Note: this site and the author does NOT have any financial relationship with the service provider being reviewed.
THREE POINT OVERVIEW:
In this week’s second review, I asked Shortcut to build out a granular revenue and cost forecast for Vail Resort Inc’s main operating segment.
Shortcut needs a lot of guidance on determining key revenue drivers, building out the structure of a granular revenue forecast, and modeling costs appropriately.
Even for a simple business like Vail, there are nuances in the business model that were not picked up by Shortcut for modeling purposes. This exercise was helpful in framing where analyst judgement will continue to be important if/when accuracy issues (some of which we identified in this week’s first review) are eliminated.
INTRODUCTION:
Over the past week I tested out Shortcut AI (link), an AI agent for Excel. This is the second review on Shortcut that I’m posting this week.
A quick recap - in my first write up on Shortcut (here), I assessed the tool’s ability to pull the current capital structure and historical financials for Vial Resorts Inc from its latest 10-K. As the first write-up showed, Shortcut was surprisingly good at pulling capital structure information but consistently made basic errors in pulling historical financials. On the bright side, I think that these issues should be fixable (they are not untraceable hallucinations). I will be testing Shortcut again soon and will update readers to the extent that there are improvements in these areas.
In this week’s second review, I wanted to test Shortcut’s ability to build a granular revenue and cost forecast if given a model that already includes historical financials. I gave Shortcut the historical segment financials for Vail’s “Mountain segment” which consists primarily of the Company’s ski resort/area operations. I also uploaded the Company’s most recent investor presentation which included additional KPIs not included in the 10-K and a press release that detailed management’s guidance for the next fiscal year. The goal was to test Shortcut’s ability to model the segment in the same detailed way that a buyside analyst would be expected to think about the business and build projections.
REVIEW OF RESULTS:
First, we uploaded an excel sheet with the below segment financials (Figure 1) and wrote a prompt asking Shortcut to use these financials along with any relevant information in the uploaded FY 2025 10-K and latest Investor Presentation to come up with a granular revenue and cost build for the segment. I know that this is a lot to expect from a very nascent tool - but when its promos are going viral with taglines suggesting imminent mass replacement of analysts, I think it’s fair to at least test it by assigning real work a buyside analyst might actually do.
Figure 1: Segment financials we uploaded to Shortcut
What We Were Looking For:
First, it is important to establish the output we were looking for - in other words, what is a “good” revenue build and forecast? Vail’s “Mountain Segment” is pretty simple at its core - revenue consists of visitors paying to ski on the company’s mountain properties (“Lift Revenue”) and visitors spending money on ancillary services (ski school, food and beverage, gear rentals), while costs are incurred to operate the mountain and provide the related service offerings. In its 10-K, the Company discloses the total number of skier visits per year as well as the “Effective Ticket Price” which is the average Lift Revenue generated per skier visit. So the revenue build should consist of Skier Visits * Effective Ticket Price, right?
Well, actually its a bit more nuanced than that. If you look through the 10-K and Investor Presentation to learn more about Vail’s business model, you will quickly see that “Effective Ticket Price” is an implied average price metric that doesn’t reflect the reality of how Vail generates revenue. Lift Revenue actually consists of two distinct sources: Pass Revenue and Lift Ticket Revenue. Passes are are sold in advance of the ski season for an unlimited or pre-set quantity of ski days, typically to more experienced skiers (with pass revenue recognized on a straight-line basis during the ski season). Lift Ticket Revenue, on the other hand, represents visits not booked in advance of the ski season and is more dependent on weather conditions (and more fickle skiers). Pass Revenue is about two thirds of total lift revenue and management discloses sales trends prior to the ski season beginning - so by the time the ski season begins (in November in North America) we already have an idea of what two-thirds of Lift Revenue will look like for the fiscal year. Therefore, it’s important to model the business correctly to take into account the significant “known” portion of revenue. So, Lift Revenue should be forecasted as (Average Pass Price * Passes Sold) + (Average Lift Ticket Price * Lift Tickets Sold).
The unit count for passes sold for each ski season is not included in the 10-K, and has to be pulled from the uploaded investor presentation. So we wanted to test Shortcut’s ability to recognize this and pull the required information.
On the cost side, a good analyst forecast would recognize that the vast majority (ie. ~80%) of costs in this business are fixed. As you may imagine, operating ski resorts involves a lot of costs that are fixed in nature (workers operating/maintaining the mountain, utilities/fuel, safety personnel, etc.) - of course, there are some variable items (retail/food costs, certain labor that can be flexed), but many of the costs are incurred regardless of how many skiers there are.
Along with the forecasts, we also wanted to test Shortcut’s ability to incorporate management guidance into projections. For example, management is in the middle of executing a cost efficiency plan so we wanted to see if shortcut could read the guidance from the uploaded PDF press release and incorporate the cost savings into its forecast.
Results:
Modeling Revenue: For the revenue forecast, I asked Shortcut to create revenue forecast that is “detailed and granular, breaking out the key unit drivers of revenue” (full prompt in Appendix). Below is a summary of the results - I think the (TLDR) summary is that Shortcut needed a lot of hand holding through the revenue build - this is not necessarily a knock on Shortcut, but just reflects the reality of why analyst discretion/input would be needed even if Shortcut wasn’t making errors pulling numbers:
The first result from Shortcut was a forecast that projected lift revenue as Skier Visits * Effective Ticket Price. As explained in the prior section, this formula does not reflect the reality of the business and is not the best way to build the forecast.
I then instructed Shortcut to use data from the uploaded materials to separate pass revenue and lift ticket revenue drivers. Shortcut was able to pull the relevant stats from the Investor Presentation to do these calculations without me specifying the pages or specific data sets (pass revenue, pass units, and % of total skier visits attributed to lift tickets) - so Shortcut recognized what it needed to look for to forecast pass and lift revenue on a Price * Volume basis. However, we should note that once again, the data retrieval resulted in an error that mixed up numbers between FY 2024 and FY 2025 (Figure 2).
Once shortcut forecasted pass and lift ticket revenue, the projections were not linked correctly into the total revenue build, so this needed to be corrected.
The pass and lift unit sales need to be translated into a total ski visitor forecast (which drives ancillary revenue which is forecasted on a per total visitor basis) - Shortcut did not recognize that it needed to assume visits per pass holder and add pass holder visits to lift ticket unit sales, instead just doing an annual growth forecast independent from the pass and lift forecasts. So we needed to prompt Shortcut to fix this.
In the end, Shortcut was able to produce a usable dynamic model - but this back and forth of reviewing what Shortcut did and providing follow up prompts took longer than it would have taken me to build the forecast myself.
Figure 2: Accuracy of number retrieval remained an issue (mixed up FY 2024 and FY 2025 figures)
Modeling Costs: Labor costs are the biggest component of Vail’s Mountain segment expenses. In general, Vail’s labor costs have a significant fixed component – this makes logical sense as a lot of labor related to operation of the mountain (ski lift attendants, safety personnel, mountain grooming and snowmaking, etc.) need to be in place regardless of the number of visitors when the ski hills are open. In FY 2024, when skier visits declined 10% Y/Y as a result of abnormally low snowfall, labor and labor-related benefits expenses declined only 2% Y/Y – this suggests that as much as ~80% of labor costs might be fixed in nature. However, Shortcut categorized the entire labor expense category as a variable expense – when I asked why, one of the main reasons was that the 10-K mentions the Company having a “seasonal workforce”. Technically, this is true and expenses are technically “variable” in that sense – Vail employs ~7k year-round employees and ~40k seasonal employees (who work in the winter). However, during the period when most mountain revenue is actually earned, past cost trends show labor costs are in large part fixed. A good analyst would at least try to estimate a fixed/variable split for large cost items such as Labor - as a placeholder subject to future diligence, I would probably go with an 80/20 fixed/variable split given FY 2024 results - but Shortcut won’t really attempt this until prompted on these details. Shortcut’s misinterpretation of the nature of costs would result in the model overstating Vail’s cost flexibility, for example, we were to model skier visitation being down 10% next year.
Incorporating Management Guidance Into the Model: Once Shortcut’s initial forecast was built out, I uploaded a press release that included Vail’s FY 2026 guidance and asked Shortcut to incorporate management’s guidance into its FY 2026 forecast. Part of management’s guidance includes making progress on achieving $100M in run-rate cost savings from efficiency initiatives. The Company achieved $37M in FY 2025 and expects to achieve $75M of the original $100M target in FY 2026 – this $75M represents $38M in incremental savings over the $37M already achieved. However, Shortcut missed the part where this $75M guidance was net of the already-achieved costs in FY 2025 and incorporated the entire $75M as incremental cost savings in FY 2026.
End Result: The end result is shown below (Figure 3) - it’s a usable and dynamic segment profitability model with sufficiently detailed revenue and cost projections. But it took a while to get there - and further work is required to fix some of the flawed cost assumptions.
Figure 3: End result was a working dynamic model (with more work needed to fix flawed cost assumptions)
KEY TAKEAWAYS:
Shortcut can do a decent job at building out forecasts if guided through the step-by-step model building process with detailed prompts. However, Shortcut does not automatically thoroughly investigate the business model and isolate the key drivers of the business, so this required additional follow up prompts. Shortcut also shows limitations in interpreting descriptions of costs when making fixed/variable cost determinations, and also made some mistakes interpreting certain nuances of management guidance. These misinterpretations can translate into bad model forecasts (more on this shortly).
Overall, once the “number pulling” accuracy and formatting variability issues we talked about in our first review are fixed (which I think is the first and most important hurdle right now), I think Shortcut in its current form will be good enough to sketch out a first draft of a model, perhaps for a “quick screen” stage of an idea in a professional setting. Analysts would then be able to shift more of their time to building out key revenue/cost drivers and determining appropriate forecasts to complete a model that would be acceptable for use in “investment committee level” materials.
As the technology progresses, perhaps it can reach a stage where if prompted for a granular forecast, Shortcut (i) conducts a more in-depth evaluation of the business model and (ii) gathers all the disclosed supplementary KPIs that are available in the reported materials. And only after fully investigating the business as per (i) and gathering available data as per (ii), Shortcut then determines the most effective way to break down the forecast into its key drivers. Not sure if this is already how Shortcut tries to handle these types of prompts (if so, there is certainly room for improvement there). Improvements in building out granular and accurate revenue and cost builds would really allow analysts to shift even more time from model building towards determining better forecast assumptions for the key variables driving the business (which is the most valuable and intellectually difficult part of the modeling!).
APPENDIX:
Prompts Used for Segment Forecast Build:
Initial Prompt: Use the uploaded template to complete a forecast for the Vail Resorts Inc. Mountain Segment. Use information and data from the FY 2025 10-K and FY25 MTN Investors Conference Presentation to come up with a 5-year forecast. The forecast should be detailed and granular, breaking out the key unit drivers of revenue. Costs should be forecasted as a percentage of revenue if variable or increase at an annual inflation rate of 2% if fixed. Show all forecast assumptions on this tab. The forecast assumptions for revenue and costs should be shown on an annual basis throughout the forecast period. Assume key revenue and cost variables show the same trends in each year of the forecast period as in FY 2025. Make sure all formatting such as font sizes and column widths are consistent with the template.
Follow Up Prompt #1: Use data from the uploaded materials to separate pass revenue and lift ticket revenue drivers and include them in the forecast. Forecast all non-lift revenue on a per visit basis.
Follow Up Prompt #2: In the top section of the segment financials building to total revenue, make sure pass revenue is using the forecast calculated using pass units sold multiplied by average pass price, instead of as a percentage of total lift revenue.
Follow Up Prompt #3: Forecast lift ticket visits independent of total visits. Also calculate pass visits assuming the ratio of visits per pass holder annually is constant instead of calculating it based on a percentage of total skier visits.