Data analytics, along with Blockchain, AI, IoT and ML, is one of the dreaded buzzwords.
It has been positioned as an elitist, mysterious marvel. This positioning is incorrect!
Data analytics is used subconsciously by you, dear reader, every day. Because of this, it is actually understandable, accessible and relevant. And, when used in business, it is the “secret sauce” that helps us make more money, and fail less.
Risk can have many forms – but it is linked, in its heart, to two levers we hold dear: greed and fear.
Here we tie fear – or, more accurately, our desire to manage risk – to data analytics.
And further, to make this really live, we bind data analytics to commercial real estate (CRE).
Example
To explain data analytics, we’re going to lean on an accessible, real life example, powered by the world’s oldest and most sophisticated computing machine (i.e. your brain).
Data analytics is a two-step process, with the steps best summarised into two roles.
The “Miss Moneypenny” (step 1) world of data analytics is the “low glamour” bedrock of the discipline: data collection, organisation, and storage.
The more glamorous analysis and translation piece (step 2) is better understood as the “James Bond” of data analytics.
As is hopefully clear, James Bond cannot succeed without Miss Moneypenny.
The unexciting “Miss Moneypenny” processes that form the bedrock of the discipline are easily understood. This don’t-pass-go-without step is worthy of a separate article in its own right. (Here, in the interim, is an excellent overview). This part of data analytics relates to the people, processes and technology required for corporate data governance. Phrases you may have heard in this regard: master data management, “one version of data truth”, “master and slave”, “deduplicate and merge” / “de-silo”, normalise, exception reporting, standardisation, data dictionaries). (Incidentally, our human brain, the best supercomputer of all, handles this deceptively high complexity exceedingly well.)
The glamorous “analyse and translate” James Bond role is less accessible, so it is unpacked as follows.
As a committed first year varsity student, you wake up on a Sunday morning with a brand new, pounding headache.
Your brain automatically starts working through the four, sequential analysis and translation steps of data analytics.
It kicks off by running descriptive analytics (i.e. what) on its various databases. As a result, it mines out the following gold: “I have various pains (but the greatest is located between my ears), as a result of a, somewhat bewildering, but impressively varied range of actions starting yesterday and ending about 4 hours ago”
Your (sore) brain moves on to diagnostic analytics (i.e. why): “Is it malaria? (No, I was in the bush over 3 weeks ago). Is it sport related? (No formal contact sport yesterday). Booze? (Jackpot). But why the intensity? Mixed drinks (check), no water between drinks (check), started dehydrated and on an empty stomach (check, check), felt it necessary to smoke cigarettes (urgh, check), went to bed late so limited sleep (check).”
Fast forward a few years. Hopefully, after collecting and storing sufficient structured data samples (aka benders), you start moving into the world of predictive analytics (i.e. “what if” / what is likely to happen). For example, every time I have this combination of activities, involving this selection of people, in this location, and a commencement state, it is likely (95% probability) that I will be awoken by a jackhammer pounding my brain.
Prescriptive analytics (i.e. what action to take) to prevent this recurring: you clearly define an algorithm to your saintly partner (i.e. when these specific variables interact, an action should happen). Your request may be to kindly trigger you with an exit suggestion, when circumstances are such that high-risk stimuli are in play, while you are still receptive to rational decision-making. (E.g. Mike, your soon-to-be-28th-best-mate, is bringing out the tequila again, now is a good time for us to exit).
To all readers, now feeling like highly accomplished data analytics experts out there, well done. And yes, only now will it dawn on your parents that your behaviour, more fairly termed “field research”, was always fueled by a deep love of data science.
Going back to relevance, this process is not only used for booze. If we are fortunate to be a reasonably competent adult, we use this data analytics-driven learning loop throughout our life, and in all aspects of our life – from relationships through to business.
Now is when we fuse together the concepts of risk (remember the fear and greed), and data analytics, and apply them to real estate.
For purposes of simplicity, we’re going to focus on a subset of our noble and beloved asset class – commercial real estate (CRE).
Here we call out 10 ways in which data analytics helps reduce risk in commercial real estate, unpacked against four select industry verticals, namely: brokers (sections 1-2), property funds/owners (sections 3-6), debt funders/banks (sections 7-8) and insurers (sections 9-10).
Remember the significant financial pain caused by the failure of a lost “dead-cert” deal? How does data analytics help reduce this risk?
First, the Miss Moneypenny bedrock stuff… All vacancies data needs to be collected. Next, the data needs to be accurate, standardised and stored in a highly structured and precise format so that it is search friendly.
For example, vacancies data:
Now for the glamorous James Bond part…
Enter search criteria (dates, locations, minimum and maximum values) into a sophisticated engine so as to identify those results matching your criteria.
Filter out properties from the universe of options, and output the info into a tenant-friendly brochure/report (linking back to a website), specifying decision-relevant property attributes.
This complimentary “dream team” of roles, working in tandem, means you don’t fail to identify the perfect sub-divisible property located in an overlooked suburb deep in your target area. Result – your client thinks you are amazing, and you also do the deal.
Your competition does not. And wonders how you did.
Time is money, and you are not operating in a competitive vacuum. Your clients expect this entire analysis and translation process to happen in seconds, not hours. And your data needs to be reliable: incomplete or inaccurate data wastes time, and takes a scalpel to your reputation.
So, if you want to succeed, tech-driven data analytics is the only solution.
Hopefully it is now clear that the easily overlooked data analysis “bedrock” of collection, organisation and storage is non-negotiable. Without it, there is no ticket to proceed to the James Bond world of analysis and translation.
As the saying goes, “junk in, junk out”. Or, the converse: high quality data equals high quality decisions.
A decision about any new lease or lease renewal is principally a data-driven decision, leaning on data analysis and translation. If you, wearing the hat of landlord, position your vacancies or lease renewals above market value (greed), you stand to either miss or lose high quality tenants, or go for periods when the space remains vacant and you lose income. Under-price your product (fear), and your widows-and-orphans shareholders get hurt, while tenants pay below market rentals.
The Miss Moneypenny role of data analytics ensures that a complete, accurate and well-structured population of relevant information is available. The James Bond solutions empower decision-makers to make data-driven decisions, quickly.
Compliance is boring, and so is how you go about preventing this risk…
A powerful antidote to this boredom: the consequences of acting negligently or recklessly when reporting wrong information to shareholders and regulatory bodies.
Risk mitigation here talks to the low glamour Miss Moneypenny of data analytics: the collection, organisation and storage piece.
Firstly, their data is not recorded completely and accurately at point of capture. This means that critical bits of missing information are not collected, organised and stored at inception.
By the time you want to report on this data, it’s too late and your ability to report is hamstrung. “Band-Aid” solutions simply dial up risk. Without recording correctly at point of capture, you’re playing catchup in a race that you have already lost.
Data silos (sometimes-duplicated clusters of data used by various divisions / in various processes) are a major issue for property funds, which often have masses of data spread across numerous, disparate systems.
Data collected, organised and stored correctly ensures that the right data can be reported timeously, and that processes to correct data early on can be implemented.
What happens when key individuals move on? When valuable, proprietary information lives in human brains, instead of being stored digitally, in structured systems?
Data assets cannot be backed up, “data debt” is created, and unhealthy human dependencies arise.
Two drivers here. Firstly, data is not stored correctly at point of capture. Secondly, data reporting is manual (instead of managed by automated, efficient processes). Therefore, the problem has to be “Band-Aided”.
The problem: missing data analytics best practices – the Miss Moneypenny correct data collection, organisation and storage, and James Bond analysis and translation solutions.
The sub-par solution: a non-scaling “Band-Aid” of employing highly skilled, expensive, high-opportunity-cost-of-time staff, working in near perpetual motion, doing largely high-risk “hands work”.
Ignoring the opportunity cost of poor utilisation of highly skilled humans (these humans are expensive!), the “Band-Aid” solution of employing humans is more expensive than the maintenance of a well-structured data analytics solution.
This difference in costs results in reduction in profits. The effect: reduced distributions.
In commercial real estate, unlike residential property, individual asset values are bigger. Portfolio valuation issues don’t “come out in the wash”, and aren’t solved by the laws of large numbers. Debt funders need to know the value of individual assets. And debt funders need this data sooner rather than later.
Two examples of this discipline enabling more proactive credit risk management / reducing delays in action are laid out below.
In a loan to value (LTV) ratio, assuming all funding comes from a single source (and assuming no cross-collateralisations), the debt funder can be certain of the debt amount.
The property value, however, is prone to variation. Borrowers, in terms of their funding terms and conditions, can be required to submit data to their lenders on an annual or as-required basis. In certain instances, such submissions may not coincide with financial reporting periods, or worse, arrive too late…
Factors that can hurt property values can be categorised into two groups:
Data points that are internal to the property. Such data can be further subcategorised:
External factors – i.e. generally, public domain data
Macro nodal negatives – such as increased vacancies, declining property rentals, changing uses and nodal regulations – are one example. These factors can and will reduce the value of all properties located within a node. Another example would be a key retailer, facing profitability or insolvency challenges, hurting various customer portfolios values.
Provided data is collected, stored and organised correctly, the foundations are in place for decision-relevant, proactive analysis and translation interventions. For example, borrowers are bound to legal entities, which in turn are categorised into parent entities, property assets are geolocated to enable hyper-accurate geospatial analytics, property category composition is accurate, tenants are uniquely identified and bound to portfolios, etc.
Descriptive and predictive data analytics builds on these foundations. By presenting lenders with user-friendly dashboards, both internal and external property value issues can be revealed. Such data enables lenders, in near real time, to identify exposure to, and mitigate, credit risk.
Ignoring the foregoing preventative measures, such data, delivered at lower cost and greater frequency than traditional property valuations, can also assist lenders in impairing loan assets.
The ability to call out a property’s true (over-)indebtedness and identify high risk borrowers using an expanded view of their debt exposure.
In less information-perfect markets, it is possible for borrowers to secure debt against a property from various sources. This action increases the sum of loans against a property’s value. In certain instances, existing borrowers may not be aware/notified of such additional gearing.
While seniority of debt reduces risk to primary funders, it can constrain cash availability for a borrower. Such cash constraints can divert funds from critical repairs and maintenance expenditure and incentivise the appointment of lower-quality service providers. This “death spiral” can result in reduced property values.
Furthermore, over-indebted borrowers will likely have a shorter loan duration and will generally attract higher client management costs.
A property portfolio is not a static asset. Property owners are free to make gearing decisions with various third parties, on various assets under their control.
When a debt funder has a holistic view of a borrower’s portfolio, by property and loan value, it provides that lender with a view of the borrower’s financial stability.
A borrower with a falling property portfolio value and a higher portfolio level of debt can be tempted to “rob Paul to pay Peter”. Across a distressed portfolio, high quality assets can tend to cross-subsidise lower quality assets. As the profitability of higher quality assets increases, it constrains resources of a property fund’s management team to successfully operate all assets.
When a lender has a view of all property and debt in a borrower’s portfolio, and access to debt and value changes / trends within the portfolio, such a lender has a greater ability to proactively act on risks. Descriptive data analytics surfaces such insights to decision-makers, while predictive and prescriptive analytics mitigate future risk.
In the instance of a high performing property fund, property values should increase over time, and operator risk should decrease.
Assuming palatable LTV ratios are fixed, as property values rise, so too does the value of potential or “unsweated” increases in debt funding.
By way of example: a highly entrepreneurial and hyper-competent property fund purchases US$10bn of distressed assets, funded by interest-only debt at an LTV of 50%, 3 years ago. That property fund has actively managed those assets, and the value is now US$15bn. At a 50% LTV, on the new values, the debt funder can safely increase funding by US$2.5bn.
Where the debt funder fails to step in, others can, and will.
Data analytics can identify these patterns, unveil borrowers with a higher propensity to borrow, and equip prospective lenders with the data to approach such prospective borrowers.
While this is not risk related, savvy debt funders can extend this model. Given the correct data, one can identify financing opportunities in competitor portfolios, and act on these.
Further, as data availability matures, it becomes possible to understand how active the market is, identify loan transactions by competitors, understand the terms, and actively counter moves made by competing debt funders.
Additionally, referencing the points regarding improved data availability in section 7 above, and applying them to competitor risk, the benefit of high quality, decision-relevant data is transaction speed, and reduced friction. When banks have access to complete, standardised, high quality (up to date, accurate) borrower information, such data speeds up the credit process, allowing that bank’s clients to get money quicker, and enable banks to disburse funds faster.
Here’s an example: red sports cars are expensive to insure. But a red sports car driven by a 25-year old female accountant with a clean driving history is significantly better risk than the average. Further, a white family saloon, driven by a 25-year old male with a criminal record, carries more risk than the red sports car.
The point of this? Making risk calls on high level data, rather than granular, risk-specific data will always create risk mispricing.
A data-driven approach to CRE risk empowers insurers to price risk more accurately and avoid the spectre of good risk subsidising bad risk.
As always, it starts with the role of Miss Moneypenny – how data is collected, organised and stored.
In processing the following non-exclusive list of fields (and ignoring office, industrial and retail), a property insurer is able to evaluate the risk of a generic property. This means not outpricing the red sports car driving accountant, and correctly evaluating the white family saloon driving ex-convict’s risk.
Applying algorithm-driven predictive and prescriptive analytics enables risk to be more accurately priced. Good risks get good pricing and are attracted. Higher risks are accorded higher pricing. Assuming policy holder decision makers are rational, these (unattractive) policies will flow towards those businesses who price based on the white family saloon data alone.
Insurers win when claims are prevented, and policy holders win when policies are cheaper. In this regard, both parties’ objectives are aligned. Data analytics can be activated to incentivise policy holders to reduce their risk, and encourage risk-reducing behaviours.
Predictive analytics: “As a result of the data we have available, your risk has reduced, your policy now falls within tier X, and you pay a reduced premium”
Prescriptive analytics: “Your next steps to further reduce your policy risk, ordered by ease of execution, and based on what our claims data is telling us, will be to take the following actions”
This level of proactive, data-driven reporting is a direct benefit of high-quality data analytics.
Because data is the oil that CRE runs on, the industry provides a highly fertile environment for data analytics to reduce risk – and increase revenues.
All of the above feels like a when-not-if…so why the delay? Why is the real estate industry not benefitting, now, from these solutions?
Firstly, CRE is exceedingly complicated: assets are not homogenous, and industry players have not been gifted with the protocols and tools to standardise how their data is stored and structured. Various verticals within the CRE industry own and manage data uniquely. The domain is deceptively complex, and insiders and outsiders alike often don’t-know-what-they-don’t-know.
Interventions by industry outsiders have followed a try and fail pattern, burning the fingers of passionate innovator/early adopter project sponsors inside of these CRE organisations. As those who have paid for this lesson have found out, competencies for engineers skilled in banking, insurance, FCMG and logistics industries are not automatically transferable to CRE. “Once bitten, twice shy” – it is understandable that decision-makers are reluctant to engage in high-risk, career-limiting behaviour.
Secondly, everyone assumes data analytics is a silver bullet. However, it is a two-step process – with the first step being time-consuming, expensive and slow to deliver the impactful, action-packed James Bond second step insights.
The first step involves the Miss Moneypenny disciplines of data being collected at source correctly, organised appropriately and stored right. Such processes are deceptively complicated, highly unglamorous and oftentimes more expensive and time-consuming than is obvious from the outside-in.
What happens: the “hard yards” work is skipped or constrained, in a quest for the exciting wins. Unfortunately, the biblical parable of the builders comes to mind – those who build their houses on sand fail, while those who build on solid rock succeed.
Thirdly, “If it ain’t broke, why fix it?”
While the solutions are in a relatively immature stage of creation, the benefits of data analytics (while growing) are neither obvious, nor significant. At current levels of solutions maturity, the cost-risk-return model is unattractive. As is standard with all tech adoption lifecycles, only when the benefits are undeniable, and the innovation is mainstreamed, will we see data analytics playing front and centre in CRE.
The three points above makes the US$29tn commercial real estate industry one of data analytics’ last frontiers. This fusion of new age technology with hard-won CRE experience is pioneering, and tremendously exciting.
James Bond villains carry guns and come in all shapes and sizes. Data analytics villains are the business evils of risk, inefficiency, and lower profits. In both instances, hard work, passion and superior wit will ensure that the heroes defeat the villains.
Hopefully after reading this you feel less intimidated about the discipline, have your beliefs reinforced about where your energies can be well spent, and are inspired to take steps forward in this rewarding and exciting journey.
We wish you good luck!
In a long-overdue industry first, PropTech is now delivering professionally-designed, automated vacancy schedules to the CRE industry. You time-starved CRE professionals can now get your vacancy schedules to work as hard as you do, giving your vacant space the best chance of finding the right tenant.
Good news – CRE professionals can now seamlessly get their stock for sale or to let onto the listing platforms. With one click of a button. These data flows are now automated, making CRE more efficient. Say goodbye to double capture.
How the best CRE brokerages are applying the economic principle of opportunity cost to automate, outsource and focus. The result: higher profits, happier staff. We go from the theoretical to the practical, unpacking this in hard numbers.
