Make your Data a Source for Peak Growth
A Framework for Efficient AI and Data Investments
In 15 years, forecasts predict that more than half of our globe’s growth will be based on a AI and data. The successful will see growth factors higher than 2, but it requires strategic focus, strong governance and long term investment to get there. In this post I describe what it takes for an organization to actually liberate their dormant data and transform it to a capitalized and well managed high gain asset on their balance sheet.
The article The Secret of Capturing Value through Digital Transformation triggered several questions from readers on how to actually govern and measure return on data and data activation. The questions kicked off a journey into how to frame the answer in terms of the theory of financial management, while at the same time keeping the links to the processes, technology and science of of data. The result is what follows. Hope you enjoy the revisit of basic financial theory and its merger with some of the excitement that is happening in data management, analytics and AI right now.
Growth from Data
Data is the new oil — Clive Humby, 2006
The 2006 quote gives a strong association on how valuable data is becoming. For a many years I thought the quote was rather silly as data is an intangible, invisible asset. Data also has virtually infinite supply and it can be reused for several purposes comparing it in value to scarcely supplied and commoditized oil becomes wrong. There is no doubt in my mind whatsoever that modern civilization would collapse in a matter of months if oil stopped flowing. Oil is about as important to the developed world as agriculture. That can not be the case for data, right? Here is a recent critique of the quote with a similar opinion as I had before.
Fortunately, I decided to challenge my presumptions and I was astonished by the findings.
Consider this: In this 2015 study, 87% of the market value of S&P 500 is shown to come from intangible assets. That share had risen with one percentage point on average each year for the last 20 years. Sure, there is a lot of brand capital in the total number, but my bet is that a large share of the growth in recent years is attributable to data and it’s children AI and analytics.
Digging deeper I found data’s importance to business performance is likely to grow even further. In this Accenture forecast AI is projected to double GDP growth rates by 2035. AI will do this by (1) automating manual tasks, (2) enhancing the abilities of the existing workforce, and lastly by (3) being a driver for new innovations. The foundation for AI is data, so if this is right, Data capital will drive the economy to twice the future growth if it didn’t exits.
For the biggest developed countries, the forecast predicts that the total GDP increase compared to baseline will be between +35% (USA) and +11% (Spain) from data & AI in 2035. This PwC forecast makes a similar estimate globally, keeping in mind that the time scope is 5 years earlier 2030. As the average is a doubling of GDP in 2035, the top organisations will achieve growth multiples larger than two and achieve this much earlier. This is an important insight because organizations can create immense value already.
Don’t think that is possible ? Consider the success from AI and data from these players:
- 80 % of all videos watched on Netflix is a result of a personalised recommendation by the company (read more)
- Alphabets (Google) makes over 88% of their revenues comes from AI and Data through their advertising business (read more)
- At the moment Facebook earns $20 per year on its 98 data points on the user (read more)
If you consider the impact on the economy, oil is directly contributing about 2.5% of global GDP. As seen above, data dwarfs oil on GDP impact both currently and into the future. The world economy has been developing with oil as its lifeblood for over a hundred years. In the next 100 years, data will be its new upgraded 2.0 lifeblood.
My view today is that the quote “Data is the new oil” is a broad understatement.
Machinery which is not used is not capital — Karl Marx, 1863
So what is data capital and how can we value it? The problem is that all businesses have data in some form, but it is worthless unless we access it. So what transforms dormant data to data capital? To guide us on our quest, a good definition of data capital can be found in this MIT article, where data capital is described as:
- Non-rivalrous: A single piece of data can be used by multiple algorithms, analytics, or applications at once.
- Non-fungible: One piece of data can’t be substituted for another, because each carries different information.
- Experience good: The value of information can only be attained by knowing the information itself. But once known, the information can be easily replicated.
Aligned with the above, data needs to be used (experience good) to capture its true value, it can be used many times (non-rivalrous), and it cannot be valued by itself since it is not a commodity (non-fungible). To value data we therefore also need to value it indirectly through its value creators: Data Management, AI, Analytics and integrations into customer and operational processes.
So my suggestion is the following: Each year calculate the project investments that have gone into analytics tools & platforms, AI-algorithms, process automation, automated marketing, data collection, data storage, data management and similar. Add this value to the same depreciated investment for all prior years (see below for depreciation). The total sum is your data capital.
Your company is doing this today implicitly by capitalising its investments. What I suggest is simply to key out explicitly the investments that go into data, AI, analytics and its integrations. The suggestion is not to measure the value of data itself. Instead we are using a proxy : We are measuring how much we invest in the field of data, AI and analytics. If the data is just lying there it has no value so you need to invest to activate it. Note however, we are not yet measuring how smart or good the investment was (more on that below). We measure just the money invested — smart or dumb allocation: This is the Data Capital.
If you are unknown to the concept of depreciation, here’s a short primer: An investment will loose its value over time due to age. In accounting this is done by setting a conservative life span of an investment and then writing down the value in equal pieces all the way down to zero for that lifespan. For some assets such as machinery or buildings, depreciation time can be set as high as 20 years. With AI and data the life is much shorter as innovation in the space is so fast and investments quickly become a burden if you do not reinvest or upgrade. Hence, probably 3 years is a prudent lifespan, maybe max 7 years for the data management assets.
Example 1 Let’s view the calculation through a simple case in table below. Imagine Brightfuture Inc and it’s investments in Data and AI. Brightfuture’s strategy is to make Data Capital their most important source for growth and profitability in a 5 year period. The lifespan is the prudently set life for each line at the time of investment. As the max lifespan is 4 years, we only show 3 years back. All other investments in these areas 4 or more years ago are valued to zero so not relevant. Based on that we can calculate the total depreciated value of the four year investments - Brightfuture’s total data capital (valued to $1.5 Million).
Return on Data Capital
An investment in knowledge pays the best interest — Benjamin Franklin, 1758
Having defined data capital, we now need to find a good measure on how well we are directing our investments and capturing value. In finance, there are many KPIs for calculating profitability, but one of the better measures is maybe return on capital employed — ROCE :
Capital employed, also known as funds employed, is the total amount of capital used for the acquisition of profits. It is the value of all the assets employed in a business and can be calculated by adding fixed assets to working capital or subtracting current liabilities from total assets (read more).
The reason why ROCE is such an efficient measure is that it can be used to reliably compare profitability among companies within similar industries. Studies also show that companies that have healthy ROCE ratios also have high Price / Book ratios and stock returns, meaning that future expectations are high. Stock markets value companies for growth, but not revenue growth for growth’s sake. Hence, high and increasing ROCE is a good measure that the stockholders’ equity is not eroded while having a healthy growth.
We use these insights to our mission to control our data and AI investments. Let’s define Return on Data Capital (RODC):
Just like ROCE, by measuring growth in RODC over time you balance both growth and long term sustainability of a company’s data capital.
Data and AI driven EBIT needs some work to set. As revenues and operational expenses are not attributed to specific activities in income statements we need to make some assumptions to set the measure, while staying prudent. Let’s leave out indirect effects like better decisions from better insights & BI because it is very difficult to prove attribution to data and analytics. The direct revenues and cost savings on the other hand is fully attributed so that balances the equation. The same goes for all expenses to drive the capability. Lastly notice depreciation of data capital. As data investments currently are rather short lived, this post can be high.
Example 2 Continuing our Brightfuture Inc case, imagine a budget Income statement for Data Capital related lines. To make it simple we have left out all other income or cost. Observe the low data operating profit. To a large degree this is due to cost of depreciation on large past investments. These are not bringing out new revenue or AI driven savings. The result is a meager RODC of 3.8%. To compare, the ROCE over the last 25 years in capital intensive industries (ex: paper, steel, cement, aluminum) was 5.4%. Brightfuture really needs to step up their data investment discipline and tune their data operations if they shall deliver on their data and AI strategy.
Efficient Data Capital Management
We don’t have to be smarter than the rest. We have to be more disciplined than the rest. — Warren Buffet.
Next question we face is really how to manage our data capital for better return. To do that, let’s turn to a century old trick. If we rewrite the RODC equation above (3), you get the following important Dupont analysis version of the equation:
What this shows is that a firm can arrive at a higher return on data capital by two main avenues: (a) Increasing data capital turnover or (b) Increasing its data operating margin. Let’s revisit our example to extract ways to do exactly that.
Example 3 Based on the RODC analysis from example 2, Brightfuture Inc’s new CDO is not very happy with the value creation from the company’s data and AI investments. A taskforce is set up to do prioritization that does not affect the total cost budget for the area. Three actions are proposed: (i) Increase the spending on personalized marketing, (ii) Terminate the MDM service as it was not fulfilling its potential and the cloud data lake architecture could take its role, (iii) Extending the CLM ambitions and automate more core business processes. The projected result is a 22 times increase in RDOC with an almost 3 times increase in Data Capital Turnover and 8 time increase Data Operating Margin.
With the example fresh in mind, let’s return to how we can efficiently increase RODC in general.
(a) Increasing data capital turnover
This factor is driven by how smart investments are made and utilized to generate revenue. As it is effected by only the data capital value and the data revenue you can impact it a few ways:
- Increase utilization of your data capital. If you have made smart investments in the past, it is probably wise to sweat the asset harder and scale the activity. This is what Brightfuture proposes in action (i) in example 3 by extending their personalized digital marketing campaign.
- Stop maintenance and impair past investments that are not driving sales. This reduces your asset base under maintenance. This is the same as action (iii) in the example.
- Make short term AI investments that drive sales immediately with a higher capital turnover than. These investments need to have a higher ROI than your current capital turnover rate. In practice, such investments are hard to find, but sometimes unicorns appear.
All the actions above makes sense in isolation for capital turnover, but can give a negative side effect for the operating margin. Action (1) typically will effect the COGS (cost of sales), so make sure the gross margin stays the same or increases (see below). Action (2) gives the side effect of an impairment cost and change in operating expenses (lowering of maintenance of the existing asset, but increase in the maintenance of the alternative data capital asset)
(b) Increasing data operating margin
Data operating margin measures how well you run your Data and AI part of the business. To increase the margin, either the gross margin must increase or expenses relative to sales must decrease (or both). This leads us to several actions.
- An increase in the selling price for sales activities. In our case above, it could be to raise the consultancy hourly fee for the insights service.
- A decrease in the cost (COGS) per unit sold. In our example, that could mean a different channel/partner mix for your digital marketing initiative (ie Facebook versus Google versus programmatic advertising).
- Stop maintenance of data activities that are not driving enough business value. Prioritization is tough but often necessary because manpower in this space can be difficult to scale.
- Replace costly data activities with more efficient ones. This is as action (ii) in the example above.
- Scale AI automation measures to more use cases. This is as activity (iii) in the example above.
Make it happen
Knowing is not enough; we must apply. Willing is not enough; we must do — Johann Wolfgang von Goethe
We all know that there is a often a gap between a optimal foundation as described here from a purely business driven perspective and what happens in practice.
First of all, organizational setup around these areas are usually based on a history from Business Intelligence close to the IT function that is not really tuned to being a core part of the business and customer operation. That needs to change, but is outside the scope of this article.
Secondly and even more important, data capital is as introduced an experience good. The value of information can only be attained by knowing the information itself. This means that data will need to be part of an experimental setup that tests hypothesis an continually refines an idea before true value is crated and captured. You cannot plan yourself out of this, you must experiment.
Third and related to the second point, getting to a stage where Data Capital is really returning value measured through increasing RODC can be a yearly process. You need to have controlling mechanisms that are much closer to the day to day operations and prioritization.
To answer challenge 2 and 3, I here propose some leading KPIs for your data and AI operation.
These KPIs are examples, and I would love to hear other perspectives on good ones. The point of the KPI is that it actually can be measured daily and aggregated up to week, month & year. The KPIs should be leading indicators measuring that data is activated and data investments are really creating sustainable growth for the company. By tracking KPIs the organization can agree on the intent and start acting in a common direction. I described in The Customer Data Revolution how data ownership is moving from the business to the consumer and that a likely price pressure is 10% from this. This is to a large degree why Consent Conversion and Data Activation above is so important for the success of the model.
There are many reasons why organizations are struggling to capture value from their data like the best digital players are doing: Culture, process, competence, structures & technology are some of the areas that many are lagging on. What I have suggested here is that if we introduce a way to manage, measure and govern our data projects and investments organizations can start moving towards a common goal.
Even though everybody are talking about this stuff, you can be reassured that there is a lot of hype around this (just look at how far to the left different data and AI technologies are in this Gartner Hype Cycle — it will take years before a lot of this becomes mainstream). Hence, start now with efficient data capital management and you may be one of the most competitive industry players soon.