As ‘big data’ gets more and more attention with its fancy algorithms and predictive analytics tools, I can’t help but think that most problems companies face are really ‘small data’ problems. These problems arise when management does not actively measure and manage the fundamental value drivers of their business. At a high-level, these value drivers are the components of the customer lifetime value (LTV) formula, such as revenue, cost to serve, retention rate and customer acquisition costs. But to really have an impact on these value drivers, managers must focus on their actionable sub-components — the drivers of the drivers, so to speak — or what I like to call ‘small data’.
Actionable Small Data: An Example
One company that I once worked with managed a chain of ~30 physical therapy clinics. To better understand the drivers of revenue, I disaggregated each clinic’s revenue into the following ‘small data’ sub-components:
# of patients * # of visits per patient * frequency of visits * # of procedures per visit * $ reimbursement per set of procedures
Similarly, I disaggregated their cost structure:
# of physical therapists (PT) * hourly labor cost per PT * labor utilization (# of labor hours worked / # of labor hours spent with patients) + # of square feet * ($ rent per sq ft + other overhead per sq ft)
‘Small data’ metrics aren’t just related to the income statement — the balance sheet also matters. So, I broke down the cash collection process into its different components, which included:
Average time taken to process insurance claims and percentage of claims denied for addressable reasons (e.g., legitimate claim that was improperly coded)
The power of measuring these pieces of ‘small data’ was huge. When we benchmarked average revenue reimbursement data by clinic, we identified insurance contracts that needed to be renegotiated for more favorable terms and determined which clinics could improve their billing procedures for higher reimbursement rates. We also identified best practices for collecting insurance information upfront which dramatically reduced the cash collection cycle (often greater than 120 days). Most importantly, we were able to start using ‘small data’ to better manage the clinics by giving them actionable metrics to which their employees were held accountable.
Why ‘Small Data’ Problems Exist
The good news is that ‘small data’ problems do not require sophisticated algorithms or an army of data scientists. What they do require is: 1) identification, 2) systematic measurement, and 3) maniacal focus with supporting incentives. Sounds pretty basic… so why don’t most businesses manage ‘small data’ effectively?
1) Identification: The absence of the first requirement is a failure of modern business education. Writing down the profit equation — and it’s ‘small data’ sub-components — for your business model should be the first thing taught in a graduate or undergraduate business school curriculum (sadly, it’s not). The Lean Start-Up movement is beginning to address this problem with its focus on actionable metrics… and at a fraction of the cost. The good news is that it’s a relatively straightforward exercise to write down the profit equation and its ‘small data’ sub-components for your business.
Recommendation: Start by giving yourself and your managers the homework assignment of writing down the ‘small data’ that drive value for their functions / business units. Then work with then to identify which drivers have the most impact on the business and make sure they focus on them.
2) Systematic measurement: The blame for the absence of the second requirement should be placed squarely on three parties: 1) the Financial Accounting Standards Board (FASB) which promote GAAP accounting, and 2) the creators of general ledger software, which is used to present GAAP financials, and 3) investors who use GAAP financials to value companies. If general managers don’t (or shouldn’t) use GAAP financials to manage their business, then why is a whole ecosystem of regulators, software vendors and investors so focused on them? Putting my soap box aside, ‘small data’ usually exists in various tools and databases that can be aggregated to calculate your actionable metrics. It may take some time (and money) to get the data to talk to each other, but it’s probably there. If not, you’d be surprised how well manual data collection in Excel can work just to get things started (though not for the long-term).
Hypothesis: There is a shit ton of money to be made for an entrepreneur to make financial software that tracks ‘small data’ for companies and translates them into GAAP financials (vs. the other way around). (Note: If you are a programmer who is interested in starting such a business, let me know!)
3) Maniacal focus: If a business doesn’t have the knowledge or measurement tools, then it’s unlikely that its managers will be able to incentivize employees using these ‘small data’ metrics. In the absence of this information, managers will instead rely on heuristics and intuition, which often leads to mediocre outcomes. As mentioned above, however, the process of identifying your actionable metrics is just a homework assignment away and many of them can be tracked with the most basic of tools. Once small data is linked to performance assessment, it’s amazing how quickly behavior changes. It’s not that your employees are bad at their jobs, it’s just that they’ve never known where to focus their time and energy.
The problems ‘big data’ are trying to solve are directly tied to ‘small data’ problems. Algorithms are often used to improve conversion rates and reduce attrition, both of which are ‘small data’ metrics linked to the LTV formula. But businesses must first learn to crawl (and walk) before running. So start by identifying, systematically measuring and maniacally focusing on small data — the fancy algorithms will still be there when you eventually need them.
Short URL: http://tinyit.cc/ceee0 Tweet: