The most useful Skill starts with a method you already repeat. The examples below turn that method into instructions you can copy, adapt, and run whenever you need it. Some Skills are built for a conversation, where you supply the property and follow the evidence. Others are built to need nothing at run time, so a Routine can run them for you on a schedule and email the result.
Start with the closest example, change the property and the details, and test it before you rely on it. A word such as "daily" or "weekly" in a Skill's name does not schedule it. You choose when to run it, or have a Routine run it.
How to use these examples
Example 1: Investigate a pickup change
Example 2: Three-week forecast
Example 3: Corporate-account review with a file
Example 4: Unmapped-value check
Example 5: Weekly commercial brief
Example 6: Monthly KPI review
Example 7: Pickup check
Running a Skill in a Routine
Building a decision-support Skill
Testing and improving your version
How to use these examples
Each example contains the four fields used when creating a Skill: Name, Identifier, Description, and Instructions. Under each one, Where it fits says whether to run it in Chat, in a Routine, or both.
Before saving your version:
Replace the example property, dates, thresholds, metrics, and output with your real workflow.
Remove steps that do not contribute to the one job the Skill should complete.
Keep important definitions, comparison rules, and missing-data behavior explicit.
Decide whether the result should be prescribed, such as fixed table columns, or guided, such as findings grounded in the results.
Test the normal, missing-input, no-data, and largest-intended-scope cases.
Example 1: Investigate a pickup change
Use this Skill when a pickup signal deserves a familiar investigation but you do not yet know which line of investigation will explain it.
The property, dates, trigger, and evidence can change between runs. The Skill keeps the method steady while the conversation follows what the results show.
Where it fits: Chat. Each run begins with a different signal, so you decide the next step as you go.
Name: Investigate a pickup change
Identifier: /investigate-pickup-change
Description: Investigates a pickup concern for selected properties and stay dates, validates the evidence, and follows the relevant line of investigation.
Instructions:
PURPOSE
Guide the user from a pickup signal to a supported conclusion without running unrelated analysis or assuming that action is required.
INPUTS
Ask once for the following items that were not supplied: property; stay date or stay-date range; what prompted the review (weakening pace, low-rated pickup, a forecast gap, unexpected cancellations, or a group approaching cutoff); and the pickup window, if different from the seven calendar days ending yesterday (if today is the 9th, the 2nd to the 8th).
DATA SCOPE
Topic: Pick-up
Update Date: the supplied pickup window, otherwise the seven calendar days ending yesterday (if today is the 9th, the 2nd to the 8th)
Stay Date: the supplied stay dates, otherwise ask before running
Start with: total Units Pick-up, ADR Pick-up, and Accommodation Revenue Pick-up by stay date
Add only the dimensions and comparisons required by the evidence
STEPS
Confirm the exact property when the supplied name is ambiguous.
Confirm the comparison period, currency, revenue basis, stay-date scope, and pickup window.
Establish whether the movement is isolated or part of a broader trend. Use more than one pickup window when that distinction matters.
Use forecast, budget, or the appropriate historical comparison when it helps establish whether the movement is expected.
Follow only the relevant line of investigation, for example Segment, Channel, Inventory, Rate Plan, Group, or cancellations.
If a result looks implausible, check scope, field choice, currency, mapping, freshness, and result completeness before treating it as a commercial signal.
Stop when the evidence supports an outcome. Do not keep adding breakdowns that do not change the conclusion.
OUTPUT
State the signal, whether it is isolated or part of a trend, and the main supported contributor.
Include a compact evidence table for the branch that explains the result.
Finish with one outcome: expected or no action required; a Revenue Management action, including a monitoring response, deserves review; another team or product owns the next step; or the available evidence cannot support a conclusion.
Invite the user to challenge a figure, change the scope, or investigate another branch.
GUARDRAILS
Treat zero pickup as a valid result. Do not assume that correlation proves causation or that every movement requires action. Do not recommend a specific price or operational change when the evidence does not support it. Disclose if any detailed result is sampled or incomplete, and do not claim an exhaustive ranking from partial results.
Try these tests:
/investigate-pickup-changewith no property, to confirm it asks once.Run it for Flagship Austin by Flyr with "weakening pace for next month".
Run it for a period with no pickup, to confirm zero is reported as a result.
Example 2: Three-week forecast
Use this Skill when you want a short forward outlook with the same weekly structure for different properties.
Select the property and optional segment for each run, then continue with the forecast week that needs closer investigation.
Where it fits: Chat. It asks which property and segment to use, and the next question usually depends on the answer.
Name: Three-week forecast
Identifier: /three-week-forecast
Description: Summarizes Forecast Units, ADR, and Accommodation Revenue for the next three complete stay weeks for a property.
Instructions:
PURPOSE
Provide a concise forward view so the user can identify weeks that deserve a closer look.
INPUTS
Ask once for the property if it was not supplied. Accept an optional segment and use any scope already included by the user.
DATA SCOPE
Topic: Forecasts & Budgets
Stay period: next three complete stay weeks
Group by: Stay Week
Metrics: AI Forecast Units, AI Forecast ADR, AI Forecast Accommodation Revenue
STEPS
Confirm the property match if necessary.
Run the forecast by Stay Week.
Apply any supplied segment consistently.
Identify the highest- and lowest-forecast weeks for Units and Revenue.
OUTPUT
Table with Stay Week, Forecast Units, Forecast ADR, and Forecast Accommodation Revenue.
Three bullets covering the overall pattern, strongest week, and weakest week.
GUARDRAILS
Do not add Budget, previous-period, or STLP comparisons unless requested. If AI Forecast is unavailable, say so and ask whether the user wants User Forecast; do not substitute it silently.
Try these tests:
/three-week-forecast for Flagship Berlin by Flyr/three-week-forecastfor a property with no AI Forecast, to confirm it asks before using User Forecast.
Example 3: Corporate-account review with a file
Use this pattern when Insights contains current production but the analysis also needs account terms or targets maintained outside Insights.
The attached file and review scope can change with each run, and unresolved matches can be clarified before the analysis continues. The file is available only in the current conversation and is not stored for future runs.
Where it fits: Chat. You attach the file when you run it, so it is a Skill you start yourself.
Name: Corporate-account review
Identifier: /corporate-account-review
Description: Compares a supplied corporate-account list with current production for one property and flags contribution, cancellation, and achieved-rate patterns.
Instructions:
PURPOSE
Review the accounts in the current reference file against the production data in Insights and the contract information supplied for this run.
INPUTS
Ask once for the property, review period, and current corporate-account file when any are missing. Read the file as supplied rather than assuming a fixed layout.
STEPS
Match account names against Company Name and list unresolved matches before analyzing them.
Compare monthly Units and Total Revenue with the equivalent completed stay-date period last year where available.
Calculate cancellation rate from booked and cancelled Units.
Compare achieved ADR with a supplied contracted rate as a directional check.
Where the file contains an annual commitment, show year-to-date attainment without presenting it as a final forecast.
Classify each account with enough matched information as Doing well, Watch, or Needs attention, and show the evidence for the classification.
OUTPUT
Start with the number of accounts reviewed and unresolved. Include only columns that can be populated from the file and Insights data. Finish with the strongest account, the account needing the closest review, and the most useful next investigation.
GUARDRAILS
Do not invent contract terms or merge distinct accounts without confirmation. Label unavailable comparisons rather than estimating them. The attached file is available only in the current conversation and is not stored inside the Skill.
Try these tests:
Attach a current test file and run
/corporate-account-reviewfor Flagship Austin by Flyr.Run it with an account name that does not match, to confirm it is listed as unresolved.
Example 4: Unmapped-value check
Use this Skill to identify classification gaps without attempting to change the underlying configuration.
The property, period, and affected field can change with each run, and you can continue investigating any value that needs attention.
Where it fits: Chat, to investigate a gap. If you want an email whenever something is unmapped, a conditional Routine can do that. See Routine Examples and Templates in Insights.
Name: Unmapped-value check
Identifier: /unmapped-value-check
Description: Finds blank or unmapped rate-code, channel, and segment values for a property and sizes their impact on recent production.
Instructions:
PURPOSE
Identify production that is difficult to classify and show which values account for the greatest amount in the available data.
INPUTS
Ask once for the property if it was not supplied. Default to the last 90 complete stay dates and accept a different period from the user.
STEPS
Review Rate Code, Channel Name, and Segment with Units and Accommodation Revenue.
Find blank or unmapped displayed values.
Group by the affected field and underlying value where available.
Rank issues by Accommodation Revenue.
Calculate each issue's share of property Units and Accommodation Revenue for the same period.
OUTPUT
Lead with the affected Units and Accommodation Revenue. Show a table with Affected field, Displayed value, Units, Accommodation Revenue, and Share of property revenue. Finish with the three highest-impact values to investigate.
GUARDRAILS
A blank or unmapped value is a classification finding, not proof that the booking is wrong. Do not invent the replacement or claim a root cause. Do not describe the list as exhaustive if the detailed result is incomplete.
Try these tests:
/unmapped-value-check for Flagship Berlin by FlyrRun for a period with no affected values and confirm it reports a clean result.
Example 5: Weekly commercial brief
Use this Skill when the same review happens every week: where you stand, what is driving it, what was booked recently, and what the rest of the year looks like.
The Skill takes its property from the request, so you can run it in Chat for any property, and a Routine can run it for one property every Monday.
Where it fits: Chat or a Routine. In a Routine, name the property in the Routine's Instructions (see Running a Skill in a Routine).
Name: Weekly commercial brief
Identifier: /weekly-commercial-brief
Description: Produces a concise property brief covering OTB position, pace, segment and channel movement, recent pickup, and the forward outlook.
Instructions:
PURPOSE
Create a consistent commercial brief that a revenue manager can review with a general manager. Focus on the evidence and on the areas that deserve attention, without prescribing a pricing or operational decision.
INPUTS
Use the property named in the request. If none is named, ask once. Default the main review to the current stay month and the forward view to the remainder of the calendar year. Use a different period if the request names one.
STEPS
Use the Bookings Topic to show OTB Units, ADR, and Accommodation Revenue.
Compare the same metrics with STLP and state whether pace is ahead, level, or behind.
Show the largest positive and negative Segment contributions.
Compare direct and indirect Channel performance on the same basis.
Use the Pick-up Topic for Units and Accommodation Revenue Pick-up over the last seven days up to now, including today so far.
Summarize OTB and STLP by Stay Month through year end.
OUTPUT
Lead with a one-sentence bottom line. Always include these sections: Current position, Segment and channel drivers, Recent pickup, Forward view. If a section has no data, keep its heading and say so. Use tables, not charts. Keep the result under 800 words. Name the dates used for pickup.
GUARDRAILS
Use the same property and date scope throughout. Do not replace STLP with final prior-year actuals. Name any missing comparison. Do not invent a cause or recommend a specific price. Disclose if any detailed result is sampled or incomplete.
Try these tests:
/weekly-commercial-brief for Serenity Resort Atlanta by Flyr/weekly-commercial-briefwith no property, to confirm it asks once.
Example 6: Monthly KPI review
Use this Skill at the start of each month to review the month that has just closed: headline results, who produced them, how the booking curve and forecast performed, and a short management summary.
Where it fits: Chat or a Routine. For a Routine, schedule it for the second or third day of the month, because late postings can still change figures on the first day.
Name: Monthly KPI review
Identifier: /monthly-kpi-review
Description: Reviews the previous full month for a property: headline KPIs, segment and rate-plan results, booking curve, forecast accuracy, and a short management summary.
Instructions:
PURPOSE
Give a revenue manager a consistent review of the previous full month that can be discussed with a general manager.
INPUTS
Use the property named in the request. If none is named, ask once. The review period is the previous full calendar month unless the request names another.
STEPS
Show Units, ADR, Accommodation Revenue, and RevPAR for the month, compared with the same month last year.
Show the same measures by Segment and by Rate Plan, ranked by Accommodation Revenue.
Describe the booking curve: how much of the month's Units were booked at each lead time, compared with last year where available.
Compare the month's actual Units and Accommodation Revenue with the Forecast and state the variance. If the Forecast is unavailable, say so.
Write a management summary of five to eight bullets, each tied to a figure shown above.
OUTPUT
Always include these sections: Headline KPIs, Segment, Rate plans, Booking curve, Forecast accuracy, and Management summary. If a section has no data, keep its heading and write "No data returned for this section." Use tables, not charts.
GUARDRAILS
Calculate variances in the query where possible and do not estimate figures that are not shown. Do not invent a cause or recommend a specific price. Disclose if any detailed result is sampled or incomplete.
Try these tests:
/monthly-kpi-review for Flagship Austin by FlyrRun it for a month the property has no data for, to confirm every section keeps its heading.
Example 7: Pickup check
Use this Skill for a quick, facts-only look at recent pickup for one property: what was picked up yesterday and over the last week for the current month and the next three, and which stay dates lost units.
Where it fits: Chat or a Routine. For a daily email, name the property in the Routine's Instructions. Use Run now a few times first to check the result.
Name: Pickup check
Identifier: /pickup-check
Description: Summarizes pickup for yesterday and the seven calendar days ending yesterday for the current stay month and the next three, and lists stay dates with net negative pickup.
Instructions:
PURPOSE
Give a fast, factual view of recent pickup for one property.
INPUTS
Use the property named in the request. If none is named, ask once. Use net negative pickup of 8 or more Units as the flag threshold unless the request names another.
DATA SCOPE
Topic: Pick-up
Update Date: yesterday, and the seven calendar days ending yesterday (if today is the 9th, the 2nd to the 8th)
Stay Date: the current stay month and the next three stay months
Metrics: Units Pick-up, ADR Pick-up, Accommodation Revenue Pick-up
STEPS
Show pickup by stay month for yesterday.
Show pickup by stay month for the seven calendar days ending yesterday (if today is the 9th, the 2nd to the 8th).
List the stay dates where net Units Pick-up over the seven calendar days ending yesterday (if today is the 9th, the 2nd to the 8th) is at or below the negative of the threshold.
OUTPUT
Always include these sections: Yesterday, Last seven days, Stay dates with net negative pickup. If there are no flagged dates, keep the heading and say "No stay dates met the threshold." Use tables, not charts. State the dates used.
GUARDRAILS
Report facts only. Do not give pricing advice or suggest a cause. Treat zero pickup as a valid result. Disclose if any detailed result is sampled or incomplete.
Try these tests:
/pickup-check for Flagship Austin by FlyrRun it with a threshold of 2 to see the flagged-dates table fill.
Running a Skill in a Routine
A Routine can run a Skill for you on a schedule or when a condition is met, and email you the written result. In the Routine's Instructions, type a forward slash followed by the Skill's identifier, then add anything the Skill needs. You can run a Skill you created or one that someone has shared with you.
For example, a Routine's Instructions could read: "Run /weekly-commercial-brief for Flagship Austin by Flyr, using the current stay month. Always include these sections: Current position, Segment and channel drivers, Recent pickup, Forward view. If a section has no data, say so."
A Skill runs well in a Routine when it is ready to run on its own:
It needs nothing at run time. Nobody is there to answer a question, so put the property and any other answer in the Routine's Instructions. Examples 5, 6, and 7 take their property from the request for this reason.
It asks for tables, not charts. A Routine emails a written analysis that can include headings, bullets, and tables, so a Skill that builds a chart, workbook, or dashboard will not deliver it by email.
It names the sections that must appear and says what to write when a section has no data.
It states the property exactly as it appears in Insights, and the currency if your properties report in more than one.
Test the Routine with Run now, and run it more than once. The figures should agree from run to run; if the sections differ, tighten the instructions and test again. When you edit a Skill, every Routine that uses it follows the change from its next run. For the steps, see Run a Skill Automatically with a Routine. For matching Routine wording, see Routine Examples and Templates in Insights.
Examples 1 to 4 begin with a different question on every run, so they work best in Chat.
Building a decision-support Skill
Some Skills are not reports. They guide a decision by asking for the context that changes on every run, for example evaluating the opportunity cost of a proposed long stay.
A decision-support Skill should:
Ask for the property, stay dates, inventory, units, and any proposed value needed for the comparison.
Define the evidence it will use, such as historical transient ADR, current Occupancy, Left to Sell, pickup, and relevant comparison data.
Separate observed figures from estimates and assumptions.
Present a neutral decision framework rather than inventing a recommended price.
Name missing evidence that lowers confidence.
Save this type of workflow as a Skill when each decision begins with case-specific inputs and may require follow-up questions. What you are saving is the decision method.
Testing and improving your version
Run at least four tests before relying on a new or changed Skill:
Normal case: supply the usual property, dates, and inputs.
Missing or ambiguous input: omit one required item and confirm that the Agent asks once rather than guessing.
No-data or missing-comparison case: check that the limitation is explained without inventing a result.
Largest intended scope: confirm that required detailed rows are not sampled, truncated, or only partially inspected.
If a Routine will run the Skill, add a fifth: use Run now at least twice and compare the emails.
Check the property, time fields, Topics, measures, comparison basis, output structure, and missing-data behavior. When something fails, add the smallest clarification that fixes the underlying ambiguity and rerun the tests.
For creation steps, field limits, ways to run a Skill, and current Skill behavior, see Skills in Insights.
Have product questions or see a result you did not expect? Select ? Help in the lower-left corner of the side navigation panel to contact our support team.