New-Data Value Proposition Test
Find investments that turn newly available data into a better decision or service
- Difficulty
- Moderate
- Time to result
- ~weeks to results
- Steps
- 6
- Confidence
- 94%
The New-Data Value Proposition Test screens an opportunity by asking what fundamental customer offering has changed because data now exists, arrives faster, or can finally be analyzed. The process begins with a real decision or operational problem, identifies the new information source, and traces how that information produces a better action. Murray's examples include using phone activity to assess borrowers without bank accounts and using industrial sensor data to predict equipment failure. A strong opportunity does more than collect data: it improves access, timing, accuracy, cost, or prevention in a measurable way. The final test considers consent, cybersecurity, operational integration, and whether competitors can reproduce the advantage. The output is a clear investment thesis linking novel data availability to an economically meaningful result.
Origin
Extracted from Masters in Business through Steve Murray's discussion of the investment pattern behind Tala and Uptake.
Core principles
- 01Start with a changed value proposition, not technology for its own sake
- 02Seek data that was previously absent, inaccessible, or too slow
- 03Connect data to a consequential decision or action
- 04Validate that the output improves speed, access, cost, or accuracy
- 05Account for trust, privacy, and security
How to run it
- 1
Name the Existing Constraint
Identify a valuable decision or service currently limited by missing, delayed, or expensive information. Specify who experiences the cost.
Pro tip Look for decisions still based on sparse records, scheduled inspections, or crude proxies.
Watch out Do not begin with a dataset that lacks a meaningful customer problem.
- 2
Identify the Data Shift
Determine which data now exists, can be accessed, or arrives at a useful pace. Explain what technological or behavioral change made this possible.
Pro tip Distinguish genuinely new availability from data that is merely easier to market.
Watch out Access may disappear if it depends on unstable platform permissions.
- 3
Map Data to Decision
Show how the data changes a credit decision, maintenance action, diagnosis, or other concrete process. Make the causal chain explicit.
Pro tip Reduce the thesis to data input, analytical process, and operational output.
Watch out Insight without an actionable workflow rarely creates durable value.
- 4
Quantify the Improvement
Measure gains in access, prediction, cost, speed, uptime, or risk. Compare them with the incumbent process.
Pro tip Use a customer outcome such as avoided downtime rather than an abstract model score.
Watch out A statistically better model may not produce an economically meaningful improvement.
- 5
Test Trust and Integration
Evaluate consent, data security, bias, regulation, and the customer's ability to act on the result. Ensure the product fits the operational system around it.
Pro tip Treat trust infrastructure as part of the product.
Watch out Sensitive data can destroy adoption or create harm when governance is weak.
- 6
Assess Defensibility
Determine whether feedback loops, proprietary access, integrations, or accumulated outcomes strengthen the product over time. Check how easily an incumbent can copy it.
Pro tip Look for lawful data advantages that improve with usage.
Watch out Raw access to widely available data is rarely a durable moat.
In the wild
Tala evaluates selected activity on a prospective borrower's phone when that person lacks a conventional bank account. Correlations with prior repayment behavior support a credit decision that traditional records cannot provide.
→ Previously excluded borrowers can receive microloans based on an alternative evidence stream.
Uptake analyzes sensor data from industrial equipment to detect signs that a tractor may soon fail. The warning prompts maintenance before a breakdown interrupts farm work.
→ The operator converts continuous equipment data into avoided downtime and repair costs.
Common mistakes
Collecting Data Without an Action
A large dataset is not a value proposition unless it changes a consequential customer decision or workflow.
Ignoring Trust Costs
Products using personal or operational data can lose adoption when consent, security, and governance are treated as afterthoughts.
Mistaking Availability for a Moat
If every competitor can access the same data and reproduce the analysis, the advantage may be temporary.
Is it for you?
Best for
It is best for investors, product strategists, and founders evaluating data-intensive opportunities in established industries.
Not ideal for
It is not ideal for products whose value depends mainly on branding, entertainment, or craftsmanship rather than improved information.
From the transcript
“So, what we like to see is what are they doing now that has changed the fundamental value proposition of what is being offered?”
“are you using the availability of data that either didn't exist previous slee or wasn't it didn't exist at the pace at which exists now…”
“Uptake, which is using data from sensors on industrial equipment to do things like tell you that the tractor that's going across the farm in…”
From the episode
Steve Murray Discusses Startup Valuation
Steve Murray