 Hello, everyone, can you hear me? Good morning. We are going to get going. Good morning and welcome to the September metrics meeting at the Wikimedia Foundation. We have a fun filled action packed hour for you this morning. First, our agenda today will have our welcomes. We're going to have a community update presentation. We'll talk about metrics as this is the metrics meeting. We'll have a research presentation, a product demo, have a feature, and at the very end, you will be able to have questions and comments and have all kinds of fun discussions about what we talked about today. But first and foremost, we'd love to welcome our new hires and contractors, interns and volunteers. In our new hires, we have Lonnie in finance and administration. Natalia and Sam in product. Nina in finance and administration. Manuel in technology. As for our contractors, interns and volunteers, we have Liana in legal. Katie in communications. Tia in advancement and Valona in TNC. Let's give them a very warm welcome. Other excitement, anniversaries. Celebrating one whole year at the Wikimedia Foundation are Jeff Allen, Karen, Yate, two years, Jake, Satie, Marcel, Marty, Bartos, Jeff, Odir, Alex, and Ellery, and Jorge for two years. Celebrating three years, Dan, Gary. Celebrating four years and a dress scoop. Five years, Chris, Satush and Kristen Satush. Celebrating eight years, Michelle and Eric, and celebrating 10 years. Mark, Marks, Ma. Quite an accomplishment. And next up, we've got Community Update. Hi, all. I'm Maria Cruz. I work in communications and outreach in the Community Engagement Department. And I am here to share some stories from the community. So what happened last month? Wikimedia's CEE had its annual meetup from August 27th to August 29th. About 60 people joined from a dozen different countries and they received training on several topics, including social media, WikiData, reporting and storytelling, and shared knowledge on programs focused on education, GLAM, WikiProject Med, and GenderGap. And they also had a space to reflect on the impact of long and of small and large meetups. And you can read more in the page there. WikiLoves, The Olympics. This was a writing contest organized by Wikimedia that took place from August 5th to September 18th. And it covered both the Summer Olympics and the Paralympics. And it's probably the largest writing contest across languages that I've seen. And just so you get an idea what language communities took part. Spanish, English, Catalan, Asturian, Portuguese, Bulgarian, Polish, Ukrainian, Punjabi, Arabic, Japanese, French, and German. Only comparable to, from what I know, C.E. Spring Writing Contest. And one cool feature is that they created lists of article requested using the missing topics tool. You can read more in that link over there. Women of the Antarctic. This was a editing marathon hosted by the Scientific Committee on Antarctic Research in partnership with Women in Red and Women in Science. It was important not only for the content added and the awareness generated about Wikipedia and the importance to take part, but also because it inspired a lot of young women scientists. As you can see from this participant quote, this wiki bomb gives us and people everywhere access to greater knowledge, strong role models, and a richer understanding of the history of this shared global space. Women included. You can read more on the link below. And finally, Wikimedia Project Milestones and Affiliates. Wikidata reached 20 million items. Chinese Wikipedia reached 9,000 articles. Albanian Wikipedia reached 60,000 articles. Applaud. And upcoming collaborations with communities. The leadership development dialogue is ongoing until October 16. Please feel free to contribute on the link there. There will be a legal consultation on Creative Commons 4.0 starting on October 4th. And Wikiconference North America will take place in San Diego from October 7th to 10th. If you have a collaboration upcoming and you expect 50 people or more, please add it to the public calendar or meta. And with that, fundraising. Who is presenting? Oh, okay. That's me. Hello. Can you hear me? Yes. Yes. Okay. Hello. My name is Megan Fernandez. I'm the director of online fundraising. And I'm sorry I'm not there with you today, but I'm happy to join remotely. And tomorrow we are posting our report on the 2015-2016 fiscal year. So today we're going to show you a few highlights from the report. It's a very exclusive preview. Let me see if I can share my screen correctly. How do I do this? Brennan, you can tell me if it's, is it okay? Looks good. Okay, great. Here you have a quick overview of where our donations came from in the 2015-2016 fiscal year. We raised about $77 million from over $5 million donations. These donations came from chapter fundraising, major gifts and foundations, and online fundraising. And I'll be focusing on the online side where we see that desktop still makes up the most significant part of our revenue. However, email and mobile are becoming more critical to fundraising. And this follows a general readership trend on Wikipedia. Our traffic is declining. And as there are less readers on our site, the portion of our revenue coming from emails we send to past donors becomes more important. And over the years, we've grown our donor contact list and made huge improvements to the performance of our email program so we can really see it becoming hugely important and that bubble is getting bigger. The amount raised from email in 2015 doubled from 2014. So really, it's very quickly becoming a major source of revenue. Our readership is also shifting from desktop to mobile. And we can see that. And this is our second year that we've run mobile campaigns. The donation rate on mobile is much lower. So readers are much more likely to donate from desktop than mobile, which is why desktop still remains the most significant portion of the revenue. And we have made some pretty great improvements on mobile so far, but it still remains a real challenge for us. And we also received donations from recurring gifts and these are people who signed up to give to us monthly. Here's a look at where the donations came from. Throughout the year, we run campaigns and languages all around the world. And similar to past years, the bulk of the money came from North America and Europe. One new area for us this year was Latin America. The fundraising tech team integrated with the new payment processor, which allowed us to accept local payment methods and currency in Latin America. So we were able to run our first big campaign there and dramatically increase the amount of money we raised in those countries. We raised 17 times as much money as we had previously in Latin America. So we run campaigns year round in different countries, but we raise about half of the revenue in the month of December alone when we run our English campaign. So December is a very intense period where we're sending millions of emails and testing hundreds of banners on the site. The work volume is exceptionally high and the deadline for us to hit the goal by the end of December is exceptionally critical. So we're almost in October here. This is coming up again pretty quickly. So right now the team is busy preparing and making sure our systems are stable to handle the volume. So if you are walking around the office and you see the fundraising team looking busier than usual, that's why this is about to happen. And here you have a look at our fundraising totals over the years. While the average donation amount has declined over the years, the overall number of donors and amount raised has increased. Our average is about $15 and we get lots of $3 donations, lots of $5, lots of $20 donations and that really make up our budget just with the sheer number of supporters that we have. We're now raising five times as much money as we were about six years ago. The steady increase on this line here reflects improvements to our fundraising appeals over time and that big jump in 2014 is where we started showing readers one large banner at the beginning of the campaign. We'll show some banner examples later on but that big jump is when the banner's got bigger. So for now I'm just going to remind you to please read the report that we're posting tomorrow. It'll be up on MEDA and we'll send a note around to the mailing list. We're just showing you a few highlights today but please do take the time to read the report and for now I'll hand it off to Caitlin Cogdill to talk about emails. You good? You good? No. Yeah, everyone can hear me? Okay, so I'm Caitlin Cogdill and I manage our email programs for the online fundraising team and I need the quicker. Cool, so as Megan mentioned emails becoming a rapidly growing part of our revenue sources for the fund foundation. In the last year we saw 103% increase in email revenue from about 8.5 million to almost 17 million. So how does that happen? Well one of the main ways is that our email list grew by 43% last year so that's because our banner team is constantly doing a better job of optimizing our banners and getting more of our users to become a part of our donor list. Every time a new donor gives they become a part of the list that I will then email the next year. But like I said that was a 43% increase. We've got 103% increase in revenue so where else does that come from? The main thing is we are constantly A-B testing our banners just like our emails just like the banner team is and optimizing to make sure that we are reaching our donors in the best way possible. Last year we ran 697 A-B-C-D-E tests in email and so that means anything from like the subject sign or the front name in the inbox to different elements of copy to the color of a button to adding graphics. Any of these is like a single element that is eligible for a test in an email. We have to go forward. So one of the most valuable things that we tested last year was adding in different buttons into the email that give a donor amount options based on the last amount that they gave and also displaying to the donor their past donation history saying something as simple as a year ago you donated $15 seems to be enough to really anchor the donor remind them why they're on our list and why they might want to give again. So as you can see here our emails perform far and away above industry standards and there are a few important reasons for this. I would say the biggest one is that the average nonprofit sends about 15 to 30 fundraising appeals a year and we send pretty much a maximum of three in almost all cases. So that allows our emails to be sort of more productive in each one that we send because our donors know when they're receiving an email from us they're not going to get one very often. They have sort of very few chances to respond to an email appeal from us and must be important when they get one. I want to call it specifically here the conversion rate which is donations over email sent. We get about an eight and a half return on every email we eight and a half percent return on every email we send just about 30% of the industry average. So over the 14 and a half million emails we send per year we know that at least eight and a half people are going to respond. Just pretty too good. So for us a really important tenant of our our funders and campaigns both for banners and emails to make sure that we're not oversaturating our donors with appeals which is why we send so few email appeals per year. When I started on this team we only sent two emails a year per donor and they were only fundraising appeals. We didn't really send anything informational. Nothing was just sort of more information about our work. So that kind of comes in conflict with another one of the goals of the fundraising team which is to make sure that we are being good stewards of the Wikimedia movement teaching our donors about what WMF actually does and sort of helping them understand that we are a lot more than Wikimedia or at least what they sort of think we do on Wikimedia. So last year we decided to see how can we educate our donors better while still respecting their inboxes and making them feel like we're only going to ask them a finite number of times a year to give. So we pulled a small segment of our US list out just under a hundred thousand people and we sent them a couple of different numbers of newsletters last year and our winning segment received a newsletter in April and October and then when we emailed them in December asking them to renew their donation they were 14 percent more likely to give than the people who had not received those newsletters and their average donation was about two dollars higher. So this was a really exciting result for us because it told us that like we actually can do the work of educating our donors and it's actually going to make them hopefully more invested stakeholders in our work rather than leading to more unsubscribes and everything else. So this year we decided to expand that newsletter cohort to a larger portion of our English speaking list as well as a small part of our Italian list. So we are sending newsletters to about two million donors in the US and Great Britain as well as a small number in Italy and we're testing sending different kinds of content as well as different frequencies of receiving this content. So our control group is receiving two newsletters a year. There are some groups that are receiving as many as eight newsletters and then different frequencies in between for testing content like videos, blog highlights, fun facts from Wikipedia and then in this fiscal year we're sending emails like next week there will be a CTA to download the app which is really exciting. Get people involved in user research. We want to see what levels of engagement are the right amount to keep our donors invested, interested while not exhausting them when it comes time to ask them to give money during the fundraising campaign. So these newsletters are getting extremely high engagement just like our donation appeals about 30% open rate and 7% clicks on the emails that we send and the unsubscribes are super low. So this tells us our donors want to receive this information. It's really exciting. If anybody in here has information that you want to share with our donors or stories you want to tell let me know and I'll see if we can work it into our schedule for 2017. And the last thing I want to mention about these is that there is no donation ask in any of these newsletters. We want to make sure that we are just providing a service to our donors and they don't need to provide anything back for now. So we have some later. We do see this as a potential revenue source in the future. So with this two million emails I would estimate we could probably make about $300,000 if we were to include a donation button at the bottom of each of these newsletters. So it's good to know that that revenue stream is there later but for now we're going to sort of not tap it and save it for when we need it. So on that note I'm going to hand it off to Sam Patton who is our campaign manager for banners. All right, great. Thanks very much. Caitlin, I think everyone can see my screen and Brendan will be advancing the slides. So yeah, I work with Megan and Peter Kuhn and Jessica Robell on the banner team. Our first slide is just an overview of some of the key metrics from all our banner testing. Number of banners tested is itself not necessarily indicative of success, right? But it speaks to the amount of spaghetti we throw against the walls. And as Megan said, we've only really been emphasizing mobile fundraising for the last two years. And so a lot of the growth and overall quantity of banners tested is coming from are the continued emphasis on mobile and iPad. You can see we tested more than 50 percent more banners last fiscal year. And then the last statistic that we've chosen to highlight is the number of donors who give the first time they see a banner. That corresponds to people seeing what we just called the large banner, which is Megan said we introduced last year. I guess, yeah, last year and it led to a enormous spike in revenue. And considering that one of our tenants is to impact the reader experience as little as possible. We only show that banner once. It is a huge point of entry for many people, and then we revert to much smaller banners. If you've seen our fundraising banners, it's hard to imagine you haven't independently or on the job. You'll definitely realize that copy does so much of the self for us. And we do see our banners though they have the core priority of raising our annual budget. We also see it as a really important place to introduce people to the values of our organization. And so we were really happy last year to adopt what we just call the knowledge copy that came from staff suggestions. And you can read it on the screen. It appears in the second half of the primary banner. So the first half is, you know, a more traditional appeal with certain talking points that we've found to be very effective. But then the second half of the banner, we do kind of expand our focus a little bit into the importance of knowledge to us. And we consider it a big win that we're able to get that copy into the banner. We're always looking at new ways to incorporate more staff feedback. And then the next copy test I'll just mention, and as Caitlin said, AB testing is the bread and butter of our work. And we throw a lot of copy out that at best will break even. But it's been interesting to see in the last year that one of the places that most consistently we've seen an improvement from copy is the simple change of the opening sentence of our banner to actually reference the country the country in which we're currently fundraising. It doesn't always work, but it has in many cases produced a pretty big impact. And it's something we're continuously trying to expand is everyone knows privacy and protection of user privacy, not collecting data. It's also really important to us. And so there are a lot of limits on how much we can actually target our banner content to the reader. This is one of the first places that we've seen a game with some kind of targeting. And we're going to continue to explore exactly what level of focus, you know, we bring to our copy and how we can keep producing these kinds of wins. Another highlight we really wanted to mention was what we just call remind me later. This was the addition of a option in our fundraising banners to actually sign up for an email reminder. It hides the banners for a week. It's an alternative to just pressing clothes or obviously just scrolling away. And this has been really exciting for us and we consider it the kind of game changer that will have repercussions for many years where not only did adding this remind me later option create a slight increase in donations. But it's allowed us to collect an enormous, you know, email list is growing all the time. You see 200,000 email addresses collected so far. Caitlin Cogdale has done a lot of forecasting and we're predicting right now over $1.3 million in additional revenue from these remind me later signups. And then the screenshots below which I admit are somewhat hard to see. Just the red circle shows where the remind me later option actually appears in each of the forms. And I think this is also really interesting because I've been involved with the fundraising team for about three years. And when you think about it, we've been optimizing like our desktop banners for over 10 years and you reach practical limits of creativity and what you can change next. But we've been able to produce significant gains to the remind me later feature over the last six to nine months. And I think it speaks to in some ways how good we are at iterating, especially when we have not yet optimized a given element. And so this was a breakthrough kind of option that Peter Coom finally got to work for us last December and that will will pay off for a really long time. There's just a couple small, you know, small elements but but highlights are some of the biggest games we've had. And then just to talk a little bit about what we're planning next. So mobile, you know, it's all mobile for us. And I think it's in many ways all mobile for much of Wikipedia as we see traffic move there. We've been working with the reading team on doing some first testing of what kind of fundraising message we can present in app. We're going to continue to focus on mobile web. Caitlin continues to, you know, every year double our email efficacy and we'll see how long we can keep that up. Payment methods, as Megan mentioned with the 18% increase in in Latin America, which came almost directly from the fact that we found payment processors that we could actually work with. That continues to be a really important function of FR tech and ops. And then recurring donors are, you know, we always offer recurring option in our forms, but we're constantly trying to both renew recurring donors who may be a fallen off and also just increase the number of donors who choose recurring option over the single. So with that, I will pass it to Caitlin virtue and Megan and major games. Better now. Okay, there we go. Okay, so major gifts is actually major gifts and foundations. We consider major gifts anything over 1000 USD. And last fiscal year, just about 1400 people gave donations at that amount higher. It's about nine and a half million dollars. Who are these people? They're individuals, they're small family foundations, they're large institutional foundations and their corporations. It's people from around the globe but it mirrors very closely the the continent, the globe map that Megan showed you before. Majority of funds are coming from from North America and from Western Europe. We default to privacy for these donors just like we do with everything in the organization. But a large number of these donors choose to be recognized on the benefactors page and the link there points you to the benefactors page. You can see who's giving and how much they're giving and where the distribution is. And so, if you're interested definitely go ahead and take a look. Most of our major gifts support the general operating budget. We have a very small number of gifts and a proportionately small percentage of the 9.5 million dollars is restricted funding and supports things like the global education program with the people here and specific engineering projects. And finally, I think a good way to think about major donors is they're very similar to extremely active editors. They're contributing a disproportionate amount to our projects and we're really grateful and lucky to have them. Thanks. Next up is David Strine who's going to talk at FRTech. Hi, I'm David Strine. I'm the product manager for Fundraising Tech. You've heard a bunch of mention of us through the presentation. And some of the things we've been working on in the last year is some updates to CVCRM which is our CRM system that is an open source alternative to things like Salesforce. We're leading the development in this actual open source project in a bunch of different ways and we're probably the largest database out of all the other projects that use CVCRM. So we see a lot of things at scale that other people don't and some of them are duplicate donor records. We have a lot of them from many, many years of fundraising and we need to kind of understand an accurate depiction of our donor history and how to communicate to them properly. So we've been implementing some scripts to de-dupe basically our database and over the last few months we've been able to clean up at least two million records and the scripts keep running keep making improvements. This is going to help us tailor more communications like custom emails and things. The email team estimates a 5% improvement or $850,000 more in revenue that we could we get from these improvements and then MajorGives has a lot of work during the big and much campaign where they're just a lot of heavy lifting of fixing records and cleaning up data and stuff and we feel like we could we can improve their workload by 40 to 52% just in that December time period. So we're going to keep working on that. You've heard a bunch about Latin America and we actually helped enable donations in eight new countries that's Brazil, Portugal, Argentina, Chile, Colombia, Mexico, Peru and Uruguay. We've had probably had some tests and some historic communication with those countries. But this is the first time we had real campaigns there mainly through the AstroPay payment processor but we also made a bunch of changes to the Royal Pay Add-In and Amazon payment processing. The vast majority of our work is either maintaining or integrating new payment processors for different countries and different locations. And then we also had we were able to ship a feature called Banner History just before English the big English campaign of last year. And that is a really interesting alternative to Unique Tokens which had been kind of a contentious topic in our organization and Andrew Green is going to actually present on that in our demo a little bit later. And I believe Joseph said is next. Hey, you all. Yeah, if you could stick over to the community agent side. So my name's Seddin. I'm the de facto community liaison for the Advancement Department which obviously includes fundraising. I'm just going to round off the section on fundraising. So over the past few years fundraising has provided areas for comment or feedback on our fundraising messaging on meta. But we wanted to be more proactive in seeking out that community input. Our fundraiser is probably one of the few occasions where we are actually engaging with our readership on the websites. And so it's really important that our messaging and designs that we use in our appeals and our campaigns are representative of the people who are part of the movement. We already actively reach out to chapters for feedback on our campaigns. And in the past we've run in person feedback sessions and and it's these that we we wanted to start running again. So back in June we started off with experimenting with in person and online feedback sessions with around 30 non fundraising staff at the Wikimedia Foundation. And that was looking at our mobile desktop banners and also our email campaigns. And recently just now in September we've run some wholly remote sessions just with community members and starting from probably November onwards we're looking at running these regular feedback sessions for the community. Going forward it'll be across multiple languages as our fundraiser runs in different countries. So if you're interested about being informed about any feature sessions please do contact me my emails seven at wikimedia.org but otherwise I'll be reaching out to communities through village pumps and mailing lists in the future. So thank you very much and that's it from fundraising. All right next up is a research update I'm Dario I run the research team at the Wikimedia Foundation and today I'm going to talk to you about something that's been involved in for a couple of months and I'm very excited about but I want to start with a quick interactive poll. So if you were to name the single most critical ingredient of a Wikipedia article what would that be? Come on I want to hear your answers. Is that references? What am I hearing? Is that again? Info boxes anything else? Internal links Red links anyone? Idols All right Okay, most of you were wrong but yes the answer is indeed references and why is that the case? Well the reason to me is pretty simple like a Wikipedia as it's famous as I say it says is not about the truth. Wikipedia is about verifiability so references and sources really are our superpowers they're the mechanism by which open knowledge can be verified corrected improved built upon in a nutshell they're really like the fundamental mechanism by which the quality control for open knowledge happens. And what is the meaning of practice? These are nice words Dario but like give us some data so I figured this is the first example that to me was a mind blowing so Wikipedia is today among the top 10 sources of traffic the scholar literature our friends at Crossref shared this data earlier on that measures the number of DOI lookups the literature and we know today that Wikipedia is one of the prime vehicles by which scientific knowledge is accessed and distributed way ahead of many publisher websites from all over the world. We also know that students researchers professionals rely on Wikipedia over and over for pre-research so having access to high quality references and sources is really critical for a massive number of people on the planet to be able to do their job and finally in the context of medical information Wikipedia is playing a tremendously important role in disseminating information that is very often is Baywald so our ability of as a movement creating content that is summarized and curated and disseminated particularly in the context of major public health crisis delivered in local languages is really a critical aspect that depends on the ability of referencing information correctly on our projects and there's another reason there's another reason why I think this is important this is not something that depends on our own projects but is a major industry trend at the moment if you look at what's happening in the context of search engines and computational answer engines by and large you will see that provenance of information or the sources for information is displayed when you look up an answer is stripped entirely from search engines is one of my favorite examples you Google the average lifespan of a goat and you get a naked number with no attribution even though this data is coming from a Wikipedia article which itself is citing some additional sources this is not just about Google basically the entire industry is doing this Siri, Amazon, you name it and you may say okay this is not really critical for goats but what about information such as a drug related information or information on policy or human rights just think about the implications of stripping references and undermine the view of verifiability when it comes to forming opinions and critical thinking about any of these topics so this is great and if by now we all believe that references are the superpowers of open knowledge what have we done technically to support this effort and I have to say that despite some tremendous initiatives that I'm really excited about you know Psytoid I'm thinking of Ref Toolbar so like really major initiatives driven by the foundation by our chapters our volunteers the way in which represent citations and references compared to any other ingredient of a Wikipedia article is still via antiquated method of marker so what does that mean? it means that citations and references are hard coded into the body of an article it means that they're very hard to manipulate they're very hard to reuse you cannot reuse the same reference across multiple articles you cannot reuse across multiple multiple languages if you want to conduct research on citation related practices good luck with that you need to parse the full content of our XML dumps so it is really a pain to deal with these objects whereas we have comments and we have wiki data for any other type of object that is reused across our projects and if this is a state of references in Wikipedia the situation is not necessarily better in wiki data so this is a plot that shows you the growth of statements over time as a function of where these statements are getting their references from so about 23% of all statements in wiki data as of today are properly referenced in external source two-thirds of these statements are either not sourced at all or they cite a wikipedia article as a reference which as you may understand is not really optimal so how about we build a bibliographic database for all wikipedia project that sounds amazing and in fact we've had about 10 years of efforts in wikipedia on meta of attempts at building such a thing and it turns out that both the technology and its community efforts were not mature enough to get us to a point where we actually build this thing today things are different and they're different because we have wiki data so wiki data today offers not just a dream of this idea but it has the vision the underlying technology the community of contributors the scale the licensing scheme are important and last but not least also the independence for building such a thing as opposed to all proprietary efforts that exist out there building a citation graph or bibliographic database and so earlier this year we brought together 60 people from our movement wiki datans wiki medians we had people from affiliate organizations so we had researchers and librarians dealing with citation modeling and citation bibliographic metadata and we hosted an event to start figuring out what this thing might look like and we're fortunate enough to have a generous funding from a number of organizations including the Alfred Sloan Foundation the Gordon Bettymore Foundation and Crossref and the event was co-hosted by wiki media and wiki media Deutschland and so the high level goal of wiki site really is to try and figure out how to build a repository of bibliographic metadata for all wiki media projects now what that means in practice is something much more complex that involves designing data models and technology and workflows for creating and curating all this data and cross-linking references with citation data figure out how to make the machine readable and it starts to comply etc etc so there's a lot of work going on that we started tackling in earnest in Berlin this year and there's a detailed report you can find on meta where you'll see each of the workgroups and what we've achieved there this is just the beginning so there's a lot of work to do and today I just want to wrap up by giving you a few highlights of totally non-representative ideas that started there but these are ideas that I like personally very much the first one is that we have data models for bibliographic metadata that are getting to a point where they're nearly complete in many areas there are some areas that are complex to model like book metadata there are some areas that are relatively easier to model like journal articles and we have a pretty complete and rich model there to start representing these citations we also have as of today all PubMed Central articles cited in English Wikipedia already imported into WikiData this is based on a work that Aaron did extracting and parsing citations from the English Wikipedia dumps and Daniel Michien went ahead and worked on importing and curating all this data with some other people into WikiData we also have an independent initiative driven by James here in DC to import into WikiData the entire open access literature I should say review papers from the past five years which are basically the main kind of content we should reuse in Wikipedia projects because review papers especially their open access allow us to find good citations for all of our contents and on top of that we also started representing not just the bibliographic metadata but also the citations that basically link any paper with each other and as of this month we hid we hid half a million statements representing citations in WikiData using a property it's called Sites that was created during the event in Berlin and finally how many of you have heard of Sica virus that's a pretty big deal right and many organizations are trying to sort out what we can do collectively to accelerate research on on this major public health problem it turns out that the entire body of the literature on Sica is basically thousand papers so it's minuscule and we're now working with different communities both volunteer communities and a community of experts to create a complete corpus that integrates the encyclopedic layer that's basically encyclopedic articles and WikiData entries and Thomas entries with an extracuration layer with where this community is representing all the facts that can be expressed in WikiData by linking them to the literature as well as a bibliographic layer and an open citation layer basically an entire graph of citations between any pair of papers in this corpus so there's something I'm very excited about there's much more going on please join us if you're interested in this initiative we have extensive documentation on meta and WikiData we have some social media handles and we have a mailing list and we're also working to host a follow-up event in 2017 so if you're excited about building the summerful citations for open knowledge please join us thank you okay hi can you hear me yes okay fantastic hi thank you so my name is Andrew Green I'm on the FRTech FRTech team and I'm going to talk to you a bit about banner history I think the word demo maybe is not a good word because it's more just tell you about the future banner history is a feature of the system that shows banners on Wiki that system is called central notice and it shows banners that are for soliciting donations and also many other kinds of campaigns calls to action and community announcements and we use central notice to target campaigns banners for fundraising campaigns based on language country device type many other criteria and so the goal of the banner history feature that we're going to show is to improve the way the information rather that we have about how the history of banner display relates to donations in a way that strongly respects user privacy we've done this in a way that that avoids collecting any unnecessary data as much as possible and thereby only get to what really really we think would help to reduce the number of banners shown that is the information we need to be able to understand better how banners are how banner history sorry I'm getting confused okay I think you know what I mean it's basically just we only want the information that we need to be able to better analyze how the history of banners impacts on donations and nothing more okay next slide please so as you know donations come from people not page views the likelihood of a person donating depends on a lot more than just the content shown on a given page view a big factor is the history and the timing of banners that are seen or not and previously our logs were treated were treating each banner view as an isolated event they weren't providing us with almost any information about the previous banners shown or not shown on this device and so we're missing with that a lot of potential data that could help us target banners better and as I was trying to say a minute ago better targeting should hopefully lead to less banners showing so next please yeah some specific questions that we can ask of data here is whether how effective is the large banner first approach I mean we know from the fundraising numbers as you saw before it has quite likely had a very important impact but we could certainly use more details about that when is it most effective in what ways it most effective how does that relate to previous banners seen when does the user get tired of seeing banners when is it just too much and it's not useful at all to show a banner and we're just annoying the user instead next slide please and so the banner history system what it does is it accumulates the information in the user's browser instead of using the more typical way of logging over page use which would be thing called unique IDs so unique ID is something IDs is something we don't do that's the more standard way of doing this or things like this it would be that standard way which we don't do is to assign a unique identifier to each device and send that identifier back with every action with a you know on every action that we wish to log and then analyze everything once we have as much as we possibly could collect using that that's a little bit like phishing by dynamite to quote our our dear colleague Adam White so that would give us a lot of information that we don't really need and it's not a great way to respect user privacy so instead we store in the user's browser the information that we think is really going to be useful the user can actually see what information is being stored there they can see the JavaScript that is processing all this so we're being as transparent as possible with this also and we're not storing articles user actions user names addresses anything that could be potential violation of user privacy or just anything that could potentially be used to to to look at user actions in a way that that's not necessary and potentially could be undesirable so that is only collected actually when users are actively targeted by a fundraising campaign no unique IDs are stored ever on the user's computer and it for users who are in the campaign we sample the banner histories stored on their computer so don't even get all the banner histories we just get the ones for people who don't donate we just get a small sample of those a small sample is really enough considering the numbers we get and then for users to click on donate we get 100% of those those histories and it's some we only generate at that particular time temporary unique ID which is just for the sake of linking the history to the donation information and that's because the donation information goes on a separate cluster also for privacy and security reasons and we just need the unique ID at that time to to link up is to bits of information that we have stored in two separate places but we never store that on the user's computer and it's not possible to correlate this information with any other history for example in web logs or anything like that and so just to give you some idea of some of the potentially more fun and advanced analyses we could do with this we could for example and this is stuff that's really a working progress that's why I don't have anything concrete to show right now but there are analysts who are working on this we can be looking for patterns in banner display history such as correlations of frequency time of day day of week language or country the banners are displayed in and then we can see how typical patterns may or may not relate the patterns of donation so just an example we could see how people who have seen several banners during work hours on a weekday might potentially almost never donate and if we see that's the case we could just try to not show banners there at all and therefore thereby reduce the annoyance factor of banners without without affecting our our donation rate at all so that's all hope there are questions at the end and thank you so much so that's all we have for our actual presentations this morning does anyone have questions comments things they would like to add anyone Euler okay hello hi sorry the microphone wasn't working so I have a general question from Pine unfortunately I think most of the C levels are on retreat today so probably won't be able to answer this question but can we get updates on the hiring for the chief technical officer is there anyone here that has any information I don't know if Amy is here but I can do a quick high-level overview we're moving forward with some finalists and we're pretty close to something that we're very happy about and hopefully we'll be able to make an offer in the next couple of days or weeks that's what I can share with Amy's around and there's no C levels so I'm just answering and another question from Matt Flaschen based on Andrew Green's presentation how do you measure the cost of showing an additional banner when comparing to the predicted benefit yeah that's a great question it's one way would be the rate at which people click on the close button I think also just the lack of donations would be an indicator as far as I know there's only two specific things but perhaps there's others that won't come to mind now yeah that's a great question I think that's an issue that should be studied more from like we should have a definition that like a good answer in the theoretical terms for you and then if necessary maybe think of more data points though I'd prefer not to add more data points to this if possible but I suppose it's also conceivable if it were worthwhile yeah great point that's all we have from IOC right now any other questions okay we'll move on to wiki love and this is where you can share fun stories about your co-workers and fun things you've got going on want to especially push on anything positive is there anything anyone would like to say I love whoever made that slide this is really fun I wanted to thank the presenters for this metrics meeting I really enjoyed the content of it and it feels really well prepared for curated designed all of that so thank you and I also want to thank you Amanda for doing some really good collaborating work with various teams in engineering and thanks to Katie and Toby and John and a lot of teams that are starting to work with various teams and community engagement in a really collaborative way it's exciting for us I want to send some love to OfficeIT because I have to say I've been presenting myself in a while and monthly metrics but it was the most seamless presentation I had smooth like smooth transitions between remote and network so big shout out that was a really big hi this is Lynette I wanted to just give a shout out to Lamayli who has been helping me and the admin team and training lots of folks with all the transition she's training folks for metrics for quarterly reviews all of it and she's just done an amazing job she'll stay in next year a couple of weeks through the end of October Robert as well and of course all that planning that goes into this thanks guys I want to thank the reading team for doing all sorts of awesome collaborations with the fundraising team from helping us test different kinds of banners to working with me on sending out emails to download the app it's just been like a great example of how our teams can collaborate and hopefully provide a better experience for everyone so thanks reading so I was going to pass on some wiki love which no one had explicitly asked me to from the IRC channel until we actually got it there's lots of people sending wiki love to the puppies that they just saw on the stream when we turned the camera around there are several puppies in here not just one okay anyone else all right fantastic thank you so much for sharing the last hour with us and until next time thank you