.svg)
Data analytics have matured from a passive look-back on previous efforts to an active indicator of future performance, particularly with the rise of artificial intelligence (A.I.) and Machine Learning. The ability to collect and analyse massive amounts of data in real time, using Machine Learning, has given commercial decision makers the ability to better understand and predict their customers' behaviour, anticipating new demands and potential issues before they arise.
But it often seems that this potential is hard to fully unlock. For digital publishers benefitting from the recent migration of physical newspaper readers to online news, supplemented by the COVID-bump in online traffic, there is no shortage of data to collect. Yet, acting on this data is another whole process, one that directly affects the bottom line of digital media companies. Being able to turn insight into action through effective commercial strategies leads to not only higher conversion rates, but longer subscription lifetimes and reduced customer churn too.
That’s why we, with the help of our machine learning partner Vidora, have decided to answer 5 of the most common questions surrounding the actioning of data, helping you demystify the task and begin extracting maximum value from your data sets.
1 - What do I need to do before acting on my data?
Data seems to be at the heart of almost everything we do these days. But it's worth noting that not all data sets are created equal. Collecting information about your readers through registration walls is just the first step in an on-going process of data set expansion. Data is worthless if it can’t be understood or utilised, and so before looking to implement actions it is important to take the following 3 steps:
Clean the raw data: Especially when data is collected from multiple different sources, it is important to unify your data sets to give a complete picture of each individual reader. With all your data in one place, it becomes much easier to cleanse the data of repeat inputs, erroneous details and irrelevant information. You’re left with an accurate and insightful pool of data from which to base your decision-making.
Connect the datasets: It's likely that your data management solution, CRM, marketing software, billings provider and analytics tool are all separate systems. For digital publishers, a single tool from which to operate gives you the ability to operate in a far faster and accurate way, accessing every aspect of your customer’s journey from a central hub. Subscription experience solutions such as ours connect with every tool in the MarTech stack to give you oversight of your entire digital ecosystem.
Enrich the database: For those looking to take their data to the next level, machine learning uses all the information available to continuously update and uncover new predictions and insights. The best part about machine learning is that the more data you add to your models, the better it operates, and the more accurate and informative your insights become. Many publishers might not know where to start when it comes to implementing machine learning within your organisation, so in the next section we’ll shed some light on the options available.
2 - How can I operationalise machine learning? Should I build my own models, hire an agency or use a vendor?
There is no ‘One Way’ to incorporate Machine Learning into your business, and as with many decisions each option has trade offs. And given that only 22% of companies using Machine Learning have deployed a model, choosing the option best for you is crucial. That being said, here are three common ways different businesses have worked with machine learning:
Build In-House: This option provides the most customisation and flexibility of any option. Because all decisions are made and all technical considerations done by an internal team, it allows the models and predictions to be completely bespoke to the given business. However, this often comes at the cost of higher expenses and longer time investments.
Hire an Agency: When companies have limited resources to build everything in-house, Agencies can be another option. Agencies can offload the work done by an internal team, and often bring additional machine learning expertise to the table which can eliminate some of the burden of doing everything with an internal team. However, fees and the need to collaborate with an external team can often limit the scope of specific engagements.
Use a Machine Learning Vendor: Vendors can often prove a great option by focusing on the specific kinds of Machine Learning needed at your company. For example, our partner Vidora has a platform that automates raw customer data into Machine Learning Predictions quickly and easily. Platforms like this can often save both time and costs by optimizing the specific machine learning Workflow. However, with any data platform, it’s important that the Machine Learning Vendor easily integrates into your existing data workflow and accurately produces the specific predictions you need.
3 - How do I take action on my data insights?
By this point you should have a fairly comprehensive and complete bank of data from which to inform your decision-making. Now comes the fun part: turning this information into results along every point of the customer journey.
It’s crucial that you use your data to maximise the effectiveness of your on-site experiences and conversion efforts. One way of doing this is by building intelligent paywalls that target reader’s at the moment they are most likely to convert, based on their details, on-site behaviour and attributes.
For example, since news readers are unlikely to buy a subscription when they are reading one of your articles for the first time, at 9am on a mobile, this would be an inappropriate and ineffective time to hit them with a paywall. But if they have visited your site multiple times in the last month and are reading at midday from their desktop, you may have a much higher chance of conversion. Dynamic paywall strategies, based on insights from your data sets, allow for more accurate value alignment between your product offering and the customer, which will turbocharge your conversion rate.
And it is worth noting that conversion strategies are as much about when not to hit a reader with a paywall as they are when to, so tailor your rules accordingly. Common data points that you can base your users’ experiences around include:
- Attributes
- Device type (mobile, laptop, etc)
- Time of session
- Channel from which they access content (Google search, social media, email click-through, etc)
- Activity
- Monthly sessions
- Session length
- Content preferences
4 - Can I experiment with my commercial strategies without risking my KPIs?
Yes.
We would go as far as to say that if you aren’t testing and experimenting your commercial strategies regularly, then you’re unlikely to hit any KPIs in the long-run. Data can tell you a lot about your readers in theory, but getting results in practice involves a certain level of flexibility in mindset and iteration in actions.
A/B testing allows you to experiment with different aspects of your product offering to see what sticks and what doesn’t. Subscription experience management solutions enable commercial decision makers to segment users in buckets of readers with similar characteristics or behaviours, before running different experiences for each. The result: the optimal UX and product packages that convert and retain.
We suggest experimenting with the following aspects of your offering:
- Segments - group readers by shared attributes and on-site activity
- Messaging - see what copy resonates with which audiences best
- Paywalls - test a mixture of different paywall strategies, including soft, hard and metered
- Timing - find out at which times of the day your readers are most likely to bounce/convert/sign-up/engage/etc
- Products - play around with content bundles, creating better product offers and upsell opportunities
5 - What next?
Keep iterating. One thing we can guarantee is that the moment your mindset becomes fixed, your performance will begin to suffer. Data collection is an ongoing process, and machine learning will keep your insights relevant even as new information about your readers is added to the model. In fact, it will become more accurate and complete as customer profiles are uncovered. Take the opportunity to revisit your datawalls to make sure you are progressively profiling to add new and valuable fields to your data sets, starting with Name and Email, continuing on to Job Title and Company Sector. In the end, the more data the better.
And this outlook applies to your actions too. Consumers change and commercial strategies need to keep up with the latest trends in behaviour. Particularly notice the subscription journeys that are especially strong at achieving your desired outcomes, and double down on these. On the other side, find the paths, messaging, product packages, etc, that aren’t working and use the data available to try to find out why. Rethink these journeys and go again. Ultimately, it's all about failing hard and failing fast, so you can move on to succeeding as soon as possible.
Looking to connect your data sets and action on these insights? We can help.
