While data science cannot explicitly help you acquire more customers on command, it can tell you when, where and how you should pitch your target audience to maximise yields and minimise waste.
Rakuten Marketing estimated that on average 26% of marketing budgets are wasted on ineffective channels and strategies. There’s even more money left on the table because spreadsheets alone can’t tell who your best customer is and which touchpoint will be crucial for “sealing the deal” with them.
So if you are struggling to understand where the highest ROI is, data science may provide you with the much-needed answers. Below are five data science use cases for the budget-conscious marketer.
Why not look at Econsultancy’s Fast Track: Advanced Data & Analytics Training Course (3 days)
1. Predict and optimise conversions
Bigger reach increases the chances of a sale. However, higher traffic does not always translate to higher conversions; it is relevancy that matters more. Paraphrasing the Pareto principle: 80% of your sales come from 20% (or less) of your audience.
Data science attempts to change that odds by helping you understand what type of exposure or actions will lead to higher conversions. This means that when you have a limited budget, you can focus on maximising the performance of your current assets – landing pages, sales funnels, cross- and upsell offers, etc. A machine learning algorithm can gauge what impacts conversions and suggest your highest performing marketing asset(s) for further optimisation.
A group of American scientists identified that the following factors majorly increase the odds of conversion in ecommerce:
- Page dwell time: the probability of conversion reaches its highest when a shopper spends 50 seconds on the product page.
- User type: Registered (returning) users are more likely to re-purchase. That’s a no-brainer; however, a lot of companies choose to focus their budgets on acquiring new audiences, rather than reactivating returning customers.
- Click entropy: Eight different clicks under one customer query on the website indicates a high likelihood of conversion.
- Click position: The highest sale probability is when someone has clicked three items. But it plunges dramatically after the fifth click.
An example of this in practice is a custom machine learning model that Sarah Raven, an online gardening retailer, used. The algorithm identified the best-performing pages on the website and provided further insights for optimisation. Improving a single product page and directing most of the paid marketing spending towards it has helped the company increase their click-through-rate by 76% and boost transactions by 194%.
2. Improve lead scoring
A lot of your marketing efforts can go to waste if your sales team cannot identify and reach out to the right people at the right time. However, more than 40% of sales professionals admit that prospecting is the most challenging part of the sales process.
In this case, data science helps you connect the dots yet again and estimate which leads are the most likely to respond to offer A, B or C. This, in turn, enables you to get smarter with budget allocation, minimise the response time and reduce productivity loss. Instead of “guesstimation”, your teams will rely on data-backed insights when making those calls.
One case study from SalesWings has demonstrated that a predictive lead scoring system – an algorithm that automatically prioritises leads using CRM data – helped an education company secure $170,000 (approx. £130,000) more revenue over 1.5 months.
3. Reduce PPC budget waste with prescriptive analytics
Programmatic advertising has made significant progress in the past several years. Today, there’s no lack of martech tools and platforms that will help you secure the best media buys and help you run razor-sharp campaigns on autopilot. But what if you are a smaller company that merely wants to know how to spend a limited budget on a Google Adwords campaign?
This is the area where data science excels. Using the right data sources, you can build simple (and more complex) models to explore one pressing issue at a time. For instance, you can learn how seasonality or even weather will impact the buyer’s behaviours and PPC conversion rates respectively if you run a campaign at a certain time.
In a nutshell, that’s what predictive analytics does in marketing – estimates when the desired action will happen and what can impact it. Here’s another example: your model determined that Friday evenings bring in the most foot traffic to your store. Then you can increase your PPC budget and gain even more exposure from retargeting or localised ads. What’s more, the model can notify you of any change and automatically pause the campaign when conversions drop below a certain threshold.
Prescriptive analytics lets you apply such recommendations to a wider range of products and based on a bigger range of factors, automatically.
4. Real-time personalisation
Undeniably, advanced marketing personalisation is a huge factor for success: it generates $20 ROI for every $1 invested.
And there’s no lack of facets worth optimising for even higher returns cross-channel:
- Micro copy on landing pages
- Product and content recommendations
- Email marketing copy and offers
- Coupon codes and discounts
- Location-based advertising campaigns
- Real-time push or mobile app messages
Speaking of the latter, Starbucks did a particularly impressive job with mobile marketing personalisation. Their mobile reward program, already 15 million users strong and counting, serves as the main source for collecting all sorts of behavioural insights that are later translated into personalised real-time messages, offers and up-sell opportunities.
What’s more, the company’s current goal, according to Scott Maw, Chief Financial Officer and Executive Vice President, is to “start to capture digital relationships so that we can use digital marketing assets, including personalisation” to reach non-Starbucks Rewards customers too.
While the company does not explicitly state how successful their real-time mobile personalisation is, an independent data scientist, Andrea Xue, attempted to provide some estimates. In a recent post on Medium, she showcases how different factors affect conversions in Starbucks offers:
Similar models can be developed for virtually any type of personalisation campaign, meaning that you can focus more on perfecting your real-time offers based on the type of data you know on the customer (or a cluster of lookalike prospects).
5. Overall marketing budget allocation
Ultimately, data science can help you answer the question of “where to spend your budgets and why?” You can start small and explore simple use cases and questions:
- How does weather impact sales of product X?
- Which landing page converts best for audience A?
- How likely is it that leads in group B will convert?
Then you can scale your exploration towards more complex problems, using the previously uncovered insights and connecting additional data sources (internal and external). The best part is that you can re-use your cleansed and consolidated data for a variety of data science models and apply different models towards it. Doing so will allow you to build a sustainable, accurate and highly-effective analytics system that can provide predictive and prescriptive suggestions on marketing budget allocation.