The quest for time-series anomaly detection at Sinch - part one, This blog post is the first in a series about our quest to find the best anomaly detection system for chatbots. Our quest is by no means over, but we’ll update you on the problems we tackle as we progress on our journey. Most of the time-series anomaly or outlier detection algorithms out there, commercial or otherwise, only function well under high volumes of traffic. One of the services Sinch…, Moving average, A simple approach to make sure the data you look at is relevant would be to work with a moving average. You can replace each measurement with the average of its past 200 measurements, for example. If you do that for the data illustrated above, you get the green line illustrated below. As you can see in the image above, the improved success rate stays nicely between 0.7 and 0.9 and is…, Time warping, Imagine your call success rate depends on the time of day. During the day time, you can expect 80% of calls to be successful, but after office hours, that success rate drops to 60%, and at night time, you only have 40% successful calls. With the traffic shown in the first graph, you could get the following measured success rate in blue below. As shown above, the success rate can wildly oscillate…, Conclusions, For non-seasonal data, you can use a moving average approach to convert your metric into something more insightful. Picking a window size that is big enough, you can guarantee that the data you look at is relevant since it inherently will incorporate the number of calls. The resulting metric can be used as input by anomaly detection algorithms requiring stable input. When you want to take…, Want to find out more about chatbot technology and how it can help your business? Take a look at the Chatlayer product page here .
Tipo: Artículo
The quest for time-series anomaly detection at Sinch – part two, Many infrastructure and performance monitoring software tools offer built-in anomaly detection. But they often generate too many false positives. This is the second blog post in a series where we describe our journey in building a better performance monitoring tool for chatbots. You can find part one here . Anomaly detection can also be formulated as a prediction problem. Anomalies are unexpected…, Statistical threshold estimatio, n, There are many machine learning methods out there to predict values for time-series. Some of them even come with tools to estimate confidence boundaries like ARIMA or Gaussian Processes. The output is typically a Gaussian, which tells you how likely it is that the actual measurement will fall between certain boundaries. A commonly used approach is the 3 Sigma rule. If your measurement is more…, Anomaly score, There are methods like Robust Random Cut Forest (RRCF) that don’t work with Gaussian boundaries. RRCF is a tree-based method that tries to model the data. Every time a new data point is entered into the model, it checks where changes are needed to better fit the data. If the prediction was accurate, no changes are necessary, but if the prediction deviated a lot, the tree will need to be adapted…, Scaled minmax threshold estimation, To tackle the thresholding problem, we took a different approach. The idea is that you learn from past data what a good threshold looks like. Imagine you have a model that predicts your metric, the deviation from the actual metric can look like the graph below. If you want to apply this reasoning to the output of the anomaly score of RRCF, you can ignore the negative values in the examples that…, Conclusions, In this blog post, we shared the approach we use to set thresholds for our anomaly detection model. For predictive anomaly detection methods, we use deviations from the prediction as an input signal. That ensures we can use time-series prediction methods to capture seasonality, but we don’t have to deal with the seasonality when determining the ideal boundaries. For methods that return a score…
Tipo: Artículo
A2P 10DLC Q&A with messaging experts from Sinch & The Campaign Registry (TCR) Conversational A2P 10DLC messaging solutions are still pretty new, so it’s only natural to have questions. Don’t worry though, to help you navigate the 10DLC space and build a rock-solid migration strategy we’ve got some tools to help you out. Take a look at our introductory webinar: 10DLC: the new standard in…, Download the guide, Still got questions? Then our Q&A blog, with advice from Sinch and TCR experts, is here to save the day. We caught up with Jonathan Vimont, Director of US Messaging at Sinch, Stefan Heller, VP of Business Development, at The Campaign Registry (TCR) to address some FAQs. What is 10DLC, and what features does it support? 10–Digit Long Codes are a carrier sanctioned cost-effective solution…, 1. Customer enters their Campaign details in the Sinch Portal and requests numbers, Vetting is encouraged and will be required for non-Russel 3000 companies: 1-5 business days , Sinch registers Brand and Campaign(s) via TCR Sinch procures and provisions the 10DLC number(s) to clients account , 2. Customer has existing numbers and wants to port/migrate to Sinch: 3-5 business days, Customer purchases their own numbers and provides them to Sinch along with their Letter of Authorisation (LOA) NOTE: Customer MUST specify if Voice will be included, if so they will need to set up or add these numbers to Sinch Inbound Voice in parallel Same process applies as Step 1 for Customer to register Brand and Campaigns via TCR Sinch submits the SMS port and A2P Tagging (OSR updates…, 3. Customer handles 10DLC procurement and campaign registration but would like to use Sinch as their secondary provider: 3-5 business days , Customer must establish relationship and create Account as CSP within TCR Customer registers Brands and Campaign(s) via TCR and selects Sinch as their ‘provider’ or DCA Customer provides Sinch with the list of 10DLCs and NetNumber ID (NNID) (if holding their own NNID**) Sinch Reviews and Approves both the Brand and Campaign(s) and sets up the numbers with the Carriers defined by the Customer…, Text Enablement, :, $10,000 USD if a company or a message sender text-enables a 10-digit NANP telephone number and sends messages before verification of message sender ownership , Grey Route:, $10 USD per message for each 10DLC message sent by a message sender via a P2P route, Program Evasion, : $1,000 USD for each unique instance of a violation (snowshoeing, dynamic routing, spam filter evasion, unapproved campaign service), Content violation:, $10,000 USD for the third and any subsequent notification of a violation from the same content provider For more 10DLC-related questions, we recommend checking out our 10DLC knowledge base articles on the Sinch Support site.
Tipo: Artículo
Say " Bem-vindos" to the new Sinchers in Brazil! Here at Sinch, we’re proud to be a truly global company with a local presence in more than 40 countries. We have offices in places as diverse as Sweden, the US, and Singapore. And now, São Paulo, Brazil! We're thrilled to announce that TWW, which we acquired in October 2019, is now officially rebranded and an integrated part of Sinch. And to…
Tipo: Artículo
We discussed in a previous blog how SMS is set to continue prospering and developing into the 5G era. In this post, we take a look at the importance of providing a fully functional SMS service for inbound roaming subscribers and why the SMSF must be an integral part of any operator’s 5G strategy. Operators who want to offer seamless continuity of service to visiting subscribers; and maximize…
Tipo: Artículo
The digital interactions such as logins, signups, and sensitive online transactions that make life so much easier come with a duty to protect from risks, and criminal exploitation. Addressing these risks is a complex problem, one that can only be managed by balancing vulnerability, cost, and usability solutions alongside a program of continuous improvement. So what’s the best way to keep your…, D, ifficult, for, identity, thieves to disrupt, intercept, or emulate, SIM, card, s, Despite what some headlines claim, exploiting mobile operator vulnerabilities at scale is not easy. For most use-cases, SIM-based verification offers the best balance of usability, cost, and risk. , C, an’t, be confused or, forgotten, With SIM-based verification, each user has only a single identity — his or her SIM-based identity. Users only need to keep their mobile devices secure rather than memorize and frequently update multiple passwords. , R, emains, consistent, regard, less of, geograph, y, SIM-based verification methods are available in any country. This, coupled with high mobile phone penetration, makes improving security via one simple method a breeze. , S, uits, many, use, case, requirements, SIM-based verification falls into three categories: SMS, voice, and data. These methods are used to onboard and authenticate new users to authorized services or add a second factor of authentication (2FA) onto existing identity solutions. Whether interacting with users directly or indirectly, all types of verification have a role to play. Identity and access management is a huge topic.…, Sign up using our self-service portal, to start testing in minutes.
Tipo: Artículo
In the post-COVID era, consumers have flocked to digital solutions to solve the challenges of social distancing. But why is mobile customer experience such a critical focus for product managers? It’s because brands can create highly personalized, valuable interactions on mobile, and do so at scale. Each customer’s experience can look and feel unique by leveraging powerful tools like chatbots ,…, , Mobile has become a center of innovation and that will continue with the rollout of 5G One-way notifications and reminders deliver value (think: a notification about a service interruption from your telco), but there’s so much more companies can do to build engagement and loyalty on mobile using two-way conversations and richer engagement. Now that people have access to faster, more reliable…
Tipo: Artículo
SMS is the most powerful communication channel for engaging your customers. In fact, people are 35x more likely to read a text message than an email. (There’s a good chance you already knew that, given that you’re on the Sinch blog, and we love talking about the power of those 160 characters .) But it’s a double-edged sword – customers love those text messages, and are so much more…, Download the free guide here!, And as always, make sure to consult with your legal team to see how the SMS compliance laws and regulations affect your business specifically. Sinch can’t offer legal advice – just tons of industry knowledge , great messaging products , and an entire resource section for your developers ., (Need something else? Our sales team can probably handle it, get in touch! ) ,
Tipo: Artículo
Inbox overload is a persistent problem—driving down success rates for spendy marketing campaigns. What works better? Messaging consumers on their phones drives quick responses unmatched by other channels. Yet, SMS—which stands for “short message service,” better known as text messaging —remains underutilized. Here’s our guide to mastering the costs, capabilities, and potential payback of your SMS…
Tipo: Artículo
If you’re in or you follow the mobile messaging industry, you’ve probably heard of “10DLC.” And for those of you who are not part of this vast mobile ecosystem, 10DLC is some off-the-wall terminology that doesn’t make much sense. But either way, you should probably learn about 10DLC . Because these 10-Digit Long Codes are changing the way we send messages in the U.S. In the past, only…, Talk to our experts and get early access to Sinch 10DLC!,
Tipo: Artículo
