Reader TTS

A MODEL TO UNDERSTAND AND MEASURE UTILIZATION RATE IN A MARKETPLACE BUSINESS A few months back, my team and I were discussing an article posted by Julia Morrongiello, an investor at

Berlin-based PointNine Capital. Her article is a collection of summaries of other pieces addressing the need to identify a set of metrics to gauge marketplace liquidity. A marketplace lives

or dies on its liquidity but the way this is measured depends on the nature of the transactions it facilitates. Referencing an iconic piece by Josh Breinlinger (a must-read for any real

marketplace nerd), Julia’s article offers some practical examples of liquidity indicators and highlights how some might be more relevant than others according to the type of marketplace,

broadly defined as _Double-Commit_, _Buyers-picks_, _Marketplace-picks_, where the first has the lower liquidity and vice versa. INTRODUCING UTILIZATION RATE According to Josh’s article, he

highlights how Utilization Rate is crucial to_ Buyers-picks_ marketplaces. The principle behind the definition of Utilization Rate is determining “_what is the level of dependence of your

suppliers to your marketplace_”. Certainly, this notion is multifaceted and it requires a logical approach more than an analytical one. One of my favorite examples is how Josh argues that a

marketplace like Etsy should focus on the number of full-time sellers. I find it simply brilliant. As Etsy helps creatives handcrafted goods, the most important measure of their production

capacity and dependence on the platform is how much time they dedicate to it. Another example is offered by Bill Gurley (an early investor in Uber) when, referring to the cab-hailing

business, he said “Utilization is a measure of the percentage of time drivers are working versus waiting… As utilization rises, Uber can lower price, and the drivers still make the same

amount”. > _Utilization Rate aims at answering the question of how high is the > dependence of your suppliers to your marketplace._ That got us thinking. HousingAnywhere.com, where I

am VP Marketplace Operations, falls among the _Buyers-picks_ category, meaning the supply side has pre-committed its selling intentions according to a set of parameters. Specifically, vacant

calendar availability. The Utilization Rate in the hospitality sector can be defined as the number of days a property is rented out divided by the total number of days that property is

available for rent (its full capacity). It’s also often referred to as Occupancy Rate. All the articles that we read didn’t offer a sufficient level of detail on how to best define it. Most

refer to the hotel industry where occupancy is one of the most analyzed metrics. There is a fundamental difference between calculating the utilization rate as a marketplace vs. an operator

like a hotel, though: a marketplace doesn’t have full visibility on the full capacity of the operator. To add on that, availability on a marketplace can evolve for reasons beyond the mere

dynamics on one (your) platform and is thus a variable factor that can increase as well as decrease over time. For instance, operators might reduce it if they sourced bookings on different

channels or increase it as more of their inventory become available. These changes can be particularly complex for a mid-term rental marketplace where bookings are for several months.

Availability is also a complicated subject on its own, as what can be defined as available (or somehow bookable by the demand) depends on the timing of search as well as the demand’s

preferences. If you were to start a search on our platform now, you’d see a different number of listings for rent according to your intended move-in date. CALCULATING THE UTILIZATION RATE

FOR HOUSINGANYWHERE To calculate the utilization rate of listings on HousingAnywhere we need two data points: * The number of days a listing was rented * The number of days a listing was

made available on HousingAnywhere The number of days a listing was rented is easy to calculate. Each booking is linked to a move-in date and a move-out date and this can be used to.

calculate the number of days a listing was rented out. The number of days a listing was made available is more complicated. Two conditions must hold for a listing to be considered available:

* The listing must have been listed on HousingAnywhere prior to the date * An availability must have been created and been in place for tenants searching within reasonable time ahead of

that availability. Consider this example, where the availability is measured on different points in time. _NOTE: THIS EXAMPLE IS PURELY FICTIONAL AND IT’S USED EXCLUSIVELY FOR ILLUSTRATION

PURPOSES._ * There are three months where the status is unchanged, namely January, February, and October the listing was and remained _available_. * In the other seven months, the status

changes. In March to Septemeber, the status changes from _available_ to _unavailable_. In Nov and Dec the status changes from _available_ to _booked_. We need a way to resolve the ambiguity,

as each day needs to be considered either _booked_, _available_ or _unavailable_ to calculate utilization. * November and December are relatively obvious to resolve. The listing eventually

got booked and will be considered _booked_ on those days. * That leaves the days in Mar, Apr, May, Jun, Jul, Aug and Sep. They were initially _available_ and then made _unavailable_ on 19

January. Since our platform had an opportunity to make a match, we could argue that it should be considered _available_ but this depends on how we account for the demand preferences in terms

of the gap between the booking (or transaction) date and the intended move-in date. We call this the AVAILABILITY WINDOW. To determine the availability window, we need some statistics on

booking dynamics. How far ahead do tenants book, and how long do they book for? To avoid making this article too long, we will not explain how we determined these numbers but we will provide

you with a visual representation only and some number later in the article. What is important to know that we will determine the AVAILABILITY WINDOW BASED ON THE MOVE-IN DATE (i.e. the time

between booking date and move-in) and AVAILABILITY WINDOW BASED ON THE MOVE-OUT DATE (i.e. the time between the booking date and the move-out). The following image explains how the

availability window can impact our calculations. What’s the right way to interpret these strategies? * NO WINDOW assumes our platform is OPERATING IN ISOLATION AND HAS PERFECT CONTROL OVER

ITS INVENTORY. The assumption here is that what becomes unavailable couldn’t have been booked anyway so the _unavailable _days are not included in the calculations of the Occupancy Rate. *

MOVE-IN WINDOW and MOVE-OUT WINDOW account for a scenario where we assume that what becomes unavailable could have been booked and it’s now _unavailable_ because some other channels were

faster than us. So we consider _unavailable _as a missed opportunity. How big is this unfulfilled capacity depends on which demand perspective we choose. ACCOUNTING FOR MULTI-CHANNEL

OCCUPANCY As described above, occupancy is calculated as the number of days a listing was rented divided by the number of days a listing was available. THERE ARE TWO TYPES OF OCCUPANCY,

HousingAnywhere occupancy versus overall occupancy. The capacity is the same for both, but the number of bookings is different since HousingAnywhere occupancy only counts days booked on

HousingAnywhere whereas overall utilization counts the total. days booked. We can re-interpret the above availability strategies to take into account the difference between HA and overall

utilization as follows: * _Booked_: Days on which the listing was booked by HousingAnywhere. * _Likely booked elsewhere_: days on which the listing was made available on HousingAnywhere but

then was changed to unavailable ahead of that day. We are going to assume that the listing was booked somewhere else, either directly by the operator or on other channels. The length of the

period for this case depends on the strategy, i.e. using either average move-in or move-out dates. * _Vacant_: Days on which the listing was made available on HousingAnywhere and remained

available until that day passed. We consider that the listing remained unoccupied on those days. * _Unavailable_: Listings that had availability, but too far ahead to be considered bookable,

then became unavailable and never again available. The length again depends on the strategy of move-in versus move-out. We cannot make any assumptions in those cases. Using our previous

example, this is how it would look like. The final question we need to solve is what kind of capacity we are going to use at the denominator. There are two options: * PROVEN (I.E. FROM AN

INTERNAL POV): only takes _booked_, _likely booked elsewhere_ and _available_ into account, but discards _unavailable_. * FULL (I.E. FROM AN EXTERNAL POV): Takes all states into account,

thus it’s the full count of days of the period of analysis. We opted for the PROVEN CAPACITY as we are more interested in answering the question “how much occupancy did we fill given the

opportunity that we were given as a platform?” Given 2 assumptions for the AVAILABILITY WINDOW FOR BOTH MOVE-IN. AND MOVE-OUT DATES OF 66 AND 225 DAYS RESPECTIVELY, this table shows the

results of our methods. COMPARING THE RESULTS: MISSED CONVERSION VS. MISSED ACQUISITION What we find particularly interesting is the comparison of the occupancy rates, across the methods or

between HA vs. Overall Utilization. When we compare the methods, we can see how the difference between the NO-WINDOW method and either of the other two is an indicator of a MISSED CONVERSION

OPPORTUNITY, in the sense that HousingAnywhere was given the availability from the supply side, did not fulfill it, while the availability was fulfilled somewhere else. In the example

above, the MISSED CONVERSION OPPORTUNITY is: MOVE-IN METHOD: 40,4% — 25,5% = 14,9% MOVE-OUT METHOD: 40,4% — 18,2% = 22,2% On the other hand, when we compare the HA vs. Overal Utilization we

get an indicator of the MISSED ACQUISITION OPPORTUNITY, in the sense that the listing HousingAnywhere was never given that availability from the supply side so, even willingly, we couldn’t

fill it. In the example above, the MISSED ACQUISITION OPPORTUNITY is: MOVE-IN METHOD: 62,3% — 25,5% = 36,8% MOVE-OUT METHOD: 73,1% — 18,2% = 54,9% DOES IT REALLY MATTER? If you reached this

point in the article you might be wondering if all this effort is even worth it. Trust me when I say, we thought the same. While we started out with the ambition of finding an ultimate

metric to measure the occupancy of our inventory, what we found is that there is more value in the comparison of the numbers than in the numbers per se. While the NO-WINDOW method is clearly

not suited for internal use as it doesn’t account for the changes over time, it’s a powerful benchmark for the other two. There is little insight in believing that our occupancy is 25.5%

vs. 18.2%, but it’s very powerful to know that, according to what method you choose, we are missing out on either ~36.88% (14.9/40.4) or ~54.95% (22.2/40.4) of what WE COULD FILL. In that

sense, tracking and maximizing this metric is important to understand how much more we could fill by attracting more demand to the platform given the current situation. Likewise, we can

conclude that ~59% (36.8/62.3) to ~75% (54.9/73.1) represents how much market share we can still gain in terms of becoming the number one sourcing channel for our suppliers. _Many thanks to

__Mark Schiefelbein__ who did most of the heavy lifting in developing this framework and who is a constant source of intellectual curiosity._