analytics translator

Analytics Translators

Just another buzz word?

Photo from unsplash, courtesy of social cut

Mckinsey has called analytics translator the ‘new must-have role’ and has predicted we’ll need around 5 million of them. ⁽¹⁾

FIVE MILLION.

That’s more than the entire population of Los Angeles.

The problem is that no one really knows what the term ‘analytics translator’ means. Mckinsey seems to have slipped this term into usage in an article they published in the Harvard Business Review in 2018 ⁽¹⁾.

What’s in a job title? For the past ten years, we’ve struggled with the ambiguous title ‘data scientist’, then ‘citizen data scientist’. Now it’s ‘analytics translator’.

I largely ignore attempts to introduce new buzz words, but it seems this new term has taken root (unlike ‘citizen data scientist’). People seem truly excited by it. Within 18 months of publication of the HBR article, the University of Amsterdam had asked me to help design and teach a new ‘analytics translator training’ for professionals. The time had come for me to start formalizing the skill set for ‘analytics translator’.

Since Mckinsey was guilty of introducing the term, I wanted to understand what they might have meant by it. After a bit of digging, I found they were recycling content from their 2016 article, which discussed a shortage of ‘business analysts’⁽²⁾. So it seemed that during the 14 months between publication of these two Mckinsey articles, the term ‘business analyst’ had evolved into ‘analytics translator’. They must have decided it was time to introduce a new job title, never mind that we still haven’t clearly defined what a ‘data scientist’ is, despite the fact that half the planet is, was, or will soon be ‘data scientists’.

What Is An Analytics Translator?

Quite simply, then, an analytics translator, is someone who can understand business requirements and ‘data science’ possibilities. This skill set is actually extremely important. In a recent survey by O’reilly, 47% of respondents indicated this as one of the biggest challenges holding back adoption of AI / ML ⁽³⁾.

An analytics translator bridges the gap between business goals and data science possibilities. As per the Mckinsey article, “translators draw on their domain knowledge to help business leaders identify and prioritize their business problems, based on which will create the highest value when solved”.

In this sense, ‘analytics translator’ is a skill set and not necessarily a role or a job title. The concept is a bit like Ranger qualification in the Army. Some companies may have roles dedicated to gather requirements, and in this case the role itself could rightly be title ‘analytics translator’. It is not, however, an equivalent label for a ‘data scientist’, ‘machine learning engineer’, ‘statistician’ or ‘product owner’ role. At best, the term ‘analytics translator’ could substitute for the traditionally nebulous title ‘business analyst’.

Even if your company doesn’t use this job title, the skills related to the analytics translator concept are crucial, and you should make an effort to ensure that a large number of your staff are ‘fluent in analytics translator’

What Skills Does An Analytic Translator Need?

Facilitating the execution of data science projects within a business contact generally requires an understanding of business goals and processes; a high-level understanding of analytic vocabulary, techniques, technologies, and processes; and the ability to communicate cross-functionally.

In particular, the analytic translator skill set would generally include

Foundational Technical Understanding

Basics of classical statistics (regression, exploratory data analysis, hypothesis testing, correlation, etc)
Overview of common machine learning techniques (deep learning, SVMs, decision trees, adaptive boosting, clustering algorithms, etc)
Overview of technologies commonly used (programming languages, database concepts, deployment tools (docker, cloud), etc)
Understanding of techniques for model building & training

Foundations for Collaboration and Communication

Understanding of frameworks and tooling used by data science teams (scrum, kanban, Jira, git, etc)
Stakeholder management: setting expectations, building trust, change management
Techniques of communicating analytic results (as per Stephen Few, Cole Knaflic, etc)

Foundational Business Understanding

Understanding the goals and priorities of the diverse horizontal and vertical elements of organizations in which data science is only one team or department
Choosing the analytics projects most likely to deliver business value in the current economic and corporate environment

Regarding skills 1-4, a large number of data science teams are run by non-technical managers ⁽⁴⁾. These managers may not need to understand the technical details of team projects, but it is critical that they understand the difference between a high-risk, high-effort technique and a low-risk technique. They should understand the background terminology of project reports so they can focus on the elements that are unique to their business case. They should be able to ask basic questions to assure that the team is not taking unnecessary shortcuts related to model training or validation, selection of training data, etc. Unfortunately, it’s very common for data scientists to push hard to use cutting-edge techniques that are completely unnecessary, and the data science manager must be able to push back on this when necessary.

Regarding skills 5-7, it’s important to understand that much of the common tooling and methodologies (such as scrum) in use today were developed over the past 20 years by and for software developers, rather than data scientists. Data science teams must adopt and adapt the elements which will make them most effective in their unique tasks. Likewise, visual design and communication skills require special considerations when the subject matter is quantitative and especially when the audience is non-technical. I’ve seen very few data scientists communicate well to non-technical audiences without special training.

I sometimes add 2 additional items to this list of skills, as part of a soft-skills training for data scientists, these are

Working in a multi-cultural environment
Dealing with office politics

Why Are Analytics Translator Skills So Important?

Analytics translator skills are crucial for two roles in particular: data science product owner and data science team manager. Without analytics translator skills, neither of these two roles will be able to bring to bear the full potential of data science within a business context. Analytics translator skills are also extremely valuable for data scientists (individuals specialized in advanced analytics), both to steer them in producing business value and in helping them work more effectively with business counterparts.

How Do We Fill Analytics Translator Roles?

I’d agree with the Mckinsey article that the best solution is to train existing staff, rather than hire into an analytics translator role. Existing staff will already have a deep, proprietary knowledge of your business and have already formed relationships with key stakeholders. To illustrate, consider an AI healthcare company here in Amsterdam, who employees licensed medical doctors as product owners for their data science teams. In such cases it’s especially clear that adding analytics translator skills to the hiring conditions is not reasonable, and that providing trainings is the appropriate solution.

Also, analytics translator skills are sufficiently general that most staff with a healthy dose of curiosity can learn them fairly quickly, and my experience is that a large number of experienced professionals are indeed eager to learn, given the opportunity.

Many companies are setting up internal analytics translator trainings. For those without such trainings, part of the skill set can be learned through online courses, but there are relatively few open enrollment programs that teach the combination of business, communication and stakeholder management skills required. I do teach some of these skills at the University of Amsterdam’s Business school, and I often give in-house trainings for analytics translator skills. Feel free to reach out to me if you’d like more information and advice on how to set up internal trainings, or if you’d like information on my open-enrollment courses. You can reach me at trainings@dsianalytics.com.

You Don’t Have to Be a Data Scientist to Fill This Must-Have Analytics Role. Nicolaus Henke , Jordan Levine and Paul McInerney. Harvard Business Review. Feb 2018
The Age of Analytics. Competing in a Data Driven World, Mckisey & Company. 2016
AI Adoption in the Enterprise, How Companies Are Planning and Prioritizing AI Projects in Practice. Oreilly 2019
The State of Machine Learning Adoption in the Enterprise, Oreilly 2018

Sample Heading 3

It was Friday the 13^thand I was almost out the door to deliver Day 4 of an Analytics Translator training at the University of Amsterdam. I somehow hadn’t connected the dots that the university closings announced the previous evening meant that my training was cancelled until, by force of habit, I glanced at my emails and saw that the full-day training I was scheduled to teach was also, most definitely, cancelled.

The following days were a back-and-forth discussion with the participants (one of whom had since returned to Hong Kong) about how to reschedule. Most preferred trying a video training and finishing their certifications as soon as possible. This would mean giving the training virtually. Given the time zone challenges, we decided to split the full-day training into two half-days. As organizers, we had a secondary motivation that two half-days would give us two chances to master the technical challenges. This turned out to be a wise choice.

And so began preparations for what would be my first virtual training (one of many firsts experienced during the Dutch corona virus lock-down).

By now we’ve all been subjected to remote meetings using the myriads of available conferencing tools. I’ve developed a love-hate relationship with many of them (UvA happens to use Zoom). If you’ve ever struggled to simply join an existing meeting, you can imagine how much more pain is involved in mastering the intricacies required to schedule and run a four-hour virtual training.

I won’t go into the details of how to use Zoom (there are manuals and videos for that) or camera placement (Dave Pond of linkedin recently poste a short video on that), but I would like to share some of the learnings I had in the context of giving my first virtual training.

Preparation Stage

I must have spent as much time preparing the technology for the first video training as I did actually giving that training.

Physical setup

I wanted to add two external monitors, but my new windows surface laptop had only a single USB-C port, and the Dell USB-C hub I’d ordered turned out to not support two external monitors. I arranged the two with the camera centered. I’ve added a photo to also illustrated the arrangement of the software windows (described below).

I brought in my high-end blue yeti microphone and my Logitech spotlight presenter, a very cool device to highlight or magnify parts of the screen as you cast. You can also see the disappointing Dell USB hub next to it.

Software setup

Big question was the video background. I was working from a spare bedroom; a passable backdrop, but not ideal. Many conferencing tools (Zoom included) allow you to substitute an artificial background. I’ve seen people use beach or mountain backdrops, but I wanted something realistic and professional.

A few minutes searching online brought me to an office background that seemed realistic. At this point I realized that the algorithm that leaves your head but replaces your background is a bit less satisfying for lighter backgrounds, so I had to fiddle with image darkening (I tried pasting photos in powerpoint and then adding a semi-transparent grey background). After spending much too much time on this, I’d found a fake background that worked for me (No, I’m not showing it here, sorry)

Much more important was how I would set up the powerpoint presentation. If I used normal presenter mode, I would have a full screen taken for the presentation, and the other taken for the speaker notes, leaving me no space for the zoom screen (questions, breakout rooms, etc).

What I discovered was that if I started the presentation in presenter mode (alt-shift-F5 on windows) and then moved the speaker notes window to the same screen as the presentation (windows-shift-left arrow), this put the presentation on my external monitor and the speaker notes on my primary (laptop) monitor.

When I shared my screen in Zoom, I would select ‘advanced’ and then share just the portion of the speaker’s page, now on the external monitor, that had the presentation screen. (Yellow boxed area on photo)

Playing around a bit more, I realized Powerpoint has a transcription feature which would allow me to also broadcast subtitle’s as I spoke. This was especially useful as English was a second or third language for the participants. The transcription is processed on my laptop, so subtitles are fine even if network quality is poor.

Practice Run

We did a practice run a few days ahead of the training, along with my (also-locked-down) university colleagues. This let me get honest feedback on my fake background, along with audio and video quality. Oddly, audio quality was best without my fancy mic or headphones (I later realized the problem was again the Dell USB Hub I was connecting through).

Training Day

The day started out smoothly, except that one participant missed the training entirely because he was using the web version of Zoom, which happened to be experiencing global outage at that time. The others used the app and joined with no problems.

There were technical glitches that we missed during preparation and discovered in real-time (e.g. related to creating in-session polls, managing breakout rooms, sharing audio, etc)

I did my best to ensure a good network connection from my laptop, running an ethernet cable down two flights of steps and directly into my router. But despite my best efforts, I did lose network connection several times during the training (latter solved by plugging directly into the cable modem, which ironically was lower-end but more stable).

Zoom has a nice breakout room feature (I’m sure other tools have something similar). The host can choose the number of rooms, with participants randomly assigned. Participants can ask questions from the rooms, and the host can temporarily join each room, broadcast messages to the rooms, re-shuffle participants, etc). Great feature, except it was just at the point when I had started these groups that I lost my connection and was dropped from the training I was giving. When I re-entered two minutes later, I no longer had access to the rooms my participants were in, so I’d lost contact with my participants

Wracking my brain how to retrieve my lost participants when suddenly a man in the main conference room started talking to me, explaining how he had just checked on each breakout room and had agreed with them all that they would wrap up the discussions and return to the main room in 5 minutes. Completely surprised, I assumed this was some university zoom administrator who had taken over. In fact, zoom had automatically assigned one of the training participants as the new host in my place. This person had decided to take over responsibility for supervising the group!

The result was that the breakout sessions went a bit longer than I had planned, so we skipped one of the planned breaks. Not ideal.

Lessons learned

Virtual training presented some notable advantages. For me, it was easier to reference my notes when needed (I turned my camera off while presenting slides). Participants, for their part, could join with less burden of travel, arranging child care, etc. (I know several would check on children during the breaks). It was also much easier for participants to share their screens with the group after breakout sessions, something I always struggled with during in-person trainings.

On the flip side, I missed being able to ask spontaneous questions with a simple ‘show of hands’. I also couldn’t read the faces to see if someone looked puzzled. With everyone on mute by default, there was no spontaneous laughter at my jokes (a few of which were funny). Of course there also were unscheduled distractions on participants, typically from housemates, pets, or children, and I’m fairly certain the daughter of a participant joined the training from her phone at one point (on mute, of course)

The only serious concern I had was that some participants did not always join the breakout rooms they were assigned to during group exercises. This would typically not happen during in-person trainings. In the flip side, I imagine these were the people who felt it benefited them most to work on tasks alone, and so they were able to make that choice themselves.

In Conclusion

At this point, I’ve taught two virtual trainings and am scheduled to teach two more later this month. Although we were initially unsure how they would compare to in-person trainings, I have to say that I’m quite happy with how they went. We also feel these worked well with the students, who were able to freely ask questions and steer the discussions towards areas most relevant to their work.

With the right preparation and with a conscious effort to invite questions and solicit audience participation, participants still benefit from the live, interactive training that differentiates these dynamic trainings from the static content available in books or pre-recorded online courses.

I don’t know what the future mixture of on-site vs virtual trainings will be. The private trainings I have scheduled with companies this spring will be virtual, but I expect the analytics translator trainings I have scheduled at UvA in September will be in the classroom.

If you’d like to stay informed about my training schedule as it materializes, you can subscribe for notifications at https://dsianalytics.com/training-news.

As you evaluate your strategy, perhaps considering a strategic pivot, gather and utilize as much data as possible to build a deep understanding in four areas:

Your customers,
Your competitors,
The external factors that impact you, and
Your own product.

You may be surprised at what the data will show you.

Your customers
Customer data is one of your most important assets, and more of it is available today than ever before. We have moved from a 2D to a 3D customer view, including motivations, preferences, and habits. The more data you collect, the more complete the picture.

Getting the data
Identify all customer interaction points, including

Hit-level data from digital platforms: web sites, apps and kiosks
Interaction with customer support: phone, email, and online chat
Social media, including direct messaging, tweets, and posts on accounts you own or they own
Location, including store videos and movement logs. There are several technologies for monitoring movement, including embedded sensors, Wi-Fi, Bluetooth, beacon, and even light frequencies in combination with smart phone apps.
Data from sensors, RFID tags, personal fitness trackers, bio-medical readings, accelerometer data, external temperature, etc.

For each interaction point, make an inventory of

Available data,
Potential uses of that data, and
Privacy and governance considerations

Link the data to get a 360 view. If a customer phones your call center after looking online on your cancellation policy webpage, connect those two events.

Use the ‘big data’ in particular for an understanding of the intents, preferences and habits of your customers. Use this customer behavior when constructing personas.

Your competitors
It’s especially challenging to get good information about your competitors. Nielsen, Comscore, and SimilarWeb will sell estimations of traffic to competitors’ sites and apps, possibly including referrer information. Google Trends gives branded search.

Scraping will show competitor inventory, services and physical locations. If you are competing on price, adjust pricing based on your competition’s proximity to your customers. For physical locations, use address and transportation routes. For online sales, consider the referrers. Customers arriving from price comparison sites should be considered price-conscious and in ‘driving range’ of your competitors.

Increase your share of wallet, the percentage of a customer’s spend that goes to your business rather than competitors’. Use the detailed customer data you have collected and see what categories and what specific products are typically purchased by the same customer segments. You’ll be able to see which customers are already making those cross-purchases, which are browsing but not purchasing, and which are active in only one category.

By using the customer segments you’ve created, you’ll be able to see if your competition is appealing more to quality-conscious customers, marketing-reactive customers, high-spenders, etc.

External factors
You may be impacted by external factors ranging from government regulation to local weather. If you are in the travel and tourism industry, regional holidays will impact long-term bookings and weather will influence short-term bookings. Commodities prices will influence production and transport costs, and exchange rates or political turmoil will influence cross-border activity.

Much of the impact of external factors will be from traditional (small) data, but newer (big) data sources will provide additional, valuable signals. Monitor changes in online activity, particularly around product segments. This will flag unexpected changes requiring your attention. As an example, Google Maps and Waze can detect construction or road closures simply by studying driver movements, without the need for dedicated data feeds from departments of transportation.

Construct regular forecasts of key metrics, including number of visits and sales projections, segmented down to the levels at which you can steer your operations (such as product and region). Monitor these at a daily or weekly level and take note whenever they move above or below expected levels, as this may signal an unexpected external factor impacting your business.

Your own product
You may not understand your own service or product as well as you think, and your customers may perceive them in completely different ways than you’d expect. What is working and what is not working? How are customers responding to your products? Where are you losing money through inefficiencies?

If your web offering is a significant part of your business, create and track micro-conversion events to see how your items are performing even before a purchase is made. Tag your sessions with A/B versions so you can deep-dive into results using your big data tools.

As you start collecting and studying these four types of data, you’ll uncover new insights and you’ll be led to ask new questions your data may not (yet) answer. Getting answers to these additional questions and regularly revisiting your initial questions will help you excel in constantly evolving business environments.

For a more in-depth discussion of digital strategy, see my review of Sunil Gupta’s book, Driving Digital Strategy

You’ve decided to make better use of your data, but what are the roles you need to hire? It’s not just the ‘data scientists’. If you want to make more data-driven business decisions and use analysis and machine learning to change how you operate, you’ll need to hire six types of data people.

Data Engineers

There is a lot of important data waiting to be collected and used, but getting it moved efficiently and stored in ways that allow effective usage are no easy jobs. In fact, preparing data for analysis is more time consuming than doing the analysis, and finding qualified staff for these specialize tasks is getting quite difficult. Even if you purchase expensive specialized software, you’ll still lose time and performance if you don’t have specially trained people who

Have expertise in using multi-purpose ETL tools as well as data manipulation tools for big data (e.g. Pig, Storm, etc.)
Have expertise in designing data warehouse tables. Depending on your tooling, this may include OLAP cubes, data marts, etc. If the database tables are poorly designed, your reports and data queries may become unusable due to instability and lag.

If you don’t get specialist data engineers, others in your data team will waste time covering this critical but specialized task, with at best mediocre results. I’ve seen it before, and it’s not pretty.

Algorithm Specialists

Your most innovative projects will be done by experts using mathematics, statistics and artificial intelligence to work magic with your data. They are writing the models that beat the world champion at Go, or recommend your next favorite movie on Netflix, or understand that now is the right time to offer the customer a 10% discount on a kitchen toaster. They are forecasting your Q2 revenue figures and predicting confidence intervals for the number of customers you’ll see next weekend.

Look for people with expertise in statistics, mathematical optimization, prototyping tools (KNIME, RapidMiner, H20.ai, SAS EnterpriseMiner, Azure ML, Google Cloud ML, etc), and strong coding skills. They should have a strong background in mathematics, usually a degree in math, statistics, computer science, engineering or physics, and they should have experience writing and coding algorithms in a rigorous language such as Java, Scala, R, python, or C/C++. They should preferably be experienced in object-oriented programming. They should have something on their C.V. that demonstrates they are really smart.

Business Analysts

Most of the ‘data scientists’ you hire will be what I would call ‘business analysts’. They answer basic but important data questions asked by your business units. They are really proficient at Microsoft Excel.

Where should they sit within the organization? Some companies group them centrally, while some embed them within the business units. Each approach has advantages and disadvantages, but the decentralized model probably occurs more often in small to mid-sized enterprises.

Web Analyst(s)

Customer online behavior is a very important data source. You can choose from are a broad selection of mature web analytics products, but whichever tool(s) you choose should be managed by a trained specialist who keeps current on developments in web analytics and related technologies (including browser and mobile OS updates). The web analyst will oversee tagging and ensure effective collection of customer activity on the website and mobile applications. Some web analytics tools can collect data from any connected digital device, not only browsers and apps, and the web analyst can assist with this data consolidation. The web analyst will set up conversion funnels and implement custom tagging and will monitor and address any implementation problems that may arise, such as data errors related to new browser releases. They will also assist in merging company data with web analytics data, which may be done on the organizations databases or on the web analytics server.

The web analyst will also be an expert in extracting data, creating segments, and constructing reports using available APIs and interfaces. For this reason, they may be actively involved with A/B testing, data warehousing, marketing analysis, customer segmentation, etc.

Reporting Specialists

You’ll benefit greatly if you hire or train staff skilled at creating good graphs and tables. This requires a mixture of art and science and should be done by people who excel in

Selecting the table or graph most suited to the use case. For example, trends will jump out from graphs much more quickly than from tables, but tables are better for more sequential tasks.

Selecting the layout and format most appropriate to the data. For example, reports with time series data shown vertically are not intuitive.

Reducing visual clutter, freeing the recipient to focus on the most important data. This is rarely done well.

Leveraging principles of gestalt and pre-attentive processing

Selecting shapes and colors that minimize confusion

On a technical level, the reporting specialists should be comfortable writing database queries to extract data from source systems and they should be trained on your BI tool(s).

Leadership

Data opportunities span the breadth of your business, but you won’t get very far with your data initiative unless you have a senior analytics leader who can set the vision, win support, prioritize the roadmap, manage stakeholders, and build and lead an outstanding data team. I’ve written more on this in my article Recruiting a Chief Data Scientist.

Pulling your data team together is a challenging task, particularly in today’s labor market, but the competitive benefits you’ll eventually reap will make you wonder how you ever managed without such a team.

Big data technologies are playing an increasing role in cancer research. Two ways that I’ll touch on are the storage of an exponentially growing quantity of sequenced genomic data and analysis of pathway disruption by large-scale graph algorithms.

There is an amazing diversity in cancer cells. A single tumor may contain 100 billion cells, each with differing somatic mutations. One patient exhibits so many types of mutations that a tumor sample will not give a comprehensive picture of the mutations present in that individual, hence the push to sequence as many genomes as possible.

Since the completion of the Human Genome Project in 2003, we have increased the efficiency of our gene sequencing to over 10k giga base pairs per week, resulting in a huge increase in the amount of archived genomic data. Parallel technologies now allow us to sequence much more quickly and cheaply, so that the price of sequencing a genome has dropped from $100 million in 2003 to $1000 in 2016. Unlike much other scientific big data, such as from the Hadron Collider, researchers will typically store all genomic data for future reference.

This drop in cost, together with the highly varied nature of cancer cells, is rapidly propelling the amount of data supporting cancer research into the realm of big data. There are already more than 1000 genomic sequencing centers in the world, and projections are that, within a few years, the amount of storage consumed by sequenced human DNA will be growing at an annual rate of 2-40 Exabytes.

Big Data analytics of pathway disruptions

Cancer is characterized by the disruptions of signaling pathways (interacting proteins) within a cell. The disruption of a pathway may be the signal of the cancer, but any given pathway may be affected in different patients by different somatic mutations, as the network could be disrupted in different ways. The result is a vast quantity of data from which we can attempt to solve a very complex and very important problem.

How is this data analyzed?

When researching a cohort of patients, each with a set of somatic mutations, we don’t initially know which of these mutations are driving the cancer (“driver mutations”). We will sequence the genes, record which are mutated, and look for unusually commonly shared mutations that are likely to be drivers for the cancer. In the past, researchers analyzed individual genes to see which mutated more commonly (there would never be a gene that mutated across all patients), an analysis which would require advanced statistical corrections across 25k genes.

Some researchers are now focusing instead on entire protein pathways. Here again, we face the challenge that pathway mutations differ between patients. For example, a 2008 study by TCGA published in Nature found a specific pathway mutated in 87% of patients in a sample of Glioblastoma patients: significant, but not consistent.

This is where big data analytics comes in. We start by constructing a gene interaction network: a graph with 25k vertices and 625m edges. Pathways are represented by subnetworks in this graph. We use an iterative big data graph algorithm library such as Gelly (part of Apache Flink) to find connected subnetworks that are mutated in a significant number of patients.

This method has produced interesting results. When applied to data from a cohort of 316 patients, researchers found 27 subnetworks (of at least 7 genes) implicated in ovarian cancer, many of which interact with each other. Thus, a more complex picture started to emerge. Twelve of these 27 were known pathways that had been studied, and the other 15 were new to researchers.

This analysis was also applied to a study of 199 patients with Acute Myeloid Leukemia. It uncovered five subnetworks of at least five genes, each of which were significantly mutated. When applied to a sample of 514 patients with breast cancer, it found thirteen subnetworks of at least eight genes each. (Publications in Nature in 2008 and 2011)

A Combined Effort to Treat Cancer

The ultimate goal is personalized cancer treatment based on a deep understanding of the mutations of each patient’s cancer genome. This brings us to the data challenge of consolidating genomic archives, research literature, trial results, and individual health records, aggregating across disparate structured and unstructured data sets.

A growing community is forming around the effort to assemble a shared genomic database for use in combating cancer. In 2016, the National Cancer Institute launched the Genomic Data Commons (GDC), referring to it as ‘a first-of-its-kind public data platform for storing, analyzing, and sharing genomic and associated clinical data on cancer.’ Since its formation, several research centers have shared their cancer genomic dataset with the GDC. Ref, Ref

As we see the convergence of computing and medical technologies in these and other cancer research methodologies, we look forward to increased benefits from big data analytics in the fight against cancer.

On a side-note, a key aspect in healthcare AI is having in place a product owner capable of steering these data science projects effectively.

Back when ‘retail’ meant 100 meters of shelving and one check-out counter, face-to-face retailers grew to understand their customers’ intentions, preferences, and idiosyncrasies, even when customers walked out without making a purchase. Retailers would go on to use these insights to help subsequent customer in an insightful and personal manner.

In our age of e-commerce, with millions of digital customers moving within our ‘shops’, recent technology developments have brought us back to the point where we can just about recapture that level of personal customer insight and interaction. This is critical to understand, as shopkeepers who only interact with customers while ringing up purchases are as destined for failure now as they were in the old days.

Online Customer Journey

If I’m running an online store, I can leverage my digital data in one of four ways:

Worst: I simply record the number of sales made (like the shopkeeper who sits behind his register all day). This is really, really negligent. I deserve to go bankrupt.

Slightly Better: I use basic web analytics. This is minimal effort, but at least now I can see basic stats: how many digital visitors, where they came from, time on my site, etc.

Standard Practice: I’ve got an analyst who set up conversion funnels, UTM codes, e-commerce tracking, integration with A/B testing, etc. I get some high-level insights but am very limited in what detailed insights I can get and in how I can help customers in a more personalized way.

Big Data Method: The way to regain the personalized customer experience of old days is to store and analyze the entire customer journey, for every customer. Now I have the ability to once again understand the intents and preferences of my customers: what they originally thought they wanted, how they modified their product search, when they reconsidered products, when they read reviews or sorted by price or size or rating or whatever. I can see what product features were selected and then unselected, and which of those selections correlated most heavily with purchases. I can then start to help my online customers in ways I never could before.

It’s the Big Data method that allows me to truly understand and eventually influence customer journey.

The Realm of the possible

These days there are also relatively painless ways to start using Big Data to recapturing personalized customer interaction (See my recent article Useful Big Data: The Fastest Way for more practical details).

With relatively little effort, we can

1. Get answers to extremely detailed questions about what customers did while browsing our sites, accounting for the customer journey and general personas of every single digital customer

2. Deep dive into multivariate tests to understand in highly nuanced ways how potential product changes affect conversion rates

How does this look? Let’s walk through the process that a traditional shopkeeper would follow, and realize that we can again do this with our millions of digital customers.

Customer Intent

Right from the start, the successful shopkeeper finds out all he can about the person visiting his store, looking to understand things such as

1. How did the customer find out about the store? A friends’ recommendation, marketing efforts, internet search?

2. What is top of mind for the customer looking? A particular product, the solution to a problem, a gift, a bargain?

By getting answers to these questions for digital customers, we can optimize our marketing efforts, our landing pages, and our product mix.

Customer Preferences and Priorities

As the customer continues the shopping process, the shopkeeper will begin to understand the preferences and priorities of the customer.

· Desired price range and features.

· Persistence of search criterion, including price elasticity

· Prioritization of search criterion, as indicated by sort order and item views

· Engagement with product reviews

The customer may express these preferences outright, or the observant shopkeeper may imply them by watching the customer’s actions…which products are ignored, quickly put back, or re-considered multiple times. Getting that level of behavioral insight for millions of digital customers is amazing, and it is possible.

Proactive Shopkeeping

The shopkeeper will help the customer. He will recommend products most likely to sell (or perhaps with the highest expected profit margin!)

The shopkeeper will keep two things in mind when making these recommendations:

1. What have I already learned about the customer’s goals, intentions and preferences?

2. What has worked with other customers? If I know what similar customers purchased in the past, I can recommend those products, thus increasing the chance that I make a sale and that the customer walks away with something that makes them happy.

The shopkeeper will be very clever in how he nudges the customer into making a transaction. For example, he’ll know when to offer spot sales. He’ll learn when to make certain inventory prominent and which features of each product to highlight for each customer.

It’s not enough to say ‘most customers bought this product’. It needs to be ‘most customers just like you bought this product’. It can’t be ‘sales increase when I run a 10% discount’. It needs to be ‘sales increase on this product to this customer base when I offer a 10% discount at a particular point in the shopping journey’.

Observant

How does the shopkeeper know how to personally interact with each customer in a way that produces the most sales… this process that we call ‘conversion rate optimization’? It’s by remembering what has worked with customers in the past. The shopkeeper observes and remembers the entire shopping experiences of all past customers. He can then make observations such as

Most people who came looking for phone X and then moved on to consider phone Y and Z ended up purchasing phone Y; if they also sorted based on price, then conversion rate increased by 7% when a 5% discount was offered on phone Y.

Wow. Now I know, for a particular segment of customers, how to both make more money and to have a happier customer.

But not intrusive

Two important things to watch here:

1. We don’t store or use personal data without specific consent. No problem. I don’t need to remember your name or anything that identifies you personally. I just need to remember your general persona and behavioral signature.

2. We don’t creep people out with what we know. Don’t do what Target corporation did and send a maternity advertisement to the pregnant girl’s home just because you figured it out based on the vitamins and skin lotion that she bought over the past few months.

And always learning from the customer

Eventually the customer makes the purchase, or perhaps doesn’t make the purchase, but either way the shopkeeper knows much, much more than he would if he’d just focused on the point of sale. Not only has he increased customer satisfaction and increased the chance of sales, he has developed his understanding of what works or doesn’t work. He’s learned what is important to customers and can direct his purchases, marketing and product placement accordingly.

Perhaps the customer was very well prepared, and the shopkeeper has even learned something new about his own products, such as cross-sell or substitutionary potential. He can later use this new information to make better recommendations to future customers.

To sum it up

Using Big Data technologies, today’s shopkeeper can again understand the preferences, intentions and habits of customers at extremely high level of detail, allowing him to understand and help them in much more insightful ways, thus both increasing his various conversion rates and also providing the customer with a much better shopping experience.

All in all, it’s a lot of potential insight, and it can make a massive difference in revenue.

This article was originally a guest posting for Adrian Swinscoe’s Customer Engagement blog.