 Di sini, dari Universiti Nasional Singapura, hari ini saya datang bercakap tentang kertas ini disebut Multi-Homing dan Oligopolistik Platform Kompatisi. Ini bergabung dengan Chun Chun Liu, Julian Wright, dan Jun Jie Chou, dan seluruhnya, semua orang kita adalah dari institusi yang sama. Mari kita mulakan dengan introduksiannya. Di hari ini, banyak aplikasi membuat value ekonomi dengan memasukkan transaksi atau interaksi antara dan penggunaan servis atau penggunaan. Platform ini bergabung dengan setiap orang. Dulu, kita mempunyai aplikasi yang bergabung seperti Uber dan Live. Kita mempunyai aplikasi yang bergabung seperti DoorDash, GrabHart, Postmade, dan UberX. Kita juga mempunyai aplikasi yang bergabung dan aplikasi yang bergabung dan aplikasi yang bergabung. Di sekarang, ada sebuah literatur yang besar di kompetiti antara dua aplikasi. Tapi jika kita melihat aplikasi di sini, ada beberapa alasan penting yang tidak terkenal dalam literaturan asyik. Pertama, aplikasi yang bergabung di setiap aplikasi adalah base transaksi. Pertama, seluruhnya mempunyai multi-home, iaitu aplikasi yang bergabung. Untuk aplikasi yang bergabung, mereka boleh mempunyai aplikasi yang bergabung semula-mula. Pertama, apabila kita bercakap tentang multi-home, aplikasi yang besar di kompetiti antara dua aplikasi adalah aplikasi yang bergabung dan aplikasi yang bergabung. Tentu-tentu dalam aplikasi kita, aplikasi yang bergabung adalah seluruh untuk multi-home dan mereka mempunyai aplikasi yang bergabung untuk aplikasi transaksi. Pertama, apabila isteri saya dan saya mempunyai aplikasi yang bergabung dan cuba berjalan, apa yang sebenarnya berlaku adalah saya akan memutuskan aplikasi yang bergabung dan dia akan memutuskan aplikasi yang bergabung kita mempunyai aplikasi yang bergabung dan aplikasi yang bergabung dan memilih aplikasi yang bergabung. Lebih gila, motivasi kita adalah iaitu lebih mudah untuk bergabung untuk multi-home. Pertama, mereka adalah aplikasi meta-search seperti Google Maps dan Google Right yang membantu pendera untuk bergabung былäng aplikasi yang bergabung tanpa mempunyai aplikasi multiple. Kebiasaan seperti ini juga digabung untuk bias untuk bergabung hoy saya akan menangguh memberi model yang dibarut untuk berujakan begini yang penting. Terbeda dalam perubahan, mereka adalah aplikasi oleh k pesan ketebat faults kEDA bias perspektif, tetapi homogenous dari perspektif selera. Ini mencapai faktor bahawa dalam banyak aplikasi 2-CyderMarket, platform selera akan mempunyai lebih kurang sama semasa bias selera mempunyai pilihan idio-syncratic untuk menggunakan aplikasi particular. Dan akhirnya, kedua-duanya akan mempunyai 2 multi-home. Selain model ini, kita mempunyai 3 menang. Pertama, kita menghargai ekonomi dari kompetisi platform dengan 2-CyderMulti-Home. Kita beritahu bahawa beberapa perintah dari model mempunyai seperti koneksi antara pengalaman dan peralatan tidak selalu mempunyai. Yang penting adalah bahawa dalam perangkatan transaksi ini, ia penting untuk mengambil perangkatan transaksi yang adalah sebuah kemampuan yang mempunyai model mempunyai ini. Selain model ini, kita melihat bagaimana perangkatan transaksi mempunyai total fee dan fee struktur. Fee struktur di sini bermaksud bagaimana total fee berada di atas belakang belakang selera dan selera selera. Di sini, kita periksa bahawa perangkatan transaksi membaikkan total fee dan membaikkan fee struktur di selera selera. Selain model ini, saya akan menjelaskan 2 kemampuan yang mempunyai pengalaman ini. Kita melihat bagaimana menjelaskan berdapat berdapat beberapa platform pengalaman mempunyai perangkatan transaksi dan perangkatan transaksi dan kita membandingkan dua kemampuan untuk membuat efek kekal. Akhirnya, kita mengatakan pengalaman pangkatan transaksi yang lebih tinggi dengan kemampuan yang kita melihat di sini di mana semua pangkatan transaksi dapat mempunyai pangkatan transaksi. Kita menjelaskan bahawa pangkatan transaksi akan membaikkan fee total dan membaikkan fee struktur di selera selera. More importantly, such multi-homing cost is going to reverse the effect of increased platform competition on fee structure, which is discussed in this point too. Let me jump straight to the model. There is a set end of platforms, a continuum of buyers and sellers. Each platform charge per transaction fee on buyers and sellers. Buyers and sellers are free to multi-hook and participation is costless. Buyer and sellers reach to transact with each other to create economic value. Without loss of generality, we assume that every buyer-seller pair correspond to one potential transaction. So for example, the blue buyer here potentially wants to transact with every seller here. So there are three potential transactions here. Likewise, the green buyer potentially wants to transact with every seller here. There is no competition between sellers. Transaction can occurs only if there is at least one platform that the pair of buyer and seller joins. To illustrate this point, let us focus on the first pair of buyer and seller here. Suppose there are four platforms and suppose that the buyer joins platform one to four, the seller's join platform one to three. Then the transaction between this pair of buyer and seller can only occur through platform one or platform two. And it is also possible to have no transaction between this pair of buyer and sellers. Consistent with our motive-rating examples such as right-hailing platforms, we assume that the buyer chooses whether to transact and if so, to reach platform. Now suppose that the buyer has decided to transact through some platform i, so platform one here. The seller is going to get per transaction net utility, which is V minus by the seller fee charged by this platform i. Here V is a gross value that is seller-specific. It is invariable across platforms and IID across all the sellers. And it is distributed according to some CDFG. You can think of V as indexing the type of the sellers. Buyer per transaction net utility is given by the expression here. V naught here is the draw that is specific to the buyer and specific to the transaction. It is invariant across platform and IID across buyer and across transactions. It follows some distribution function, F naught. Epsilon here is a buyer-platform-specific match component and it captures a buyer idiosyncratic preference for platforms. It is IID across buyer and platforms and follows some distribution function, F. If there's no transaction between this pair of buyer and sellers, both of them are going to get zero utility. And we impose the standard log-concrete assumption on the distribution functions. Platform profit is the sum of total transaction fee minus a marginal cost times the total volume of transaction and I will explain this later. Our solution concept is SPNE with symmetric fees whereby all the platforms set the same buyer fee and seller fee. And here's the timing to wrap up the model setup. First all end platforms simultaneously set their buyer fee and seller fee. Then sellers and buyers observe all fees. Sellers observe their realised draw value and buyers observe their idiosyncratic preference for platforms. Then they simultaneously decide rich platforms to join. After the participation for each potential transaction, buyers observe their realised transaction-specific utility component and they choose whether to transact and if so, through rich platform. In some scenarios, it is possible that buyers observe their preference for the platform only at the point of transaction. Also, sometimes buyers may not observe seller fee and this reflects their vertical arrangement between platform and sellers could be confidential. Going with these alternative assumptions won't affect our analysis. So let's go into the analysis. Let us analyse user decision after platforms have set the fees. Buyer makes two decision. First, participation decision is straightforward. Notice that it is a weekly dominant strategy to join our platform. This is due to zero joining cost and the fact that buyers can choose the final transaction median. This means that buyers get access to more transactions options by joining more platform. And so in the equilibrium, all buyers are going to join the set of all platforms. Next, consider buyer choice of transaction median. To do so, let's suppose that there's a seller fee that joins a set data of platforms. Then the mess of buyers that transact with this seller through platform i is given by the expression here. This is the mess of buyers who prefer transacting through platform i over other platforms in set data and who prefer transactions over no transaction. This object is the standard discrete choice random utility maximisation and it depends only on buyer distributions F and F naught. We now consider sellers. They may only participation decision. It is fairly easy to show that on the equilibrium path where all platforms set the same fee, all sellers with value above the equilibrium fee is going to multi-home on all platforms. And all sellers with value below this fee will join nothing. To pin down the equilibrium fees, let us consider off equilibrium path in which a platform i deraids by increasing its seller fee. Let's let that be platform one and suppose that there are only three platforms. Given that all platforms except one set the lowest seller fee he had, it is easy to see that sellers either join nothing, join all platforms except one and join all platforms including platform one. Notice that what really matters to the derating platform one is the cutoff here, which is the seller indifferent between joining all platforms and creating platform one only. So let us try to pin down this cutoff. Consider a seller fee that is joining all platforms. The number of transaction that comes through platform one is B1 here. For each such transaction, the seller gets fee minus platform one's seller fee. And likewise for transaction that comes through platform two and platform three, which is B2 and B3 here, the seller gets fee minus the symmetric seller fee that this platform charge. Now suppose that the seller wants to quick platform one in response to its higher fee. What will happen? Consider the buyers who have been using platform one to transact with this seller. When platform one is unavailable, some of the buyer will simply switch to use other platform to continue translating with this seller fee. And so the seller is going to gain from diverting these buyers away from the expensive platform one to other cheaper platforms to entry. And so this allows the seller to save on the fee difference here. At the same time, however, some buyers who really like platform one and dislike other platforms will simply stop transacting. So they will switch to no transaction here as indicated in the green arrow. The seller is going to incur a loss from losing access to these buyers. In sum, when trying to decide whether to quick platform one, the seller trace off between the game from diverting buyers to cheaper platform with the loss of foregone transactions. Notice from here that it will be less profitable to quick platform one if this sigma which is the fraction of buyers will stop transacting is large. In the paper, we call this sigma buyer loyalty to platform. The idea is that if buyers are highly loyal to platform one, then it is hard for each seller to divert their transaction. Even though buyers are multihorming. From here, if we solve the indifference condition, sellers with value below the cutoff indicated here will quick platform one and notice that the cutoff depends on buyer loyalty here. Now we want to determine the volume of transactions. We know that when platform one sets the highest fee compared to the other platforms, only sellers with value above this cutoff will join platform one together with other platforms. So the set here 123 these sellers are joining all the platforms. Each of the sellers over here will generate B1 transactions for the platform. And so the volume of transaction on platform one is simply the mass of participating sellers times the transaction brought in by each of these sellers. When the platform once increase its fee further, it trace off between a higher fee versus fewer participating sellers as indicated here. Notice that the elasticity of seller participation is high if buyer loyalty is low. So far, we only consider the case of outward duration by platform one. The case of downward duration downward duration is fairly similar. So let me briefly describe it. When platform one derails by undercutting its seller fee, the logic of of seller participation is similar. Each seller trace off between creating the more expensive platform to divert buyers to the cheapest platform one versus losing access to some buyers. However, the exact participation profile is more complicated. Of equilibrium path, sellers may join nothing single home on platform one, join all platforms or anything in between. Notice here that sellers that multi-home on fewer platforms is going to bring in more transaction for platform one. And the reason is that buyers who want to transact with this sellers over here is more likely to use platform one for transaction. In fact, in this case, they can only use platform one to transact. Over here, they can use platform one or two. And over here, it is either one, two or three. When platform one lowers its fee further, the range of single-homing sellers become larger. And that is going to bring in more transactions for platform one. And so in this case, the platform trades off between lower fee and less multi-homing by sellers. And also essentially, there's a trade-off between the fee and the transaction volume. And you can see that the trade-off is quite parallel across the two cases. To state the equilibrium, we define buyer inverse semi-elasticity as s s x over here. So it is just the equilibrium demand on platform i divided by this demand derivative. Intuitively, this x captures the standard competitive markup with n competing firms and it is followed from the standard perlosal type of model. Likewise, we define buyer loyalty index sigma here as the fraction of buyers who stop transsektin when one of the platform i is unavailable for transaction. Intuitively, this sigma which is between 0 and 1 captures how difficult it is for sellers to divert buyer transsektions across platform. Given these two definition, we can state the equilibrium as follows. In the equilibrium, all platforms set buyer fee and seller fee that is the unique solution to the condition here. This is saying that the platform total margin equals to x defined earlier which represent platform market power over buyers which equals to this is like a standard monopoly markup discounted by buyer loyalty index defined above here. Notice that it is smaller than 1. This composite term represent platform market power over sellers. If we try to interpret this equilibrium, then we can rewrite it as followed. The equilibrium buyer fee is the sum of platform marginal cost plus platform market power over buyers minus a cross-subsidit term that is due to the extra revenue from the seller fee. Likewise, the same decomposition applies for the equilibrium seller fee. From this equations, I want to make 3 observations. First, the source of cross-subsidisation here is different from the membership pricing models like Unstrong and Talented Joke. In those models, the platform wants to subsidise for example sellers because extra seller participation increase buyer willingness to pay for platform membership. However, that kind of dynamic is absent in our seller because transaction fees are like an insulating tariff to use Glambal's terminology. Instead, in our set-up, the cross-subsidisation here reflects what we call usage externality. Additional participation by sellers is going to increase the transaction by all existing buyers on the platform and that is going to generate extra buyer fees that offset platform costs. Second observations, recall that in our model, both sides are free to multi-home. Yet, this does not necessarily imply a specific market outcome because we have to take into account this SIGMA. In particular, when SIGMA buyer loyalty index is very high then buyers are highly loyal and it is almost impossible for sellers to divert their transactions. In this case, if SIGMA equals to 1 you can see that platforms have monopoly power over the sellers and this results in classic componentif bottleneck outcome as if buyers are single-home-made but recall that they are actually multi-home-made. So, transaction behaviour drives the result here. To the other extreme when loyalty index is small, buyers has high willingness to switch to other platforms and in this case platforms have very big market power over the sellers because it's quite easy for sellers to divert their transactions. Third, recall that in our model, sellers view platforms as homogeneous and based on this standard intuition will suggest that firms should have zero market power at least over this side. However, in our transaction platform setting seller difficulty in diverting buyers create platform market power over sellers. In other words, elasticity of seller participation in discount setting has to be derived from buyer transaction behaviour. We next look at the implication of increased platform competition. Let me state the result first. An increase in the number of platforms and is going to decrease total fee. Decrease seller fee if buyer distribution has weakly decreasing density and increase buyer fee if both buyer and seller distribution have weakly decreasing density. This result is suggesting that increase competition in this two-sided multi-homing setup is going to shift fee structure in favour of sellers decrease seller fee and increase buyer fee. So the question is why is it that competition favour sellers? What's the mechanism behind the result? To discuss the intuition, let us first reinterpret the equilibrium condition graphically. The buyer fee equation can be interpreted as a blue curve here. For each given seller fee, the blue curve tells us the equilibrium condition for the buyer fee. Likewise, the green equation here, the seller fee equation can be interpreted as the green curve here. For each given buyer fee, the green curve tells us the equilibrium condition for the seller fee. So the overall equilibrium is simply the intersection between these two curves. That is the intersection of the equilibrium conditions for both sides of the market. Now increase competition so let us focus on the buyer side first. From the blue equation, increase competition makes buyer quality demand more elastic and that is going to intensify competition for buyers. So x decreases. Graphically, the blue curve shift downwards and that is going to exert a downward pressure on buyer fee. At the same time, however, from the green equation, we notice that the lower buyer fee means that it becomes less profitable for platform to attract sellers. And that is going to exert that outward pressure on seller fee. This is the so-called seesaw effect in the literature and graphically, it is just a movement along this green curve seesaw effect. Finally, our log concurrency assumption implies an incomplete pass-through property and so the increase in seller fee is always smaller than the decrease in the buyer fee and consequently the turtle fee is going to decrease. Next, let us focus on the seller side of the market. From the green equation, the low increase competition is going to make the platforms more substitutable. And so that affects buyer loyalty index and makes buyers less lawyer. When buyers are less lawyer then it becomes easier for sellers to divert their transaction through seller participation and so that is going to intensify the competition for sellers. Graphically, the green curve here shift invert and that is going to exert the downward pressure on the seller fee. And again, we have the seesaw effect because from the blue equation here the lower seller fee implies that it becomes less profitable to attract buyers. And so we have the seesaw effect over here where we have a movement along this curve along the blue curve. Again, we have this incomplete pass-through property so the increase in buyer fee is going to be smaller than the decrease in seller fee. So the turtle fee again decreases. Combining both pictures we know that platform competition intensify the competition for both buyers and sellers. And both effects are going to decrease the turtle fee but the change in fee structure is ambiguous. Notice that the two effects shift the fee structure in opposite directions. To proceed further, we intuitively intuitively we can think of the effect of competition on the fee on each side as decrease in the market power over the side adjusted by the next seesaw effect that I've discussed earlier. This next seesaw effect is going to be positive if and only if the decrease in buyer loyalty index which represent platform market power over sellers is greater than the decrease in buyer inverse elasticity which represent platform market power over buyers. The mathematical condition is listed here. This condition depends only on buyer preference distribution F and F naught. And it turns out that it holds quite generally for example if F has quickly decreasing density or if F and F naught are gumbo so that we have a logic in mind formulation. When the when the condition stated in the proposition holds then this next seesaw effect is going to be positive and from the equation here next seesaw effect aligns with the decrease in market power over sellers and so seller fee decrease unambiguously. If we impose stronger assumption then this next seesaw effect is going to be positive and large and is is going to dominate the decrease in market power over buyers and in this case buyer fee is going to increase unambiguously. So from here you can see that this condition plays an important role. So why should we expect it to hold? Why is the intuition? Intuitively having more platforms is going to affect buyer behavior in two ways. First platforms become more substitutable and that is going to decrease the inverse elasticity and decrease biologi which is fairly standard. However recall that our model has incomplete market coverage and so that is going to trigger market expansion effect that increase the inverse elasticity but more importantly it is going to decrease biologi index and this is because when there are more platforms buyer outside option of not transacting is relatively less attractive now so that buyers are less likely to transact whenever seller creates a platform. Notice that the two effect on sigmas are online whereas the two effects on X are offsetting each other and so this explain the condition here. From the analysis we notice that there's this normal object called buyer loyalty index. In the paper we try to unpack this and we ask what are the factors that influence it? We look at multi-homing cost of buyers value of transacting for buyers buyer heterogeneity and seller side factors We also explore how do these factors interact with the effect of increase platform competition? For the sake of time I will just focus on multi-homing cost of buyers Consider the following extension Suppose that buyers obtain a state along participation benefit which can be zero from joining the first platform Then they incur a multi-homing cost which or a benefit if this cost is negative for each additional platform joint Buyers have heterogeneous multi-homing cost distributed according to some CDF For simplicity we will assume that all buyers observe only buyers fees and not seller fees This is going to simplify the analysis of buyer participation decision but does not affect the main insight we want to deliver here This set-out will give rise to an equilibrium with partial multi-homing by buyers Buyers with negative multi-homing cost will multi-hom on all platforms and buyers with positive multi-homing cost per single home So let us denote the mass of multi-homing buyers here as lambda The equilibrium fee will take the form that is similar as before Notice that the market power over buyers is unaffected by the multi-homing cost The idea is that there is always some form of competition for buyers regardless of their homing behaviour For buyers who are multi-homing then platforms compete for their usage For buyers who are single-homing platforms compete for their participation And so market power over buyers remain the same as before However as for the market power over sellers notice that we have a redefined loyalty index When lambda increases so that more buyers are multi-homing the loyalty index decreases So this means that sellers finds it easier to double buyers and that is going to weaken platform market power over sellers So a higher lambda could generate a downward pressure on seller feed and through the seesaw effect that is going to create an outward pressure on the buyer feed Formerly we have the following result stating that more multi-homing by buyers decrease the total feed and shift the feed structure in favour of sellers Let's try to compare this result with the classic competitive bottleneck type of result In those setting for example Armstrong and Armstrong and Wright and Belovin and Pides buyers are single-homing and so platforms have monopoly powers over the multi-homing sellers This is similar to our setup in the spatial case of lambda equals to 0 except that we are delivering this result in a transaction feed setting whereas those people deliver it in a membership feed setting and more importantly Proposition 3 here is saying that more multi-homing by buyers is going to mitigate the monopoly power here over the multi-homing sellers In other words multi-homing by buyers open up the competitive bottleneck and restore the platform competition for sellers Next we look at the interaction we find that homing behaviour of buyers can actually reverse the effect of competition Specifically we show that an increase in the number of platforms regardless of lambda is always decrease is always going to decrease the total feed However if the mass of multi-homing buyers is sufficiently large we get something like a baseline model where the feed structure shift in favour of sellers However when most of the buyers are single homing we will get a reverse result whereby the feed structure shift in favour of buyers now What's the intuition? When most of the buyers are single homing then platforms basically have monopoly power over sellers and there is going to insulate them from competitive fragile and so increasing the number of platform is not going to effect the competition for sellers but it intensifies the competition of for buyers and so in our raw competition will shift feed structure in favour of buyers in this case friend lambda is mock most of the buyers are single homing In the paper we also look at other key competitive statics so we look at increase buyer heterogeneity and to do so we introduce a skill parameter for non deterministic component in buyer utility we also look at increase value of transaction for buyers where we introduce an additive shifter to buyer utility from transsektik we also consider seller side factors for example increase value of transaction for sellers and the introduction of competition between sellers so we call that in the baseline sellers do not complete with each other in all this exercise we show that buyer loyalty index sigma place an important role in understepping this comparative statics so here's the summary of our result today I present a model of oligopolistic platform competition where the platform completes in transaction fees and there's two-sided multi homing we show that increasing extent of buyer multi homing is going to decrease the total fee and shift the fee structure in favour of sellers we also show that number of platforms when it increase is going to decrease the total fee but its effect on fee structure can be different depending on the homing behavior of buyers we also look at buyer heterogeneity and show that more heterogeneity increase total fee and shift the fee structure in favour of sellers and we also look at value of transactions for buyers and sellers showing that those increase the total fee and shift fee structure in favour of sellers and buyers respectively so what are the key aways from today's exercise first we show that the effect of competition with two-sided multi homing work very differently compared to competition with competitive bottleneck in particular notice here that the effect of platform competition on the fee structure can be reversed depending on the homing behavior the second highlight is that we show that transaction based model works quite differently with membership model which is finally adopted in the literature in environment where the transaction decisions of buyers is important we have to take into account transaction behavior and we show that this buyer loyalty index provides a good measure to summarise such transaction behavior and I think I have a bit of time so let me conclude with the relation to the literature so our model has the following key features and we know that by now there's a large literature in two-sided market based on the works of these semitem papers our paper is the most closely related to the three papers listed here Tan and Joe consider the effect of olicopolistic competition we show that the homing behaviour can potentially reverse the effect of olicopolistic platform competition on the fee structure so homing behaviour is important second belafin and pipes analyse competitive bottleneck and the implication of that we show how their insights extend to a transaction fee setting after taking into account that buyer loyalty index plays an important role in our environment and finally Berkels at Halabutah consider two-sided multi-homing and they show that in such setting cross subsidisation is absent when you have a membership pricing model we show that their insight does not extend to our transaction fee environment because our model has incomplete market coverage and finally notice that these three papers consider membership pricing whereas we consider per transaction fees finally our investigation into multi-homing behaviour is related to those in the work on advertisement platforms and media markets however the asset mechanism and setting is very different from ours because of this mechanism of sellers diverting the buyers and so the insights are not quite applicable to one another and that will be the all of my presentation Thank you, Ted Helm and thanks for being on time by the way Marcus, do you want to do a discussion? Ya, tentu saja Ya, saya rasa masalah saya untuk meminta pertanyaan dan berbincang dengan tanpa memberi sumber jadi, saya akan berbincang bahawa saya rasa ini adalah sebuah kertas yang bagus dan terutamanya model sangat menggantikan sangat seluruh dan terutamanya ini adalah pertanyaan yang susah mengenai multi-homing apa yang mengenai multi-homing dan saya rasa yang sangat bagus untuk menggantikan model yang menarik dan berikan perintah yang menarik jadi, terutamanya saya suka kedua-duanya dan juga Tatao memasangkan namun, ada satu sisi di sini yang membuat pilihan yang mempunyai platform untuk menggantikan saya rasa yang sangat mempunyai apa yang berlaku di sebuah pakaian dan juga, mereka menggantikan ini menjadi pakaian yang mempunyai pakaian yang baru bagaimana untuk mengenai pakaian dan bagaimana berbeda keadaan keadaan keadaan seperti keadaan keadaan keadaan keadaan keadaan dan sebagainya keadaan keadaan keadaan keadaan dan apa yang kita boleh memasangkan dari ini untuk keadaan dan juga untuk keadaan keadaan keadaan keadaan sekarang, mengatakan ini jadi, biar saya beri komentar atau lebih fikiran di depan saya dan saya membaca dan juga apabila saya mendengar keadaan jadi, pertama ia, dan ini juga mengenai daripada Tatao dalam slide yang terakhir atau bukan yang terakhir tapi slide yang terakhir dalam literatik yang berlainan yang banyak ini di luar untuk keadaan keadaan keadaan jadi, pakaian ini mempunyai secara sepatutnya untuk keadaan keadaan keadaan jadi, pertanyaan yang mudah dan saya rasa juga juga mengenai dengan keadaan keadaan apabila kita mencari keadaan keadaan keadaan yang biasanya keadaan keadaan keadaan seperti Amazon atau banyak orang jadi, ini sebuah kombinasi dari dua dan juga dalam adiknya ada keadaan keadaan keadaan jadi, ini bukan hanya per transaksi saya mengenai keadaan keadaan atau saya perlu mempunyai sebuah platform untuk keadaan keadaan dibuat oleh pemerintah tapi keadaan keadaan juga dibuat jadi, untuk keadaan keadaan keadaan apa-apa 20 atau 10% bergantung dan bergantung keadaan keadaan keadaan dan untuk keadaan keadaan keadaan dan jadi, satu soalan adalah bagaimana berhasil atau bergantung atau ada sebuah idea apakah keadaan akan berhubung jika satu mengambil dua kata-kata antara pemerintah dan per transaksi atau juga bergantung bahagian yang dibuat keadaan keadaan keadaan keadaan keadaan atau ada similiti antara keadaan keadaan atau bagaimana ia bergantung untuk per transaksi sebuah kata-kata yang mungkin saya tidak bersyukur saya mempunyai kata-kata ketika mendengah keadaan juga dalam banyak keadaan ia bergantung pada kata-kata Pada awal-awal berbual atau awal-awal berbual, yang sepatutnya mempunyai pelajar untuk pelajar yang mempunyai dan mempunyai pelajar untuk pelajar, sehingga ia patut menjadi sebuah keputusan antara mereka. Pada poin lain, dan sebelum datang ke dua poin, yang saya rasa adalah lebih banyak, saya rasa apabila anda benar-benar menjelaskan bagaimana pengalaman multi-homing berhasil, apa yang anda ada pada akhirnya adalah pengalaman sematrik dalam mana apabila tidak ada kos untuk pengalaman multi-homing, semua pengalaman multi-homing dan semua pengalaman single-home. Tapi, maaf, semua pengalaman multi-home juga, maaf. Tapi cara yang berlaku adalah apa yang akan berlaku jika satu platform meningkat atau meningkat satu pengalaman. Dan seharusnya, jika satu platform meningkat satu pengalaman, apa yang anda telah menunjukkan adalah, kemudian bahawa pengalaman pengalaman dalam termasuk pengalaman multi-homing adalah sama, tetapi beberapa pengalaman tidak lama berlaku mengalami multi-home, dan jadi ia jadi jika pengalaman itu adalah sedikit merasakan tanpa kaki-kaki pengalaman multi-home, semua pengalaman multi-home, tapi jika sesiapa tidak mengalami multi-home, ia sangat mempertahankan pengalaman pengalaman kerana mereka mengubah dari platform lebih besar dan mengusahkan pengalaman pengalaman untuk menggunakan mereka pada platform yang lebih kelihatan. Dan jadi, pengalaman saya pada jika ia sedar-sebelah, Sehingga biasanya ia sangat mudah untuk pembelajaran untuk multi-home dan mereka melakukannya. Namun, untuk pembelajaran, mungkin mereka mempunyai platform perjalanan. Mereka bukan hanya pembelajaran dan hanya pembelajaran di sana, tetapi mereka hanya aktif di platform ini. Jadi, masalah saya adalah bagaimana untuk melakukannya dengan sebuah orang. Jadi, ini adalah sebuah feature yang menarik dalam model. Sebenarnya, ia mempunyai pembelajaran yang indistinguisable untuk kami. Mereka hanya menggunakan platform sepenuhnya, tetapi mereka berpotensi aktif untuk banyak orang. Atau mereka hanya aktif untuk satu platform dan mempunyai pembelajaran yang indistinguisable untuk orang lain. Jadi, saya menurut saya, masalah ini mempunyai pembelajaran yang indistinguisable dan bukan hanya mempunyai pembelajaran. Dan saya juga mempunyai, jika saya mempunyai sebuah minit lagi, pemeriksaan yang saya rasa lebih banyak adalah pengalaman. Saya menurut saya, salah satu pembelajaran yang lebih besar, saya menurut saya, yang sangat baik-baiknya dibuat dalam pemeriksaan dan juga dalam berbicara, adalah idea ini mengapa mempunyai pembelajaran sebagai pemeriksaan, saya dapat menghormati pembelajaran untuk melakukannya dengan saya dengan platform yang saya suka dan platform yang saya suka, bermaksud platform yang saya perlu mempunyai untuk mempunyai pembelajaran yang paling tinggi. Dan cara yang saya lakukannya adalah saya tidak aktif pada platform yang lebih tinggi, yang membunyai pembelajaran yang mempunyai pembelajaran yang lebih tinggi, tetapi ini tentu-tentu sebuah pembelajaran yang lebih tinggi. Ia hanya sebuah pembelajaran yang lebih tinggi. Dan kemudian, juga sebuah pertanyaan global, sejak pemeriksaan konsumer atau pemeriksaan biolojik ini adalah sangat pemeriksaan. Saya rasa ia adalah pemeriksaan novel yang saya tidak pernah lihat sebelumnya. Pertanyaan adalah bagaimana ia mungkin untuk mempunyai pembelajaran untuk mendapatkan pemeriksaan pada pemeriksaan ini. Jadi, sebaiklah, kita boleh mencari ini dengan potensi elastisiti, tetapi elastisiti ini sedang berbeza pada pemeriksaan standard, kerana walaupun saya aktif pada semua platformnya, saya hanya menggunakan beberapa pemeriksaan, dan apakah ada potensi jalan untuk mendapatkan lebih banyak konten empu-pembu pada pemeriksaan ini, saya akan kata sangat baik pemeriksaan.