I often get invited to speak at finance conferences about how Consumer Tech Organizations differ from traditional Consumer Organizations, and how the role of Finance is evolving in this context. This blog is Part 1 of my views. In Part 2, I will cover the specific soft skills and technical skills needed for finance partners to deliver in such “Digital” Organizations.
“It is not the strongest of the species that survive, nor the ones that are the most intelligent. The species that survive is the one that best adapts to change.”
This quote from Charles Darwin best explains the declining longevity of Organizations that fail to adapt themselves to rapid and accelerating changes around us. These changes refer to the four dimensions that matter most to businesses – Consumer, Technology, Competition and Regulations. Consumers are spoilt for choice will continue to aspire for products that offer better quality, value, convenience and service. Products, services and the underlying business models that serve them are increasingly turning global. Think Uber and Ola, Tik-Tok and Facebook, Airbnb and Oyo. Computing power is now cheap, commercially available and ubiquitous. A startup, a kirana shop and a small enterprise can access the same computing power as an IBM or Google, with a mere click of a signup thanks to the consumerization of Cloud computing.
“Digital” Organizations recognize these sweeping changes and have evolved to serve customers better than their peer “Traditional-Analog” Organizations. “Digital” Organizations are more likely to be found in the Consumer Tech vertical given their endemic DNA, however many traditional “brick-and-mortar” Organizations are also rapidly evolving into Digital Organizations. Likewise, the Consumer Tech industry also has its share of Organizations that are stuck with the DNA of a Traditional-Analog and are less likely to survive.
What do Digital Organizations look like? How do we know if a traditional company is evolving into a digital skin and altering its DNA? What does such an evolutionary journey look like?
The 5-Level scale through which organizations evolve on the scale of data maturity is best represented in the above visual. Digital Organizations operate at the right-end of the spectrum, while Organizations that have the DNA of a traditional-analog company tend to find themselves at the left side of the spectrum.
The core of the difference is around how Digital organizations capture, process and consume data and how they link this to the way they serve their customers. Let’s understand these levels as they progress from left-to-right.
Traditional-analog organizations have a smaller data universe and harvesting of data is restricted to core financial and transaction systems (think SAP and other ERPs). They have limited avenues for data capture beyond core financials. Finance tends to act as a gatekeeper to this data (often like the priest controlling devotee access to the temple deity), ostensibly because traditionally finance was entrusted to ensure integrity of data. Data processing is largely in batch mode, and data custodians (think “MIS officers”) often report to finance. Other functions need to go “through” finance to access data.
Digital organizations, in contrast, cast a wider net to capture data. Captured data pertains to customer interactions, thereby expanding the dataset universe beyond financial reporting and transaction systems. Their data architecture is such that the captured data is stored in a “data lake”, and from which any functional team can extract data autonomously (subject to access rights, of course). Each functional team is supported by its own pool of analysts, unlike in the traditional Organizations where analysts are found only in finance teams. Finance is rarely a gate-keeper to this data “lake”.
In Level 1, the information architecture in digital Organizations delivers greater democratization of data within the organization.
In an epitome of traditional-analog Level 1 low-maturity, in one of my earlier finance roles on a business unit, an MIS officer reporting to me would dutifully email the rest of the organization every Monday morning a sales report for the previous week – something that other teams would eagerly consume!
Level 2 defines the degree of sophistication with which data security and access protocols are embedded into the organization’s hardware and software. Data is meticulously classified into customer and personal data (and hence most sensitive), sensitive financial data (especially for listed Organizations), confidential data and non-confidential data and “read-only” access restricted to roles (not hierarchies) purely on a need-to-know (and not seniority, for instance); more importantly, such access protocols are hard-wired into the organization. Sensitive and non-anonymized customer data is hosted behind firewalls, and employee badge-access physically restricted to zones where sensitive data is processed or exposed.
Picture this hypothetical nightmare scenario at an organization that has low Level 2 maturity. A disgruntled and IT-savvy HR employee plugs in into the server of their consumer products company, downloads super sensitive customer and distributor contact details and credit history into a pen drive and walks off the premises to join a competitor. Protocols did not exist to prohibit access to data not normally required for pursuit of regular HR role; nor was physical access restricted to potential access points for such sensitive data for non-essential personnel!
Digital Organizations have Level 1 and Level 2 maturity as a matter of hygiene; traditional-analog Organizations differ because they lag on the data and access standards that define such digital maturity.
The next three Levels define the organization’s ability to harvest data to deliver superior business performance.
“Information is the Oil of the 21st century, and analytics is its combustion engine”.
Most are familiar with the first half of this quote, but few understand the crucial importance of the second half of the quote. Once organizations reach maturity on Level 1 and Level 2, they risk being drowned in data. Raw data, by itself, is like a commodity; even worse, once plumbing is built, it can be over powering and its importance be frequently overestimated. Data can also go stale and lose its relevance, and it can be expensive to store, catalogue or process unless you have the right combustion engine.
Maturity on Level 3, 4 and 5 defines who has a better combustion engine to convert raw data into energy that can power the organization to deliver superior and competitive business performance.
In Level 3, a company extracts insights from its data, and does this better than its competitors.
In Level 4, a company converts these insights into actions, and does this faster and better than its competitors.
In Level 5, possibly the holy grail of digital prowess, the company makes predictions using data, and determines or recommends the best actions to deal with the predictions.
In my view, there are four controllable dimensions that enable a company to raise its maturity in each of Levels 3, 4 and 5, thereby transforming them from traditional-analog Organizations to Digital Organizations.
First, Digital Organizations have superior tools and capability to mine data for actionable insights and to make predictions.
All Organizations worth their salt do a certain degree of analysis in their quest for insights; however, the practical constraints imposed by limited data sets, clunky excel or excel-like tools that are not designed for large data set handling and the necessity to frame a hypothesis that needs to be validated imply that conventional approaches to analysis yield only finite insights with low productivity.
On the other hand, digital Organizations that have invested in Machine Learning and tools that build on ML have dramatically increased their capacity to unearth the volume and quality of insights from data sets. They also close loop such insights back to business decisions much faster than traditional-analog Organizations
The above picture is a vivid illustration of the difference between Machine Learning (ML) and Traditional Computing. In ML, the input that is ingested is both a historical data input and the historical outcome (output) for that input. The output of ML is a program, and this program is then used as input in future to generate the desired output from new data sets.
As illustration, let’s look see how ML can work in Foodtech company. The Foodtech company surfaces restaurant listings in a certain ranking sequence to a browsing customer looking for “Chinese Food” to order. The goal of the foodtech company is to deliver a favorable customer outcome viz a successfully placed customer order and minimize instances where customers browse restaurant listings and then exit without placing an order presumably because they didn’t find something they liked. The ML algorithm mines previous data sets viz. restaurant listings surfaced as input, and corresponding output (whether the customer successfully placed an order after viewing that listing, or whether the customer exited the site without placing an order – null action). Basis such “training data” of past datasets, the algorithm then recommends an improved and revised restaurant listing ranking sequence (i.e. the “program”) that uses the training data to deliver a higher likelihood of a desired output (viz. successful order placed).
Such ML tools and the capability that goes with it is valuable for digital Organizations and places them on a higher maturity level on their ability to mine customer insights, make predictions and recommend actions vs traditional-analog Organizations.
Second, Digital Organizations use data to weigh how different actions that impact their customer stack up when measured in a common currency in terms of net present value (NPV) of long term cash flows for themselves, thereby helping them make better choices. Such tight linkage between customer-facing actions and their own NPV of cash flows of such actions ensure that the customer sits right at the center of their business model.
Let’s take two examples.
The first example is that of a cab hailing company (hypothetical, ala Uber or Ola). Let’s say that on a given day, the cab hailing company has potential customers who have hailed cabs, but doesn’t have enough supply of cabs to serve those orders. The cab hailing company needs to determine which customers will be served with matched cabs, and which customers won’t be. So, what logic can its cab allocation algorithm use?
It could simply allocate available cabs to customers following a real time First-In-First-Out logic (“FIFO fairness”). Or maximize its revenue in a batched time interval (say, 90-seconds) and deny serving rides that have lower revenue per ride (“maximize current period revenue”). Or it could raise its fare via surge pricing till it determines a level where it maximizes its profits and ensures that supply of cabs equals demand for cab rides, knowing that at higher fares, some potential customers will drop off; and repeat this logic every time it finds demand exceeding supply (“maximize current period profit”). Or it could prioritize certain customers who are more likely to repeat future rides if they are served now, irrespective of their current ride fare (“maximize positive customer actions”) Or it could prioritize customers who are more likely to switch off the ride hailing platform in future if they are denied a cab at this moment (“minimize negative customer actions”). The algorithm that ultimately makes this complex decision in a jiffy blends at the back end a formidable computational engine that integrates insights (“Customer X is a loyal and regular vs Customer Y who is fickle and occasional”), prediction (“Based on past data, X out of Y ride hails are expected to drop off if the fare is raised by Z”) and action (“Raise fare by X, allocate cab Y to customer Z1 and deny customer Z2”).
One doesn’t know how cab hailing Organizations actually operate cab-to-ride allocation in supply constrained situations, but it’s fascinating to see how analytics and predictive logic that mines data on past behavior in similar situations can potentially be fine-tuned to maximize net present value of future cash flows by choosing policies that lead to customer behavior that maximizes long term cash flows for the cab hailing company.
Let’s take another example, this time from an online retailer that serves beauty products. The online marketplace realizes that a certain low unit priced item (say, a daily use consumable item) can never be served profitably given the structural low unit selling price and high relative cost of shipping that item to the customer. Should the online retailer then delist such low unit selection, and avoid potential losses?
The right decision would be to determine the differential net present value of long term cash flows predicted based on past customer behavior observed for a cohort of customers who bought that low-unit price product vs net present value of future long term cash flows for an identical cohort of customers who historically didn’t buy that product. Next, assess if the present value of predicted future differential profits outweighs the definite loss in serving that low-unit price product today. This decision framework weighs that customers might potentially be delighted to find a low unit price and frequently consumed item available at their favorite online shopping site, and that this resulting delight might drive measurable and predictable enhanced long term stickiness to the marketplace and hence incent incremental future shopping for other non-low unit selection such that the future profit on that incremental shopping makes up for the present loss of serving that low-unit priced item.
Much of this is common sense and such gut-sense decisions to stock loss leaders are anyway followed by offline retailers who bank on making up for such loss-making selection through additional shopping basket from the incremental walk-ins; however, a digital organization can bring in a greater degree of science to such decision making by building superior analytic engine that can simultaneously combine insights, make predictions and recommend customer-facing actions that maximize net present value of future cash flows.