inspection, record the date range, number of triggers, your episode count, missing data and review horizon. Keep those historical observations in their own section. Do not label them as future expectations in this workflow.

Add a record of every parameter version you compared. Reserve a separate period for evaluating the unchanged specification, and note whether you selected the rule after inspecting earlier results. Our proposed checklist asks you to make sample size and possible fitting to the inspected history explicit before accepting the exercise.

If you extend the work to hypothetical transactions, require stated assumptions for fees, slippage, liquidity and order size. Mark unknown inputs as unknown. Keep that extension separate from the notification test, and make no execution or return conclusion from a message-count exercise.

Our proposed review sequence

Specify, observe, then accept

  1. Freeze the brief

    Assign a version and record expected behavior.

  2. Run controlled cases

    Check entry, persistence, exit, re-entry and timing boundaries.

  3. Observe separately

    Plan a separate evaluation period with the specification unchanged.

  4. Accept or revise

    Record agreement or document the unresolved difference.

Keep the written expectations available at every stage. A staged review from freezing the specification through controlled testing, separate observation and acceptance.

7. Use a clear acceptance checklist

Our proposed completion rule requires agreement among the written brief, saved settings and test log. Check the instrument, data definition, condition, event, time fields, repeat policy, invalidation rule and version number. Add the person responsible for reviewing the setup and the next review date.

Begin with notification-only observation under this workflow. Treat any later connection to an automated action as a separate project requiring its own specification and supervision. For the final handover, ask a reader to reconstruct your intended alert from the brief alone. Accept the exercise when the documented expectations and recorded behavior agree, or clearly record the unresolved difference.

Sources

  1. MaxData Help Hub – Pomoc i przewodniki Mój workspace — agent.maxdata.app; accessed November 10, 2026
  2. Pro Alerts — agent.maxdata.app; accessed November 10, 2026
  3. Backtests — agent.maxdata.app; accessed November 10, 2026

Start working with your own agent

The agent knows this article context and still follows the assigned department checklist.

Create a MaxData account
Max Agent
Alert specifications: events, intervals and repeat control