Is data really that oily?
Identifying what is driving value in data business models with the help of venture capital investors.
Data here, data there. Recently, the topic gained a lot of attention from different angles and from all kind of industries. The expression ‘data is the new oil’ is already overused to such an extent that people only start rolling with their eyes and vent a loud sigh.
However, the wave of attention is not too surprising given the growing flood of technology, and hence more produced data, into every corner of consumers life and industrial processes, currently headed by the expansion of IoT and IIoT (big data — duh) and hereafter most likely further accelerated by the spreading of 5G.
But why are people so obsessed with data? What can you do with it? How do you ‘use’ it? To shed a light on the topic, we will drill into the topic to explore where the oil really is hidden. For this, venture capital investors have been interviewed to identify the most valuable components of data-driven analytics ventures.
The ground rules
Before we dig deeper into the value differences within the utilization of data, we need to set a marked-out playing field. First of all, we concentrate on the holistic usage of data in a business context, thus, we will analyze business models. Secondly, data is used nowadays in mostly all departments and areas to document, analyze, and improve current processes. We will focus, however, solely on business models where data is a key element and essential to the daily operations of the business. Therefore, only business models along the data value chain are taken into account commonly referred to as analytics startups — from data generation to preparation to actual knowledge extraction to the visualization and finally the distribution of such. Thirdly, and lastly, we want to analyze the front of innovation while simultaneously prevent distortions from other business activities. What’s better to focus on than innovative startups focusing on analytics?
How to achieve meaningful insights?
The basic structure for the analysis follows the three business model clusters along value creation, value delivery, and value capture. The 9-component model of Osterwalder builds upon those three pillars, which are used as the detailed framework (Osterwalder, Pigneur and Tucci, 2005).
Some clever researchers translated the general business model components to data specific equivalents (Hartman et al., 2014). This data model has been used to define the different activities within each component and to assess the value differences between each activity. And who can identify the potential value of startups better than those who do it on a daily basis? Therefore, several venture capital investors in Germany, from the largest funds to smaller and specialized ones, have been interviewed to share their perception on different analytics startup structures and their operating business model components. So what was their shared conception?
Analytics business models, data activities and value
Both the value delivery and capture pillar have their distinct characteristics, but for business models within the data field the key differentiators are found in the value creation pillar. We will therefore neglect most of the first two pillars and concentrate on the latter. Only so much: First, although the value proposition can be raw and unrefined data or related non-data services, 94% of all offerings contain information or knowledge (Hartman et al., 2014) Business models in the analytics field are therefore not able to circuit delivering knowledge. Second, one is quick to realize that consumers mostly don’t need analytics at all, while businesses can monetize knowledge and therefore seek as advanced analytics as possible. Hence, although not all focus on business segments, the larger value is typically found in offerings focusing on such. The value creation pillar can be separated into data as a resource and key data activities.
1. Data as a source: Exclusivity and meaning [Value creation — Key resources]
Data is the foundation for all subsequent activities and therefore obviously depicts a vital part for all analytics ventures and a crucial role for businesses as a whole. Value within datasets resides in the exclusivity and the intrinsic meaning.
1.1 Put a ring on it — Exclusivity
If the data used is publicly available, the defensibility of the business model is at stake. The single most important aspect is the exclusivity of data. Machine learning algorithms and data analytics applications need large amounts of data. Owning a proprietary dataset or even generating new data provides a strong advantage and a large moat towards competition.
At least some data should be owned by the firm or contain an exclusive access. This access yields an incredible competitive advantage.
Data, however, is not necessarily the issue, it is access to it. Young ventures usually do not own large data streams and generating new data is time-consuming and cost intensive. Commonly, data is sitting in large corporations. Legacy firms, on the other hand, are often lacking the innovation possibilities to utilize their own data.
Data isn’t a shortage. It is becoming increasingly not the issue, it’s access to the data, which is the issue.
To achieve and retain exclusivity to some degree but also being able to access data, is the driving motive of many start-ups to utilize the data of external firms with software products.
Exclusivity is vital for the defensibility and sustainability of the business model. But even access to a proprietary dataset does not necessarily translate into value. Obtained data must hold a meaning.
1.2 It’s the inside that counts — Intrinsic meaning
Often you stumble upon startups which aim to gather data which essentially hasn’t much value before the young organizations disappear into the valley of the dead. A good example are supermarkets. Let’s assume you put some digital counter on the entrance doors of your favorite supermarket of choice. You will be able to view how many customers enter over a period of time and the number of people within the store at any given time. However, most days will look similar with peaks in lunch hour and after work. Weekends will have a slightly different but most likely a repetitive customer distribution as well. The major flaw with the data is that it doesn’t address an existing problem, you can’t extract knowledge by connecting shoppers within the supermarket to buying behavior and that the data is repetitive.
Key elements for meaning are having a pattern to analyze but with different outcome possibilities. Knowledge of this difference in the outcome should have a large impact for the customer.
Compare the previous example with analyzing x-rays for cancer prevention. The results yield different outcomes for each patient with a tremendous implied impact.
One of the most important aspects is that an existing problem of the client can be solved [with the data]. Crucial is that the problem or its impact is large enough.
Often the internal meaning within some datasets is overvalued by the owning firm. The quality of data and algorithms are difficult to assess from an outsider view, especially for the early-stage phases of startups. One approach to approximate the issue are the ‘Data Canvas’ and the ‘Data-Need-Fit’, which has been translated from the Value Proposition Canvas to a data equivalent (Mathis and Köbler, 2016).
The Data Canvas aims to identify which data is accessible, segmented in internal or external sources as well as in a periodically-recurring or continuous flow. The Data-Need-Fit describes the value capabilities of the identified data sources along products and services, gain creators and pain relievers on the one side and the customer profile along tasks, pains, and gains on the other side. This process helps to identify the value potential based on the firm’s capabilities and to match it with customer needs.
Valuable meaning hence must cater two things: First, the data can solve an existing problem of any customer segment, second, the problem is large enough that the client is willing to pay to get it solved.
If there is no need-fit, a proprietary dataset is without use. A valuable dataset should, therefore, contain both, exclusivity and meaning.
2. Key Data Activities of the Data Value Chain [Value creation — Key activities]
Prevailing is the understanding that data can contain knowledge. Therefore, a strong emphasis on analytics business models is put on activities extracting knowledge from a certain dataset. The range of activities follows the data value chain.
2.1 From past-describing to future-predicting — Analysis as activity
Most valuable activities are often attributed towards the end of the data value chain, of which analytics is stated as the most performed activity. Analytics is split into descriptive, diagnostic, predictive and prescriptive analytics. It describes ‘What happened’, ‘Why did it happen’, ‘What will happen’ and ‘What should we do’. Value resides in utilizing data to analyze the past in order to predict the future and present actionable tasks to a client.
The vital point by now is not only displaying and visualizing data but presenting actionable items to the client.
However, although most promote their solutions as groundbreaking, widely performed by startups are typically only descriptive analytics. Those young ventures often tackle the challenges for large corporations arising from the big data characteristics of volume, velocity, variety, and veracity to restructure datasets with the aim of a live-visualization. The subsequent steps into future-predicting analytics are yet often neglected.
Future-predicting analytics have a higher value than simple past-describing and visualizing analytics, but with a limitation. Theoretically, prescriptive analytics are most valuable as they offer the highest optimization potential. Practically, predictive analytics with ultimate decisions taken by humans are sought for most. The reasoning is a lack of trust in algorithms to operate fully automated, but also regulations prohibiting certain developments. Therefore, for the short-term view, predictive solutions will be in the foreground of analytics solutions. This might change in the mid-term view with growing trust and a more accepted widespread use of algorithms.
2.2 Don’t wait for others but produce yourself — Data generation as activity
A difference in the most valuable activity arises when adding a longer time-horizon to the key activities. Although predictive analytics is perceived highly valuable, the utilized algorithms may become a commodity in the future, while the databank on which the algorithms perform will turn out to be the differentiator of a good to a great business.
Short-term, analysis systems are surely more interesting as you can sell your value faster to a potential client. In long-term, however, the firms active in data generation hold more potential.
Many focus only on analytics as the most important knowledge-discovering activity, but the identified criteria for valuable datasets may shift the weight perception of value along the data value chain. As demonstrated above, a proprietary and meaningful dataset presents value in itself. Therefore, in the long-term, data generation might surpass analytics as the most valuable activity.
In a nutshell
Not everything which is sticky and burns well is valuable — or something like that. But there are definitely oily treasures hidden in the ever-growing flood of data. Paying attention to the details within analytics business models supports identifying valuable paths into the future.
Good luck finding your personal oil well!
As I’m new to writing here, I’m happy to here your thoughts. Hit me up on LinkedIn or leave a note here.
