How to maximise your profits from AirBnB holiday rentals in Edinburgh?

What factors influence price ? A data-driven analysis

Photo by Clark Van Der Beken on Unsplash

I live in Edinburgh and for I while I’ve been considering purchasing a buy-to-let property in the city where every year (minus the Covid months/years) I observe thousands of tourists explore various parts of the city.

Investing in a buy-to-let is a serious enterprise and I want to do my research to make sure I maximise the returns on my investments. In particular, I want to answer the following questions:

  1. What affects price most?

To answer these questions, I turned to two AirBnB data sets:

  • Detailed listings data for Edinburgh: 6,536 listings with information about the neighbourhood where the property is located, the host (location, name, picture, etc), the property type, availability, number of bedrooms, amenities available, price etc.

Price is the variable I was mostly interested in, and in particular, I wanted to know what factors affect price most, i.e., what are the determinants of price. Therefore, I treated questions 1–4 as a regression analysis of estimating what feature-values have strongest relationship with the target variable — ‘price’. For the purpose, I built a machine learning model where the features were neighbourhood, number of bedrooms, property type, room type, amenities, property availability. Features about the host or reviews were not relevant for this exercise so I filtered them out.

I approached question 5 as an observational analysis.

Question 1: What features are the most important determinants of price?

Below you see the top 20 features that have biggest impact on predicting the price of an Airbnb rental. The number of bedrooms was of highest importance to the model, followed by the number of days for which a particular listing is available in a year, the maximum number of nights the property could be rented for, and next was the number of people the property could accommodate.

Top Features for Predicting Price

Interestingly, neighbourhood features didn’t bear much weight in the predictions, but let’s still focus on this feature alone.

Question 2: What neighbourhoods were strongest influencers of price?

Highest ranked neighbourhoods as price predictors

This one doesn’t suprise me — Dean Village (featured in the picture at the top) is a cute little area near the city centre where I mostly see tourists. Tollcross is mixed, and in fact I’m surprised that this even made second place.

Old Town, Princess Street and Leith Street cover a fairly large area in town and I would like to know which area from those three carries most weight — I suspect it’s Old Town but I would need to further explore the data to see whether a more fine-grained neighbourhood information is available.

Question 3: What amenities were strongest predictors of price?

Top Amenities for Predicting Price

The following one surprised me — I expected things like ‘TV’ and ‘wifi’ would make the top list but ‘fire extinguisher’ seemed to be of higest importance for predicting price. Keep in mind though the relative importances are fairly small so based on these findings alone I wouldn’t come up with conclusive observations as to what amenities to equip my property with.

Question 4: What property types affect price most?

Highest Ranked Property Types

Renting entire flats is one of the most popular options on AirBnB. Most flats are serviced in the sense that cleaning etc is taken care of by the host and in this respect the distinction between ‘entire apartment’ and ‘entire serviced apartment’ is not clear-cut.

Intuitively, I wasn’t sure whether renting private rooms or entire flats would have a stronger effect and the result above shows a clear answer.

Question 5: Is there any seasonality of holiday rental prices?

High seasons are generally the summer months, and lower seasons are usually winter, excluding the Christmas and New Years holidays. This question interests me from the following perspective — if there’s a particular month/ day of the week when prices significantly drop down, then I could reserve the property for use by family and friends without major financial loss.

To answer this question, I looked at the calendar of listings on Airbnb in Edinburgh.

I grouped the observations by month to obtain the following bar chart:

Price by Month

August is clearly the most expensive month with an average price of $89.785 followed by July with an average price of $88.3597 per night. This is the festival season attracting millions of tourists each year. Then we’ve got June and September with an average price of about $86. From this data, February is the lowest season with an average price of about $76.9. Surprisingly, December is on average as expensive to visit as is November.

Some of these numbers might be affected by Covid — for instance, are February and March representative of what we should expect in reality? We clearly need more than one year to see what the trend is.

Second, I looked at the trends by day of week.

Average Price by Day of Week

Friday and Saturday are unsuprisingly most expensive for holiday rentals but not by a huge margin — the average rental price is $86 compared to an average of $83 for Monday to Thursday and $84 for Sunday.


In this article, I used AirBnB listings and calendar data for Edinburgh to understand what factors influence price most and whether there’s such a thing as holiday rental seasonality.

Some key takeaways:

The number of bedrooms is the strongest predictor of price.

Looking at the feature ‘categories’ — neighbourhoods, amenities, and property types — we had Dean Village the highest ranked neighbourhood, fire extinguisher & first aid kit the highest ranked amenities, and entire serviced apartment the highest ranked property type feature.

Having said that, we should keep in mind the differences were fairly minimal.

Finally, we established that August and July are the most expensive months to visit Edinburgh, as are Fridays and Saturdays. This gives some clear indication when I’ll be busy looking after my guests :)

The next question is — do we see the same trends and same results when analysing other buy-to-let data? Do we have more years available for further analysis?

To see more about this analysis, see the link to my Github available here.

Data scientist, linguist