Changes to our Twitter Monitoring App Twilert close
In Apps
Twilert is a Twitter monitoring tool that helps users to find the tweets and mentions that matter to their business without having to spend hours on Twitter. Users create searches for keywords, brand names or hashtags and receive an email alert when they are mentioned on Twitter.
We first developed the app back in 2008, as a way to help agencies and brands to effectively monitor their online reputation. Since then, we have made changes including relaunching the platform as a subscription service in 2013 and adding various new features such as searchable tweet history and realtime email alerts.
At the beginning of this year, we started to feel again that there was more we could do with Twilert to benefit our users. To see if our assumptions were right, we conducted user testing to discuss the feature and update ideas.
The results were incredibly interesting - some features which we thought were essential were clearly not and some which we had added as an afterthought proved to be invaluable. This, paired with the acquisition of one of our service providers, meant that Twilert needed some pretty big changes to its architecture and design if it was to continue as a market leader in Twitter monitoring and alert services.
Here's what we did:
New Features
Improved Geolocation filters
One of the things that came out of our Twilert user testing sessions was how useful targeted geolocation search would be for agencies and brands managing Twitter accounts. Twilert has always had the ability to filter results using the ‘near:’ and ‘within’ operators, (e.g. ‘near:London within:15km’) but this was not always accurate enough. Twitter Search also offers the ability to search by Country or City but we thought it would be much more useful, if a user could monitor a much smaller area such as an event, venue or street.
This led to the creation of a new Geolocation filter which allows users to circle an area on a map that they would like to view tweets from.
This could be a City, a postcode, or even something as specific as a restaurant or street, as shown in the example below.
Riffle integration
One of the features we were keen to develop for new Twilert, was in-app reporting that would allow users to analyse Twilerts and other Twitter users.
Due to time and resource constraints, this feature ended up staying in our backlog and is something we will look to implement later this year, but in the meantime we were able to integrate with a powerful chrome extension called Riffle.
For those who may not have heard of it, Riffle is an easy to use tool that allows you to click on any Twitter user and instantly see information on top mentions, hashtags they use, their number of followers and many other important insights.
Riffle results for ‘BBCNews’ after the Riffle icon is selected within the search resultsEnterprise plan
Within this version of Twilert we will also be offering an Enterprise plan alongside our usual Basic, Pro and Agency subscriptions. This has been built specifically with larger companies and agencies in mind, who need more than 10 user accounts, more searches or have specific requirements that may not be fulfilled within our other packages.
Search results preview
As well as an improved onboarding flow, we have also developed a ‘Live preview’ panel which allows users to preview the results as they are creating a search. We hope that this will help users to experiment with the search filters and find the most accurate results possible, which will save them from sifting through irrelevant results later on when they receive their alerts.
Live preview on the righthand side shows what a Twilert would look like with the selected search termsChanges to existing features
Extended free trial
After listening to comments from Twilert customers and user testing groups, we extended the 15 day trial period to 30 days. This is something that our users explained would help them to fully evaluate Twilert and also share it with their colleagues and clients where required.
Changes to the trial process
In order to continue to provide ongoing customer support and to justify the build of new features in the future, we also made the difficult decision to remove our free post-trial account.
As we have experienced before, there is often a sense that if it is online, it should be free. Unfortunately, this is not sustainable when you have high monthly running costs and want to continue to invest in developing and improving a product.
The Techie Part
As part of Twilert's latest round of upgrades, we rebuilt significant chunks of its backend architecture from scratch. As time went on, the previous architecture based around PiCloud and Elasticsearch (which we wrote a bit about here) didn't age well - PiCloud was bought by Dropbox back in November last year, who promptly proceeded to shut down the service. This, coupled with some issues in Elasticsearch's stability and data integrity, meant we decided to re-engineer our tweet processing platform from scratch.
Twilert's architecture now comprises of three tiers - the codegent Social Data Service (a highly scalable aggregation platform which connects to Twitter and can receive and process massive amounts of tweets at any given time, performing real-time analysis and de-duplication in the process), Twilert's backend and the Twilert app itself. We're hoping to open up access to the Social Data Service API in the future so that other developers can take advantage of Twilert's processing capabilities.
Both SDS and Twilert's backend are written in Python, backed up by Postgres. We use Celery for task scheduling (effectively replacing PiCloud) and Flask for our web APIs. We're still using Amazon SES to send email, and host our infrastructure on a mixture of horizontally scaled DigitalOcean and AWS virtual servers, moving away from Heroku.
The Twilert app has also been completely overhauled - we've moved away from a lot of server-side rendering and now do everything on the client using AngularJS.
As a result of all these upgrades, Twilert's user experience is vastly improved and we're able to offer a considerably more reliable service which will scale far beyond the 45 million or so tweets we currently store for our users.
We hope that current and future users will enjoy using the new features we have in store for Twilert over the next 6-12 months and we’ll be posting regularly on our progress and the lessons we've learnt along the way, so subscribe to our newsletter if you'd like to join us on the journey. Or, to give the new Twilert a go, head over and start a 30 day free trial of Twilert here.