Part 2: How does Smartify work?

In part 1 of this trilogy we already told you about what Smartify is and why it will benefit your newsroom. In this part (part 2), we will dive a little more detailed into how Smartify was build and how the system actually works. The last part of this trilogy, part 3, is about the future of editorial analytics and Smartify and will be filled with predictions from our CEO Erik van Heeswijk.

Our CMO, Rutger Verhoeven, chats with our CDO, Ilija Susa

From day one we designed smartocto to be useful in the newsroom. Because - let’s be honest - fancy numbers, pretty graphs and lists of the best performing stories on a big screen might look sexy, but in reality and in practical terms, they do little more than give your ego a nice massage.

In the past we’ve had many conversations with newsrooms and storytellers where they’ve said something related to this: having data wasn’t the problem, but knowing what to do with it is. And that insight was one of the main reasons we created smart notifications - to help you go from data shock to action stations. To simply tell you what the best course of action is for every story, at any given time, and always, always, in relation to your business model.

But we also have to admit that sometimes newsrooms are a hectic place to be - especially on days when the world or your local area seem to be on fire (literally or figuratively) and time is a scarce resource.

Times like these, stories get published thick and fast and there is hardly time for reflection - let alone to keep an overview.

So, we realised that our editorial analytics system could benefit from a different, complementary approach - one that was better connected to what journalists, storytellers and editors actually do. A super practical solution to help every day, but particularly when it’s really manic.

And that’s how Smartify was invented. This new technology doesn’t take the story as the key point of focus, but rather puts editorial action at its heart. And with this reversed approach it becomes possible to match specific stories with concrete editorial tasks.

Now, I can hear the cogs turning. I can hear you asking, how does this work? Excellent question. The best person to answer this is Ilija Susa - CDO, one of smartocto’s founders and also the brains behind the tech that powers Smartify.

As you might imagine, Smartify is a system that’s built on top of the existing platform of smartocto. It combines different types of collected information which varies from the raw data that clients share with us (which is actually the beginning of the data flow) to the notifications that we produce (at the end of the existing data pipeline).

Ilija puts it like this:

"Smartify is like the extended brain of the smartocto system. Smart and self learning. It looks at all the stories produced and published, and creates a new intelligence layer for editors on top of that."

The power and beauty of Smartify is that there are predictions at the story level for every suggested editorial task. One of the biggest challenges is to make these predictions as accurate as possible. This is of course work in progress because the system needs data to learn and become more precise, but the obvious upside is that the more you use Smartify, the more accurate it gets.

Here’s Ilija again:

"To simplify in non-technical language: at the beginning of the process we classify a new story by its different features (like its potential reach) and by various editorial decisions (e.g. the time and day of publication, and its lexical and readability characteristics). Those features are then compared with other previously published stories and, based on their performance and results, we can predict what will happen with new stories."

"What we have seen so far is that predictions are working well enough for editors to be able to make decisions based on them. So, for example, if you want to look for a story to share on a Facebook page, we will suggest the top 3 stories to share and in 90% of cases, the order will be accurate. If you are looking at the predicted metrics, the results are less accurate (around 60%) but in the next version, which is coming soon, we will improve this."

The power of smartocto’s notification system is that it works automatically, analysing everything without the user even noticing. All they see are the resultant notifications - and they are able to set these notifications to ‘done’. This isn’t just a novelty - it fulfils an important purpose too. Let’s find out why.

Ilija: Smartocto is a very powerful editorial analytics tool with actionable notifications based on client data. With Smartify we introduce another layer of notifications that require very precise data. We’re actually creating notifications on notifications. An example of such a high-level notification would be: “This week you finished 15% fewer tasks, therefore you missed 152,000 views and 3,500 user engagement actions”. To have that accuracy we need to collect as much data as we can about how clients are responding to our smart notifications. Setting a task to ‘done’ gives us the means to finesse the predictions in Smartify.”

By now you are probably wondering why a super smart system like smartocto isn’t able to see for itself if a particular action has been done. Let’s ask Ilija.

"Actually we are doing that, and I can say that we are doing it very efficiently. But we can’t do it for all the types of actions that we suggest. To give you a good example, we can’t automatically see if the tip “Enrich your story” has been followed up or not. And the simple reason is that we don’t know what the ‘enriching’ component has been."

"Also, a very important result of the whole process is that when you look at the percentages of finished tasks, split by type, you can see which notifications are not configured correctly or not in use at all. Newsrooms configure a set of notifications of type, like “Share on Twitter” but it might be the case that they aren’t actually using Twitter, or only doing it now and then. So, looking at the percentage of ‘done’ tasks can help to configure the system more efficiently for the needs of that specific newsroom."

There is often a big difference between devising the technology and actually implementing it. What you initially come up with cannot always be made. And during the process of creating Smartify, Ilija’s development team faced multiple challenges. Let’s ask him what the biggest challenge in building Smartify has been, shall we?

"There were many. Creating the predictions and the algorithm to identify when a task is done was tricky for sure. But, maybe the most challenging task was to create a ‘dry-run’ feature in the notification manager which calculates a number of notifications that the user will get based on the parameters and thresholds he can enter himself. This is very important because we don’t want to spam users with notifications. But, at the same time it’s not easy to predict the approximate number of notifications for some random combination of parameters. But the good news is: we actually cracked it!"

We’re pleased to say that Smartify is already gaining a lot of traction in the market - and much of the reason for that is because it’s so simple and straightforward.

All editors, storytellers and journalists have liked working with the first version of Smartify because they immediately understand what the tech does for them and can see how they can use it in their daily practice - and that’s because it’s well adjusted to the workflow of the newsroom.

But tech development relies on the feedback loop, there must be a lot of potential new features and improvements that are in the pipeline, right? Let’s see if Ilija is willing to share some of that juicy stuff with us….

"A lot of new ideas were raised during the development process. In the next few months, Smartify will be enriched with much more useful information and filters. But while there’s still plenty of things to add and improve, we’re committed to keeping its simplicity. Also, I can predict that we will produce a huge amount of useful notifications [algorithms] based on the data and usage of Smartify, for specific departments or roles within a news organisation."

"And aside from the development of multiple smart notifications, maybe the most exciting thing we will develop to help enrich the predictions is a ‘ticketing’ system which will predict how much FB engagement you can expect if you get a certain amount of impressions. It means that we will predict possible results of a FB campaign before a story is actually published on Facebook, which means we can get the insights before Facebook itself!"

During the pilot phase, when Smartify was rolled out to some clients, we collected a lot of information. We discussed the progress and got feedback in weekly sessions, where our clients also shared their comments and made suggestions.

One thing that editors requested were more filter options. And this is of course very understandable. If you want to place an article on Facebook, the system can advise you to use a story, but first extensive filtering is required to make sure the right one is selected.

And this depends on the publisher - each has their own needs and requirements. For example:

• Newsroom A: I want to see all the stories that need to go to Facebook but NOT the articles on Soccer
• Newsroom B: I want to see all the articles that need an update but only those that ALSO have the notification that there were a lot of comments on the Facebook post
• Newsroom C: I want to see which articles need to go in the newsletter but NOT the ones covering the topic on POLITICS

These filters make the selection more accurate and therefore useful, but while filtering might look easy, it’s not. So let’s ask Ilija why this is so complicated to build.

"Filters are very important. We noticed during the testing period that for some brands, we are making better predictions in certain sections than others because of their users' behaviour. Not all types of stories are equally important to all departments and publications. One of the biggest challenges is to set up a filtering system that can support different use cases and user needs equally successfully."

Earlier, Ilija described Smartify as an extended brain that you can add to your system. So naturally that leaves us with one more question: would it be possible to add this brain to other systems as well, like an Intel Inside solution? (Making it ‘Smartify Inside’?!) And, would that bring intelligence to these systems? What does Ilija think of that?

"Well that’s an interesting thought. And yes in theory it’s totally conceivable. But we need to optimise this existing system more first and tackle all the challenges that newsrooms give us. If we solve all of that in 2022 I’m sure the first solutions for this Smartify inside will become reality. What we’ve accomplished so far is a bit of a paradigm shift for the industry and time-poor editors, and I don’t think it will be long before editors will be telling their journalists to “just Smartify it!”."

As a final note, you might be wondering how this impacts other technology…

Dennis Laupman of Pluxbox - who joined us for the recent webinar ‘Smartify your system’ - explains why he believes that adding intelligence to various systems is the next big thing in digital publishing:

"The world is changing fast. More and more collaboration is being sought between different media departments, like tv, radio and online. We have built a platform for distribution and creation that is a very scalable outcome of that change. With this platform we love to work together with all sorts of tech to make a stronger tool box. We believe in vendors who are specialised in skills that we are not, but we can help connect the dots and work together. That’s why we are enthusiastic about the Smartify solution and we’re looking forward to combining this tech in our platform and help our clients overcome their challenges in connecting better with their audiences."

Stay tuned!

In part three of this series, our CEO, Erik van Heeswijk talks about the future of newsroom and editorial analytics - and explains how and why Smartify fits into that vision.

Are you ready for a demo, then please allow us to show you how Smartify really is the next step in editorial analytics and how it can help your newsroom! In the mean time, download our Smartify whitepaper!