RESPECT for the DIY Mar-Tech + ML Pioneers
This is a post long overdue.
Reflecting on 2020, I want to say on behalf of the Syntasa team that we are incredibly fortunate and grateful to have been part of the journey with so many innovative businesses and leaders in the last year. We are happy and proud to celebrate the work that has completely embraced Covid-era ecommerce by innovating and delivering programs to improve their company’s Customer Acquisition and Customer Experience metrics!
Here are seven of the most interesting programs we’ve witnessed up close this year (not in any particular order):
1. Ecommerce – Conversion rate optimization (CRO)
You’d think that one of the oldest and most established branches of e-Commerce would lack on-going innovation. Surprisingly, it has brought the most potential for immediate revenue generation for anyone selling online.
The point of CRO is to ensure you convert as many prospects as you can, whilst you have their attention via your content/offerings, closest to the time of transaction. The two most common areas of focus in this category have been:
* Cart Abandoners – Smart retargeting via email and advertising to the people that have abandoned a shopping cart. Depending on the size of average weekly revenue and average basket size, we have seen customers generate 2-5% of total daily revenue via a simple automated program like this with zero incremental on-going effort.
* Basket recommendations – Suggest relevant product and content recommendations based on a customer’s online journey by combing browsing sessions from the recent past. Moving beyond recommending products based on static merchandising and bundling, smart recommendations produced using machine learning (ML) techniques applied on Digital Clickstream data sets, produce up to 300% increases in basket attachment rates on their ecommerce websites.
2. Marketing – Propensity based retargeting campaigns
Marketing campaigns most commonly use various business rules to select the audiences to target. Whilst this is an established approach, the typical challenge is to find the right set of business rules, in conjunction with A/B testing, that will produce large enough audiences to measure, along with producing sustainable conversion rates.
We’ve seen customers using automated Machine-Learning based retargeting campaigns produce 15-20% better results in conversion rates, with up to 200% increased audience sizes—while simultaneously reducing marketing spend by 80%; ML propensity-based audiences can select very granular targets as opposed to the scatter gun approach of the past.
The final piece in the puzzle for Propensity based campaigns is to share propensity scores downstream into target marketing eco-systems. For ongoing campaigns, it’s important to quickly setup data pipelines that continuously activate audiences for the best performance. Beware that this step can be prohibitive for the uninitiated in terms of setup cost.
3. Analytics – Journey mapping & Fraud prevention
Websites, products and content are changing constantly. In our conversations, digital executives express a deep desire to have a complete view of how their customers are using their websites and apps–not simply each individual browsing session, but also across multiple sessions spanning varying time periods (1 week, 2 weeks, 1 month, 3 months). Adding transaction data to this information makes it vastly more actionable, considering a transaction is the positive outcome from a browsing session.
A journey mapping solution has not only provided analytical and insightful information to digital executives, but also helped create meaningful KPIs to track the performance of the business based on how the digital visitors continue to interact with the web property. Any dip or spike in these KPIs point to actionable information on the intentional or unintentional changes that may have occurred recently, and how it has impacted revenue numbers.
Another key area in analytics we’ve seen recently is the use of analytics for fraud detection and prevention. For obvious reasons, a user’s behavior on the website, or lack thereof, is an excellent indicator for whether the activity is perhaps fraudulent or BOT traffic. A program to include this fraud metric in an existing ecommerce platform has yielded more than $2.5M in quarterly revenue and savings (revenue from more timely processing of previous falsely flagged orders, and savings from chargebacks and fees that ecommerce payment partners would have charged had the transactions turned out to be fraudulent).
4. Customer Experience – Home page and Care
When it comes to Customer Experience, personalization is the name of the game, and the biggest bang for the buck is often from seemingly simple, yet significant interaction points on digital properties. In 2020, we have seen leaders pay special attention to personalizing the home page and customer care pages.
Home page personalization can take many forms, but the two most common methods we’ve seen are to have a handful of landing pages designed specifically catering to your website’s biggest audiences, and/or to have containers on the home page that show tailored content (products, content, creative) based on the visitor’s journey. The key here is to have journeys defined for each visitor (non first-time visitors, obviously) that reflect their past behavior on the website and their transaction history in CRM. The better the journeys, the more likely the personalized home page will engage the visitor for longer and will create up to 30% better click through rates CTR.
Care personalization can take many forms as well, but again the lowest hanging fruit on the digital channel is to promote the most likely content at the top of the help or FAQ pages, based on the recent behavior of the visitor on the website. Combining digital behavior with transaction history can provide greater value, but bear in mind that it may not be as easy, and it could become computationally costly.
5. CRM – Churn and Lapsing
As with many other business functions, we’ve seen Customer Relationship Management (CRM) teams be super switched-on when it comes to using transactional, demographic, and other meta information to identify candidates for their next campaign. In these times of the pandemic, many businesses have paid extra special attention to inactive and churn-risk customers because it’s difficult to replace these with new customers given the market conditions.
We have witnessed how creating better churn predictor models and better ML models can identify lapsing customers and provide a 5x upside to businesses, especially those that rely on repeat customers or subscription pricing.
Using ML models to identify the customers at the highest risk of lapsing and launching tailored CRM or retargeting campaigns have consistently produced strong results.
6. IT – Advanced Data management
IT teams have embraced the cloud already, but 2020 has further accelerated adoption and use-case implementation. Key adoption barriers from the past have been overcome, like the security and governance concerns, and the management of personally identifiable information (PII).
Syntasa has been fortunate to support IT departments with establishing advanced data management protocols on GCP AWS and Azure, and thereby bringing all the promised returns on investments like reduction in TCO, just-in-time provisioning of resources, reduction in red-tape for launching new business programs and so on.
7. Supply chain – Forecasting
Needless to say, a strong and robust supply chain and sales forecast has been the lynchpin for many businesses in surviving and thriving during the pandemic lockdowns. While almost everybody understands the principles of forecasting, during the pandemic the techniques that businesses have relied upon for decades all of a sudden just stopped working. The metrics that drove these forecasts, such as orders, transactions, and footfall, changed dramatically in the last 9 months.
We have found that digital activity, in the form of clickstream, has provided a reliable way to account for new behavior, while also providing a more granular level of insight into which products pages and product combinations are getting the most views at a certain time of the week and so on.
While supply chain, sales, and inventory forecasting models are specialized fields with established statistics practices, adding a new data source of digital behavioral data has proven a successful tactic. And with Brexit looming in the UK, businesses there must prepare for further disruption to the performance of their established modeling techniques. During these uncertain times, it makes even more sense to incorporate any data sources that bring the footfall behavior from the digital world into forecasting.
It has been an incredibly rewarding year for Syntasa, and we are excited and passionate about sharing the power of digital behavior. If you’re curious to know more about any of the work above, or if you’d like to brainstorm with some incredibly awesome and highly experienced digital practitioners in the Syntasa family, we’d love to have a conversation.
