21 July 2026
The word “doctor” meant “teacher” in Latin. The meaning spent a few centuries drifting from the person who instructs you to the person who examines you. That drift is the fate waiting for “test and learn”.
Anna Goss has a great post about the Riskiest Assumption Test here: From the RAT to the Magenta Book.
“It had long since felt like MVP had lost its way in the five years since The Lean Startup had come out. I can’t find it in myself to go and remember all of the many different arguments around what is an MVP, definitions of what it means, how to scope and test it, but the phrase jumped the shark, and the RAT turned up into a wider product environment that was desperate for a new way of talking about validating propositions, starting small, and testing as you go.”
How MVP felt to me, managing a software product in 200x was: the MV in “minimum viable product” meant that the market will decide if the product is minimally viable. You’d ship, see what the market thinks, hmm not much, that MP isn’t V yet, so let’s go again. But I think once teams saw the MV in “minimum viable product” as a tool of product management - a matter of deciding how much scope - the MVP term lost its force. But really the term had been ambiguous and evolving since birth 1.
Rather than re-reading Saussure and Derrida I’m going to confidently handwave that the semantic drift happens faster if the names are popular. And it’s probably slower if there’s an enforcing mechanism - words in regulated industries like “doctor” hold their meaning better. Either way, meanings change, get diluted, co-opted, redefined. So when you’re talking to someone about “agile”, “digital transformation”, “brand” or “comms”, you have to check what they mean. And they have to check with you.
This drift happens to all words, doesn’t it? It is what language and communication are, meaning negotiated and renegotiated by a speaker and a listener and the context, and the same happening at societal level, with no final anchor for meaning, so you hold hands and try work it out together.
“last month, the Treasury published their Test and Learn annex to the Magenta Book, the government’s evaluation handbook. The word assumption shows up on page 3 in the glossary:”Criteria that need to be met for a policy or intervention is to have the intended impact. These may or may not hold true in practice. Test and Learn activity prioritises assumptions that are most uncertain or risky, as these pose the greatest threat to achieving outcomes. […]
But it’s not done, is it. So we keep chasing Gatsby’s green light (or the Treasury’s Green Book), boats against the current.”
The work it stands for isn’t going anywhere - but the name might. So we get to watch what codification does: does making “test and learn” official in the 2026 Magenta Book slow the drift, or speed it? Or maybe neither, the tides of political change may matter more than the handbook.
Either way, “test and learn” will one day be borne back ceaselessly into the past.
“Minimum Viable Product” is itself a case in point: In 2001 Frank Robinson defined MVP in terms of product scope that delivers value: “the product with maximum ROI divided by risk”. In 2005, Steve Blank had it as a “minimally scoped product that gets some job done for your chosen customer archetype”. By 2011, Eric Ries had MVP as a validated learning loop: “the most minimum version of their product that will still allow them to learn”. More useful reads on the history: Adeocode, Paul O’Brien.↩︎
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.