The single most efficient way to increase marketing effectiveness
In order to capture the data that would make a significant impact you should stop from time to time and look at where the platforms responsible for the majority of your marketing budget are heading. Where is Google focusing most of their development? What was Facebook’s latest revelation? For the purpose of this post, I have consolidated some of these marketing giants’ remarkable shifts, in this context, for you:
To say that the progress in the past years has repeatedly made your first party data more important is understating it times 10 million. Nearly every piece that used to delimit good marketing from great is now automated, putting the pressure on what value you provide instead.
In 2020, feeding the right data relevant to your core objective is the most efficient way to increase the effectiveness of your marketing efforts.
Take a look at the reports provided by your agency or marketing team, and reflect on what you base your marketing efficiency on. There’s a general perception that ROI is a good thing; without much appreciation of what it means. Most people talk as if profit and ROI are interchangeable when they’re not. Reason being that ROI doesn’t reflect which campaign is most effectively reaching your objective, only the efficiency in doing so. Secondly, what value does ROI add, if measured on gross sales when your business targets are based on net profit?
Not contemplating our data foundation nor diminishing return curves, we are prone to errors in our decision-making from lacking data that directly addresses the question at hand, a form of sampling bias.
Don’t settle for a correlation if you want causation
At Precis, our clients’ core objectives are mainly to maximise the effectiveness of something, being brand consideration, profits or LTV. What you, as an executive or digital marketing manager must be hyper-aware of, is whether your marketing optimization technology is fed data relevant to the objectives you would like to reach. As explained earlier in this post, the marketing giants have automated the manual processes that required lots of head count years ago – but they won’t know what they don’t know and that is the core objective of your company. Still, most companies don’t invest resources in activating the business-critical values to optimize more accurately. Worse, they may not even collect the data needed to do so.
We end up seeing banks optimise for leads instead of issued loans and retailers optimise for revenue instead of net profit. This is a consequence of focusing on things that can be measured easily. Not what’s important. This is not to say that a company trying to optimise for profitability (or predict LTV, for that matter) can’t solve a similar problem using the data they do have. But that would require even more expertise and resources, in order to shape the proxy-based outcomes into accurately reflecting the business objective.
If you’re not measuring the long-term, think again
Think back on when you launched your last campaign. A common practice when launching a campaign is to measure the outcome during, then immediately after, the campaign was launched. This is a flawed approach, considering that marketing effects can last far longer.
With access to customer data, measurement of long-term outcomes is arguably more accurate. It allows us to see what a particular customer has bought and when, and how much revenue or profit resulted over different time periods. This is what we consider the Lifetime Value (LTV) of that customer.
LTV will never be a universal formula or tactic, it’s calculated and activated in many different ways. In some companies, it’s mainly used as a key performance indicator towards investors, in others it’s heavily integrated in their bidding or reporting and for some it remains an uninvestigated path. Our recommended approach is to use a machine learning model to compute the relative difference between customers, while the opposite end of this spectrum is that you’d start by setting a rule for how different your customers are (your customers LTV might be more nuanced than “High” and “Low”). What happens when a human sets these relative differences? It might enhance the comprehensibility, but it’s most likely averaged, missing hidden dimensions in the data and static. If covid-19 didn’t tell you, customer behavior can shift from one day to another, teaching all of us the gains of responding quickly to shifts.
A common assumption for those only moderately familiar with data science, is that the bulk of a predictive analytics project is building the model. In reality, at Precis we typically spend 60-70% of our time on understanding, cleaning, preparing, and validating the data into a useful and efficient format. This is known as Extract, Transform, and Load (ETL) processes, and when spent time on, the modeling process runs more smoothly and produces better results.
The final step of any LTV model is putting it to good use. Contrary to popular belief, its sole purpose is not only to report how sustainable your business is. At Precis, we’ve built a scheduler for daily calculations of our models that are then exported to the main marketing platforms through pre-configured data pipelines. This enables us to inform our clients’ BI or marketing teams of what marketing campaigns contribute to high-LTV customers, as well as target customers labeled “churn-risk” or “one-off” with the purpose of increasing LTV. For a fully automated approach, we are able to include LTV as a signal in the bidding process, allowing bid engines to take into account subsequent purchases from a user across devices and channels, as opposed to only intent or same-device purchases made online. Needless to say, this adds a tremendous value and a competitive advantage to our clients’ marketing capabilities.
How can technology be used to reframe the problem?
The technology to collect and store data is declining in cost, and the tools to analyze that data are even cheaper (in many cases free), so what opportunities emerge from this?
If you didn’t read this part of the blog series prior to this one, please return when you have. I will wait. Following Matilda’s post, Cloud is not meant to just run databases or storage spaces, that in itself is not a particularly new use of Cloud technology. What you can do with the data is the true differentiator. At Precis, we’re making new applications possible in the Cloud by marrying technology with people and business, then sharing it exclusively with our clients. Making use of immense computing power and integrated solutions for activation, a marketing data warehouse in the Cloud doesn’t just solve “old use cases” but comes with exciting opportunities to enable new ones too.
1. Infuse new thinking of what data and models your marketing KPIs are based on
Project the long-term value of marketing
My ask of you regarding LTV: have your agency or analytics team analyze the different long-term multipliers and include that dimension in your marketing efforts.
Get comfortable with using predictive models to fill your data gaps
Don’t get trapped in the crazy out there stuff, there are many applications where predictive models outstrip rule-based models, forecasting outcomes is one of them.
2. Data will never be perfect, but don’t be sloppy in your preparation
Measure the cause and effect of what you want to achieve (not nearly what you want to achieve)
Instead of reporting what’s easily measured, integrate your business objective into ad and analytics platforms. In a future with less data, optimising on the right data becomes even more important.
Don’t aggregate when you don’t have to
Your organization may be activating data, but if it’s aggregated to the point where it distorts the reality underneath, important nuances will be lost and you will lose out on potential.
3. Stop capturing data if you can’t take full advantage of it
Traditional tools do not solve modern problems
Lots of data may be available, but with a culture rooted in top-down decision making and traditional tools like weekly reports means you can’t take full advantage of it.
Ask (for consent), and you shall receive
Expectations from users are high, and so are the stakes for compliance. Ensure the data captured is handled responsibly and in a transparent manner.
Now that we’ve fed our marketing platforms with a more valuable signal to our business, how should credit be divided between different channels and which customer interaction counts? In our next and final blog post of this series our Senior Data Scientist, Pierre, explains the key ingredients to unlocking the power of a well-known marketing challenge; Attribution.