The 4 V's of Data
Size up a data problem - and remember veracity decides the rest.
Faced with a data or analytics problem, the 4 V's — volume, velocity, variety, veracity — give you a quick way to size it up and ask the question that matters: is it worth acting on? Three of the V's describe the data; the fourth decides whether the other three mean anything.
TL;DR · Key Takeaways
What you will be able to do
- Size up a data problem across volume, velocity, variety, and veracity.
- Treat veracity as the gate — untrustworthy data makes the other three worthless.
- Avoid over-investing in scale and speed before confirming data quality.
- Keep it as a niche lens for data-and-analytics cases.
The four V's
Volume — how much data is being generated (enough to matter, or too much to handle)? Velocity — how fast does it arrive and need processing (real-time, or overnight batch)? Variety — how many forms and sources (neat tables, or a mix of text, images, logs)? Veracity — how accurate and trustworthy is it? Vast, fast, varied data is worthless if it can't be trusted.
How to use it
Size up the data problem across all four V's — but check veracity first. There's no point engineering for huge volume and real-time velocity if the underlying data is wrong; you'll just make bad decisions faster. It's a niche lens for data-and-analytics cases, not a general business framework.
Optimising the wrong three V's
Teams love to build for scale and speed — volume and velocity — because they're tangible engineering goals. But if veracity is poor, that investment amplifies bad data. Confirm the data can be trusted before scaling how much and how fast you process it.
Worth acting on?
A retailer wants to build a real-time recommendation engine off its customer data and is excited about the volume it has. How would you frame the data question?
I'd size it on the 4 V's but lead with veracity. Volume — they have plenty, good. Velocity — real-time is ambitious but doable. Variety — transactions, browsing, maybe support logs, manageable. But veracity is the question I'd press: how clean is the customer data — duplicate accounts, stale profiles, mis-tagged purchases? If veracity is poor, a real-time engine just recommends confidently off bad data, which erodes trust faster than no engine at all. So before investing in the volume-and-velocity build, I'd fix data quality. Veracity decides whether the rest is worth it.
Puts veracity ahead of volume.
The candidate reordered the excitement — putting trustworthiness ahead of scale and speed — which is exactly the discipline the fourth V enforces.
Where this connects
It's a specialised lens for data and analytics cases — a structuring aid that sits alongside the broader tools in Structuring Fundamentals when the subject matter is data.