Five years after its launch, the Autolib’ oneway carsharing service has not yet reached financial balance. Judged by the number of subscribers per year, the profitability horizon of the service has been repeatedly postponed by Bolloré Group.
How can we explain the fact that financial balance is more difficult to achieve than expected, while the number of subscribers per year continues to grow steadily? Using data made available to the public by the Autolib’ Métropole joint association, we show that the more the number of subscribers per car in service increases, the more the frequency of use of the service by subscribers decreases. The threshold effect thus identified appears to be a real “glass ceiling” that could prevent Autolib’ from achieving financial balance.
In conclusion, we propose options for increasing profitability of the service in Paris. On the other hand, if achievement of financial balance is difficult in the Paris area, it becomes even more uncertain for offshoots of Autolib’ in other cities where the service.
Our article on the profitability of the Autolib’ service published on December 15, 2016 attracted the attention of the French media. In order to improve our contribution to the debate, we are updating this article by integrating the following modifications (detailed in appendix):
 Change in the methodology to calculate the number of trips made in 2014 and 2015
 Change in the methodology to estimate the number of cars in service
 Updating the analyzed data series, integrating the new data set for December 2016
These changes confirm the conclusions of the article: the more the number of subscribers per car in service increases, the more the frequency of use of the service by subscribers decreases. While the Autolib’ managers present profitability results expressed in terms of subscribers, the results of our study show that the increase in the number of subscribers alone can not ensure the permanence of the service .
In order to continue to contribute to the reflection on Autolib’ business model, we will start publishing a monthly barometer on the evolution of Autolib’s uses.
Introduction
Five years after its launch, the Autolib’ service is the largest oneway carsharing service with stations in the world[1], in terms of number of cars in service and number of subscribers.
However, data for establishing the economic viability of the carsharing scheme proposed by the Bolloré group are still lacking. According to press reports, the profitability horizon of Autolib’ has been postponed several times. A year after the service was launched in November 2013, Vincent Bolloré declared that 50,000 subscribers per year would be required to repay “annual expenses”[2]. However, in early 2014, the Bolloré group estimated that 60,000 and no longer 50,000 subscribers would be needed to achieve financial balance[3]. At the beginning of 2015, Vincent Bolloré announced that the service would be profitable with 82,000 subscribers[4]. On 27 November 2016, the service had 131,328 subscribers per year, and on the occasion of the 5^{th} anniversary of Autolib’ on 5 December 2016, service managers announced they were expecting financial balance to be achieved by 2017 or 2018[5]. How can it be explained that, despite the steady increase in number of subscribers, the service is struggling to achieve this balance?
The question of the economic model of Autolib’ is all the more pressing given that oneway carsharing services with stations, operating on a similar model, have already been deployed by the Bolloré group in France, in Lyon and Bordeaux, as well as abroad, in Indianapolis (United States) and Turin (Italy), and other stations are under way (London) or planned (Singapore)[6].
This article provides an analysis of data published by the Autolib’ Métropole joint association, responsible for control of delegation of the Autolib’ public service and territorial deployment. We present an estimate of the annual revenue of Autolib’ over the last three years, and then we compare this estimate with usage data to try to explain the trends observed. The data and methods used are described in an annex at the end of article.
Overall revenues increasing but income per car slowing down
Autolib’ obtains revenue through two channels: subscriptions and journeys made by subscribers. Using monitoring data published on a weekly basis and progress reports published annually by Autolib’ Métropole, we constructed estimates of revenue related to subscriptions and journeys for the years 2014, 2015 and 2016.
The available data allow us to compute the revenues related to the annual subscriptions (also called “1 year”) and the journeys made by the subscribers annual plan, but not to take into account the other subscription plans and the trips that are related to them. However, Autolib’ Metropole’s 2014 activity report indicates that “1year subscribers represent 99% of all active subscribers and account for 96.8% of trips”; The 2015 activity report indicates that “in December 2015, “1 Year” subscribers accounted for almost 99.5% of all active subscribers and made 96.7% of trips”. Thus, it appears that the annual subscribers represent all or almost of the subscribers and journeys made, in a stable way over time. Therefore, the study of the population of”1 year” subscribers and their use of Autolib allows to obtain a sufficiently precise idea of the profitability of the service. The complete method is detailed in the annex.
According to our analysis, service revenues increased by 33% between 2014 and 2015, and by 25% between 2015 and 2016. The breakdown of revenues shows an increase in the share of subscriptions in revenue : 23% in 2014, 26% in 2015 and 27% in 2016. The relative share of revenue related to journeys has thus decreased[7]. This is due to the strong increase in annual subscriptions sold (over 30,000 more subscriptions from one year to the other), but also to the slowdown in the growth of the number of journeys made in 2016, compared with the previous year. The number of journeys made annually increased by 12% between 2015 and 2016, versus 29% between 2014 and 2015.The maintenance of journeyrelated revenue between 2015 and 2016 (+ 22%) is therefore almost exclusively due to the 9% increase in rates on 1 February 2016[8]. According to our calculations, if the price of journeys made with the “1 year” subscription had not increased, service revenues would not have increased by 25% but only by 17% in 2016. Higher overall revenues hides a decline in usage growth.
The question is thus raised of Autolib’s ability to reach financial balance if the decrease in the number of journeys made by subscribers continues in 2017.
In addition, in parallel with revenue, the number of cars increased (2 645 in 2014, 3 309 in 2015 and 3 923 in 2016) implying an increase in the operating cost of the service. If we compare annual turnover with the number of cars in service at year end, revenue per car increased by 6% between 2014 and 2015 and again by 5% between 2015 and 2016. And if the tariffs had not been increased at the beginning of 2016, the turnover per car would have decreased by 1% in 2016. Going into more details, we highlighted that the number of trips per car has decreased by 5,6% between 2015 and 2016. Thus, growth in revenue is primarily extensive (increasing the number of stations and cars is likely to attract new subscribers) and not very intensive (there will be very little increase in revenue generated per car).
The slowdown in growth in the number of journeys, as noted above, is probably the main obstacle for the Autolib service to achieve profitability. It is therefore necessary to understand the reasons for this slowdown.
The higher the number of subscribers per car, the less users use the service frequently
How can it be explained that the number of journeys made in 2016 is slowing down while the number of subscribers increased? Is this a cyclical variation or is it related to the structure of the offer?
Autolib’ Métropole data suggests that the reduction in the number of journeys is linked to a structural threshold effect, linked to the reduction in the availability of cars. Over the period studied, from June 2014 to November 2016, the number of annual subscribers[10] per car in service increased steadily. In December 2016, the service had 34 “1 year” subscribers per car, compared with 22 in June 2014. The number of annual subscribers per car increased by 50% in two and a half years. Over the same period, the average journey time remained unchanged at 38 minutes[11]. If the frequency with which users use the service had remained the same, the 50% increase in the number of users per car should therefore have resulted in a 50% increase in the revenue generated by each car. Actually, this increase would have been greater than 50% given that rental rates were increased by 9% in February 2016.
Fig. 3 : Number of “oneyear subscribers” to the Autolib service per car, by month
Yet over the same period, the frequency with which users used the service decreased significantly. While each “1 year” subscriber averaged 1.4 journeys a week with Autolib’ between September and December 2014, each “1 year” subscriber made about 0.8 journeys per week with Autolib’ between September and December 2016. After a two year gap and over a comparable period, users on average almost halved their frequency of use of the service.
Fig. 4 Number of weekly Autolib rentals by “oneyear” subscriber and by week, by month
In other words, the 50% increase (multiplication by 1.5) in the number of subscribers per car over the period studied was more than offset by a 50% decrease (division by 2) in frequency of use of the service. As a result, the number of journeys per car per day is stagnating, and even seemed to be starting to decrease by the end of 2016: in December 2014, each car on the service averaged 4.5 journeys per day; in December 2015, 4.4; in December 2016, 4,1. The trend curve represented by dots in Figure 5 shows that over the period studied, the number of rentals per car per day is tending to decline.
Fig.5 Average number of Autolib rentals by car and by day, by month
A “glass ceiling” that questions the economic model of Autolib’ and its offshoots
The first survey on Autolib’ service carried out by the 6t in November 2013 and January 2014 showed that the success of Autolib’ in Paris was largely due to the density and availability of vehicles, which encouraged intensive use of the service: with users assured of the likelihood of always finding an available car nearby, they developed an “Autolib’ reflex”. If vehicle availability decreases, users may become discouraged after several fruitless searches for vehicles and lose their “Autolib’ reflex”. This is what appears to be happening, according to analysis of Autolib’ Métropole data: the more the number of subscribers per car increases, the more the number of rentals per subscriber decreases.
If this trend persists, it could become a true “glass ceiling” preventing the service from achieving financial balance. Vincent Bolloré himself has already described the Autolib’ service as “ruinous” [13]. So what are the options if one does not want to see Autolib’ cars disappear from Parisian streets?
 Like a patron, the Bolloré group could subsidise the “shared automobility” of the Paris area: difficult to believe.
 Affixing advertising on cars would diversify sources of revenue: even if the approach has been tested on 10% of the fleet since October 2016, its generalisation and sustainability will prove complicated to legitimate, not least given the law which prohibits the parking on public roads of vehicles used as a medium for advertising[14].
 The local authority could contribute to covering the deficits: thus financing through local taxes.
 Finally, the user could pay more for the service: this is perhaps the most equitable solution when it is known that Autolib’ users have incomes 27% higher than the average of Parisian households[15] and that one kilometre using Autolib’ costs them an average of €1.13[16], against €2.40 for a kilometre by taxi[17]. It appears to be one of the solutions chosen: in 2017, it is expected that the rental rate for “1 year” subscribers will increase from 6 to 7 € per half hour.
Finally, if the economic model of Autolib’ is problematic in Paris, its offshoots in less densely populated cities are even more so. In effect, the density of Paris seems particularly conducive to direct track car sharing: population density implies a greater number of potential users in the vicinity of each station, which makes it possible to provide a large number of stations, thus a large number of potential journeys (each car needing to be returned to a station), which in turn encourages a “reflex” use of Autolib’. This virtuous circle seems difficult to reproduce in less densely populated cities, where the potential public of each station and number of stations, and thus feasible journeys, are lower . For example, three years after its launch, the Bluely service in Lyon recorded less than one rental per car per day in 2016[18] against 4,1 in Paris in December 2016. The difficulty in achieving financial balance in Paris therefore questions the economic model of all car sharing systems proposed by the Bolloré Group.
Hopefully Autolib’ will not suffer the fate of another French innovation, the Concorde: a political success, a technological success but an economic failure.
Methodological annex
For the drafting of this article, we relied primarily on two data sources:
 Autolib’ Métropole activity reports, which provide many usage data on the Autolib’ service. They can be downloaded at http://www.autolibmetropole.fr/telechargements/rapportsactivite/ (accessed on 01.19.2017, content in French);
 “Open data” files (which have been provided since June 2014 and for each Sunday), data on the number of cars in service, their unavailability rate, the number of stations and charging terminals in service, the number of journeys carried out over different durations, as well as the number of active subscriptions (“1 year” and overall number).

Estimated revenue for the Autolib’ service over the period 20142016 (Figure 1)
Calculation of the estimate uses the following variables (Update version 01/20/2017)
In the first version of the article, we calculated the number of journeys made in 2014 and 2015 by multiplying the average number of journeys made for each type of subscription by the number of subscribed plan, based on the data from Autolib’ Métropole activities reports of 2014 and 2015. However, the number of average journeys per subscription for 2015, as indicated in the Autolib’ Métropole activity report, concerns only the month of December 2015 and not the whole of the year. December being a month when the use of the service is stronger than on the whole year, this method led us to overestimate the number of journeys in 2015, and by the same to conclude erroneously to a decrease of number of journeys between 2015 and 2016. In this update, we use a method of substitution, which makes it possible to compute precisely the number of trips made, no longer by based on subscribers, but based on annual subscribers . However, since “1 year” subscribers make almost all trips, this method seems to us the most relevant and the most accurate, based on available data, to estimate the revenue of Autolib’ over the period of 2014 and 2016. The available data does not help to compute precisely the number of journeys made by type of subscription over the period analyzed. As a result, we restrict our analysis to the actual subscriptions and the trips made by subscribers of the “1 year” plan. However, Autolib ‘Metropole’s 2014 activity report indicates that “1year subscribers represent 99% of all active subscribers and account for 96.8% of trips”; The 2015 activity report indicates that “in December 2015, “1 year” subscribers accounted for almost 99.5% of all active subscribers and made 96.7% of trips”. It thus appears that these subscribers represent almost all subscriptions subscribed and journeys made, in a stable way over time. The study of the population of annual subscribers and their use of Autolib ‘thus allows to obtain a sufficiently precise idea of the profitability of the service.The calculation of the estimate uses the following variables:
 The number of “1 year” subscriptions sold: Autolib ‘Métropole’s 2015 activity report indicates the number of “1 year” subscriptions sold in 2014 and 2015. Observing via the crosschecking of Autolib’ Métropole and the number of active “1 year” subscriptions “at the end of the year is very nearly identical to the number of”1 year” subscriptions sold in a given year, therefore, we considered that the number of “1 year” subscriptions sold in 2016 is equal to the number of “1 year” active subscriptions at the end of December 2016. Those information are available via Autolib’ Métropole’s open data files.
 The price of each type of subscription, found by consultation of the autolib.eu site on 25.11.2014 and 12.12.2016 and invoices over the period studied. Concerning “1 year” subscriptions, it should be noted that the price used in this article (€ 120 per year or €10 per month) is overestimated, insofar as Autolib’ regularly offers promotions, both on the price of first subscriptions and renewals.
 Number of journeys made according to duration of the subscription taken out: for the years 2014 and 2015, the number of journeys made on average for each subscription type is indicated in the Autolib’ Métropole activity report. For 2016, the number of trips made each month is available in Autolib’Métropole’s open data files. We therefore added up the number indicated for each month in order to obtain an annual total and then multiplied it by the share of trips carried out by the “1 year” subscribers in December 2015, according to theAutolib’ Métropole2015 activity report. Indeed, the share of trips made by “1 year” subscribers is almost identical (0.1%) over the year 2014 and during the month of December 2015 (sources: 2014 and 2015 Autolib ‘Métropole activity reports). We therefore assume that this share is a constant and use the known data for December 2015 on the figures for 2016.
 Average journey time: according to the Autolib’ Métropole 2013 activity report, the average journey time was 40 minutes in 2012 and 38 minutes in 2013. Over regular visits to the http://www.autolibmetropole.fr/leserviceautolib/leschiffresen1clic/ page, the average journey time was 38 minutes between October 2014 and September 2015. At 12 December 2016, we note that the duration reported on this page is 37 minutes. In the absence of knowing when the change was made to the site, we consider the average journey time as 38 minutes throughout the period studied. Lacking more accurate data, we consider that journeys have an average duration of 38 minutes, regardless of the type of subscription taken out.
 The hourly rate for each type of subscription, resulting from consultation of the autolib.eu site on 25.11.2014 and 12.12.2016.
The calculations made for estimating revenues are :
 revenues from subscriptions: multiplication of the number of subscriptions of each type sold per year by their respective prices;
 revenue from journeys: multiplication of the estimated number of journeys made with each type of subscription by the number of subscriptions of each type taken out during the year.

Annual revenue per car (Figure 2) – Update version 01/20/2017 :
To calculate the ratio of annual revenue per car, we no longer rely on the number of cars in service at the end of the year, but on an estimation of the average number of cars in service over the year, according to Autolib’ Metropole open data. This new method makes it possible to estimate more accurately the annual revenue per car.
Our survey of Autolib’ Métropole’s open data files allows us to know for each end of the month, over the period studied, the number of cars in service. By averaging these values for each year, we obtain an average number of cars in service each year (the rate of deployment of cars seems to make the scale of the month sufficiently precise for this type of calculation). We obtain by this method an average number of cars in service which amounts to:
• 2,649 cars on average during the period JuneDecember 2014. The earliest available data date from June 2014. The number of cars deployed was probably lower before June 2014 than after June 2014. However, in the absence of data available for the period from January to May 2014, the average figure for the whole year is 2,649 cars. This probably leads us to overestimate the average number of cars in 2014 and thus to underestimate the annual revenue per car in 2014. Our estimate of the growth in turnover per car between 2014 and 2015 is therefore likely to be higher To reality.
• 3,309 cars on average over the period JanuaryDecember 2015.
• 3,923 cars on average over the period JanuaryDecember 2016.
Then, we compute for each year:
• Revenues from car subscriptions, by dividing the total subscriberrelated revenues, previously calculated (see section 1 of this annex), by the average number of cars available;
• Income from journeys outside the tariff increase, taking the preliminary calculation of travel receipts (see section 1 of this annex) and introducing an alternative calculation for 2016 in which the hourly rate remains fixed at € 11. Then by dividing the amount, we obtained by the average number of cars available;
• the surplus revenue resulting from the rate increase in 2016, by subtracting from the estimate of receipts with an increase in the hourly rate the estimate of receipts without an increase in the hourly rate and then dividing the amount thus obtained by the average number of available cars

Number of “1 year” subscribers per car, by month (Figure 3)
Update version 01/20/2017 :
Integrating the data up to December 2016 to the analyzed data files
Regarding the car unavailability rate, that is to say the proportion of cars not in service, including for repair, Autolib’ Métropole open data only indicates the rate for Sundays. It is difficult to calculate an unavailability rate over a month on the basis of partial daily data. Considering that the unavailability rate of vehicles, for days when it is known, is often around 10%, we decided to neutralise this variable and not take it into account.
To calculate the number of active “1 year” subscribers by car, we used the following data:
 the number of active subscribers at the date nearest to the end of each month in Autolib’ Métropole open data;
 the number of cars in service at the date nearest to the end of each month in Autolib’ Métropole open data.

Number of rentals per “1 year” subscriber and week, by month (Figure 4)
Update version 01/20/2017 :
Integrating the data up to December 2016 to the analyzed data files
On the choice of “1 year” subscribers rather than all subscribers, see section 3 of this annex. To calculate the number of rentals per “1 year” subscriber and per week, we used the following data:
 the number of active “1 year” subscribers at the date nearest to the end of each month in Autolib’ Métropole open data;
 the average number of rentals per day on the same month, indicated in Autolib’ Métropole open data files.
We multiplied the number of rentals per day by 7, then by 0.967, in order to reflect the 96.7% of rentals made by the “1 year” subscribers (see section 3 of this annex), then we divided the result by the number of active “1 year” subscribers.

Average number of rentals per car and per day, by month (Figure 5)
Update version 01/20/2017 :
Integrating the data up to December 2016 to the analyzed data files
To calculate the number of rentals per car and per day, we used the following data:
 the average number of rentals per day on the same month, indicated in Autolib’ Métropole open data files.
 the number of cars in service at the date nearest to the end of each month in the open data files
We divided the average number of rentals per day by the number of cars available.

Calculation of cost per kilometre of a journey with Autolib’ in 2016
We wanted to calculate a cost per kilometre that takes into account the hourly rate and the subscription price. To this end, we decided to split the annual subscription cost (€120) over all journeys made in 2016.
We do not know the average number of journeys per subscriber in 2016 but, on the basis of Autolib’ Métropole data, we calculated the average number of journeys per “1 year” subscriber for each month from January to November 2016. By calculating the average of these averages, we obtain an approximation of the average number of journeys per “1 year” subscriber and per week in 2016, which stood at 0.92. We therefore estimate that “1 year” subscribers made an average of 48 journeys over the year (that is, 52 weeks multiplied by 0.92). Spread over these 48 journeys, the subscription cost is €2.5 per journey.
Given that the average journey time is estimated at 38 minutes (see paragraph 1 of this annex) and the hourly rate per “1 year” subscriber is €12 per hour, the average journey cost can be assessed as €7.60.
By adding up the costs related to subscription and the hourly rate, a journey therefore amounts to €10.60 on average for a “1 year” subscriber. According to the 2016 activity report, the average distance travelled per journey is 9.3 km on average. Given the stability of the average duration of journeys over time (see paragraph 1 of this annex), the average journey distance should also be maintained in 2016. We can estimate the average cost of a journey using Autolib’ at €1.13 per km.
[1]With Autolib’ and similar services developed by the Bolloré group, the user can return the car they used to a station other than the one they departed from. These similar services are distinguished in this from loop car sharing, where the user borrows the car and returns it to the same station, and direct track car sharing without a station (or “freefloating”), where the user borrows the car and returns it not to a station but within a given perimeter of a city.
[2]ALBERT Laurence, “Autolib’ celebrates its first anniversary and aims at reaching financial balance by spring 2014”, Les Echos, 30.11.2012. URL: http://www.lesechos.fr/30/11/2012/LesEchos/21324107ECH_autolib–fetesonpremieranniversaireetviselequilibreauprintemps2014.htm (accessed on 12.12.2016)
[3]“The Autolib’ package sold off”, Le Parisien, 04.02.2014. URL: http://www.leparisien.fr/espacepremium/paris75/leforfaitautolibbrade040220143556359.php (accessed on 12.12.2016)
[4]DESAVIE Patrick, “Autolib’ gains ground and is copied”, L’Usine Nouvelle, 28/01/2015. http://www.usinenouvelle.com/article/autolibgagneduterrainetfaitdesemules.N310310
[5]“A still not profitable success: the fiveyear balance of Autolib’”, France 3 ÎledeFrance, 05.12.2016. URL: http://france3regions.francetvinfo.fr/parisiledefrance/paris/succestoujourspasrentablebilan5ansautolib1148395.html (accessed on 12.12.2016)
[6] DESAVIE Patrick, “320,000 subscribers and 165 million kilometres travelled in five years for Autolib’”, L’Usine Nouvelle, 05.12.2016. URL: http://www.usinenouvelle.com/article/320000abonneset165millionsdekilometresparcourusencinqanspourautolib.N472488 (accessed on 12.12.2016)
[7]The share of subscriptions in total revenue is overestimated because reductions in the price of subscriptions are offered regularly, for both initial subscriptions and subscription renewal. However, given that these promotions were offered regularly over the period studied, the growth of the relative share of subscriptions in total revenue from one year to the next remains valid.
[8] In February 2016, the price of a journey made with a “1 year” subscription increased from € 5.5 to 6 per half hour, up 9%. Timelimited subscriptions have been replaced by a “readytoride” subscription, which is not limited in time and whose rates are those of the former “1 day” subscription: no subscription fee but an hourly rate of €9.
[9]“A still not profitable success: the fiveyear balance of Autolib’”, France 3 ÎledeFrance, 05.12.2016. URL: http://france3regions.francetvinfo.fr/parisiledefrance/paris/succestoujourspasrentablebilan5ansautolib1148395.html (accessed on 12.12.2016)
[10] Until January 2015, Autolib’ offered timelimited subscriptions in addition to the “1 year” subscription. The data available about these subscriptions do not allow us to include them in the calculation of ratios (for a detailed explanation, see the methodological annex). We therefore only take into account 1 year subscriptions in the ratios. However, according to the Autolib’ Métropole 2015 activity report, “In December 2015, “1 Year” subscribers accounted for nearly 99.5% of all active subscribers and 96.7% of rentals”. We therefore postulate that the failure to take account of other subscriptions in the ratios presents only an extremely minor bias.
[11]According to the Autolib’ Métropole 2013 activity report, the average journey time was 40 minutes in 2012 and 38 minutes in 2013. Over regular visits to the http://www.autolibmetropole.fr/leserviceautolib/leschiffresen1clic/ page, between October 2014 and September 2015, the average journey time was 38 minutes. At 12 December 2016, we note that the duration reported on this page is 37 minutes. In the absence of knowing when the change was made to the site, we considered the average journey time as 38 minutes for the entire period studied.
[12] 6t, Oneway car sharing survey (ENA.3) – Oneway carsharing:what alternative is it to the private car?; ADEME, 2014. URL: http://www.ademe.fr/autopartagetracedirectealternativeavoitureparticulierel (accessed on 13.12.2016)
[13]Razemon Olivier, “What really is the use of an Autolib’? “, 10.12.2014. URL: http://transports.blog.lemonde.fr/2014/12/10/aquoisertvraimentuneautolib (accessed on 13.12.2016)
[14]Article R58148 of the Environmental Code, accessed on 13 December 2016, on the https://www.legifrance.gouv.fr/site
[15]Comparison of the average disposable household income of Autolib’ users and Parisian households. Source: 6t, ibid.
[16]The calculation is detailed in the methodological annex.
[17]6t, Working conditions for Parisian taxis, 2016. URL: https://6t.co/conditionsdetravailtaxisparisiens/ (accessed on 13.12.2016).
[18]SAMARD Francis Bluely, car sharing service, continues its momentum, 19.10.2016. URL: http://www.leprogres.fr/lyon/2016/10/19/bluelyservicedautopartagecontinuesursalancee (accessed on 13.12.2016). According to this article, the service has 302 cars at the end of 2016 and the number of rentals for 2016 were expected to be around 100,000. The number of rentals per car per day in the year 2016 should be about 0.9 (100,000/302/366).