The first of our two indicators is the Export Potential Indicator. Actually, this indicator comprises two sub-indicators. First, we have the Potential Export Value. This value is a standard value of how much a country can expect to export to a particular market for a given product in the coming years. This value will depend, of course, of the capacity of this country to export the good. But it will also depend on the fact that this market is actually importing the good, and would not be an important role if there is no import in that particular market for the good. And a third component would be the market access conditions for this particular exporter for this good. Once we have computed the potential export value, we can compute the second sub-indicator, which is the unrealized potential. And this unrealized potential just results of the comparison between what is currently exported and what we have computed could be exported by this country to that market. To do this exercise, we need a number of data. So for instance, we will need data about trade, trade flows, export, and import values. These data come from a database that is produced at ITC. Then, we also use tariff data, ad valorem tariffs that come from another ITC database called Market Access Map. The third set of data that is needed in the calculation is price elasticities. These price elasticities come from a database that is from the consortium GTAF for the GTAF database. Then we need also distances between countries and these distances have been computed by SEPI. So, this is a SEPI distance database. And eventually, we will also need projections of GDP growth. And these data come from the IMF and they are released in the World Economic Outlook twice a year. The problem when we use trade data, is that the quality of trade data is not always great. The data reported by exporters and importers may differ. There may be a number of issues with these data. To solve that problem, to tackle it, we apply a number of treatments on these data. First, we will compute averages over five years instead of just taking one year of data. And we will filter out some product. We only consider products that have been exported during the last three years by a country if we want to compute the EPI for that product. In terms of markets, we will only consider markets that regularly, every year of the five years, has imported the good. To address the fact that the quality is not always very good, we will combine direct and mirror data, that is data reported by the exporter and data reported by the importer. We also want to assess the reliability of the data imported by one exporter in particular for a given year. Because sometimes, some exporters can really provide very low quality data, and using them in our calculation would create a problem even for other countries through mirror data. So to do that, we will compare their data reported by country to the same data reported by all its partners, all its reliable partners. And when these data are very different, we would consider that our country is not reliable. So since we have to do these comparison with reliable partners only, it means that there is an iteration needed to eventually obtain a list of reliable countries for a given year. Now, when this work has been done, we will combine data coming from reliable exporters with data coming from all reliable mirrors. And if a country is not considered as reliable, we will only use mirror flows. Now, for some specific projects, we also validate ex-post all the little that we have obtained with some country experts that can check if our results make sense. Now, a second step in the data treatment is about products that we consider. So, we start at the finest level that is of the Harmonized System that is a six-digit level, the six-digit level classification of products. However, this classification has been revised several times since its creation. And countries do not necessarily apply the revisions at the same time. Also, we need to compute some Dynamic Indicators. So, it means that we need to be able to compare data in the past and data now. So, it could create problems with the fact that we have several revisions. To solve this issue, we have decided to group products to obtain groups that are insensitive of these data revision, these classification revisions. So for instance, if we have a product like salmon that was corresponding only to one code called 030212 until 2007, and that has been split into two set products in 2012, then we will create a group of products that will collect the data corresponding to the three codes, the earlier code that was used until 2007 and the more recent code that has been used in the latest revision. So at the end, we end up with product groups. We will not, for our analysis, consider all the products group possible. There are some products that are never really used in any export promotion activities. So, for those products, we don't need to show them in our results. So for instance, that will be waste and pollutants, because there are very rarely export promotion about waste and pollutants. We don't want to encourage countries to export those goods. The same thing for arms and ammunition, tobacco products, and also some products from the extractive industries, because there is little to say in terms of export promotion activities for those goods. So we have eliminated a small list of products that we considered were not relevant for export promotion. At the end of the process, we have a little more than 4,000 product groups that we can merge into 15 broad sectors or 70 sub-sectors.