The first version was leak detection. The idea was sound: use operational data to identify potential water main issues before they become emergencies. The technology worked. The pitch didn't.
Turns out nobody wants to hear about infrastructure failure. There's no emotional hook. Homeowners don't wake up thinking about their water mains. They think about their kids, their pets, whether the water they're drinking is safe.
So we pivoted.
Iteration two: bacteria screening
The second version focused on bacterial contamination. We could flag potential contamination events from water quality data faster than traditional lab testing. Technically impressive. Also terrifying.
The deeper problem was that we had put a public-health function inside a consumer engagement product. Formal contamination notification belongs to the utility and the regulator, who have the verification steps and the legal duty to issue it. We had no business being the channel that tells a household to stop drinking its water. Building it that way would have been the wrong call even if it had converted well.
So we pivoted again.
Iteration three: lab testing
Version three tried to split the difference. Instead of alarming language about contamination, we positioned the product around proactive water quality testing. A health checkup for your water.
The problem was speed. Lab testing takes days. By the time results came back, the urgency that drove someone to request a test had faded completely. Engagement rates were abysmal. People would request a test, forget about it, and never open the results.
We pivoted.
Iteration four: real-time monitoring
The fourth version was closer. Real-time water quality data, updated continuously, accessible through a simple interface. No waiting for lab results. No scary contamination language. Just data.
But data alone isn't compelling. A dashboard showing pH levels and turbidity measurements means nothing to someone who just wants to know if their water is safe to give to their dog. That's the actual question. We still weren't answering it.
One more pivot.
Iteration five: personalized risk context
The version that worked combined everything we'd learned. Real-time water quality data, personalized by ZIP code, with risk context tailored to what people actually care about: pets, infants, and personal health.
Instead of "Your water's pH is 7.2," the system says "In your ZIP code's most recent reported samples, no contaminant exceeded its federal legal limit." That is a statement about measurements against an enforceable limit, not a health verdict for a household, and the product is explicit about the difference on the same screen. Instead of contamination alerts, it provides context: "Seasonal runoff in your region typically affects turbidity. Here's what that means for your household."
This version is the one that landed. Not because the technology was different from version one. The underlying data pipeline is essentially the same. What changed was the framing, the personalization, and the emotional context.
Why most firms stop at iteration two
Here's the thing about pivoting: it's expensive. Not just in development hours, in organizational patience. Every pivot requires the client to believe the next version will be better than the last. Every pivot means explaining to stakeholders why the previous version didn't work.
Most consultancies can't survive this. Their engagement model is built around fixed-scope deliverables. Version one is the deliverable. If it doesn't land, they write a recommendations document and move on. The recommendations might even be good. But nobody's there to execute them.
We survived those pivots because our engagement model is built differently. We don't scope deliverables, we scope outcomes. The outcome was a product that drives customer engagement. It took several tries to get there, and the client funded each one because they could see the learning in it.
The lessons
Emotional resonance beats technical accuracy. Every version of the product was technically sound. Only the one that connected emotionally succeeded.
Speed kills slowly. The lab testing version was accurate and trustworthy. Too slow for human attention spans. Real-time feedback, even if less precise, wins.
Personalization is the unlock. Generic water quality data is a commodity. Water quality data contextualized for your ZIP code, your pets, your kids, that's a product.
Iteration requires trust. Five pivots only happen when the client trusts the team. That trust is built by embedding, not by presenting. You can't earn pivot-level trust from the outside.
What I'd tell my earlier self
Start with the emotional hook. Always. The technology is table stakes. The question isn't "can we build this?" It's "will anyone care?"
If I'd started with iteration five's framing and iteration one's technology, we might have gotten there in two pivots instead of five. But honestly, I'm not sure we would have understood why the framing mattered without living through the failures first.
Some lessons you can only learn by building the wrong thing.
PurviewX is embedded AI leadership for companies sitting on real operational data. We find out whether your AI actually works, including ours. Start a conversation.