Thirty indicators for something everyone called a culture

Continuous improvement existed at every site. Nobody could say whether it was working.

An 8 to 10 week sprint. One senior analyst and two juniors. A €14 billion global fertilizer and crop nutrition company. Part two of three on operational improvement. Part one was about standardising the work. This one is about whether the improvement engine on top of it runs.

The company had continuous improvement activity across its sites and no standardised way to track it. Execution was inconsistent and the impact on outcomes was unclear. Not disputed, not argued over. Unclear, which is worse, because an argument at least has two positions in it.

Underneath that sat the thing that makes this hard. Site level variation in improvement tools and behaviours meant there was no shared view. One site ran improvement one way, another ran it differently, and both believed theirs worked. Neither belief was testable, so transparency was limited, alignment was slow, and identifying where the gaps were, or sharing what was genuinely working, was close to impossible.

The word everyone reached for was culture. It is a comfortable word precisely because it explains an outcome without committing to a cause, and it cannot be acted on.

Measuring a culture is a specific technical problem, and most organisations have no instrument for it

It is not that nobody wanted to measure it.

Count the improvement ideas submitted and you measure enthusiasm, or a submission target, not improvement. Count the ideas implemented and you measure the approvals process. Ask people whether they feel empowered to improve things and you get an engagement survey, which tells you the mood and not the mechanism.

The thing you actually want to know is where an idea dies. Whether it dies at the point of being raised, or in a queue, or at the resourcing decision, or after implementation when nobody follows up. Each of those is a different failure with a different fix, and a single satisfaction score conceals all of them equally.

Building something that separates them means deciding what indicators stand for what, testing them on real sites, and accepting that some will turn out to measure nothing. That is a build rather than an analysis, and it is the sort of build a line organisation cannot easily justify starting, because the payoff is a measurement rather than a result.

There is a second reason, and it is the one from part one wearing different clothes. The people who know why an idea died at their site are the people running that site, and they are running it.

Three outcomes, and a narrower honest reading of all three

All three are stated flatly, as things that happened.

An eight to ten week sprint that builds a measurement instrument and pilots it on four sites has produced the instrument and the baseline. It has not produced a year of readings. Improvements in conversion rates and in how fast bottlenecks surface are things you observe over a period of operating, not at the moment a tool is handed over.

Part one of this series made that argument at length, so it will not be laboured again. The short version: read these as what the instrument was built to reveal and enable, and ask us for the second year’s numbers, because that is when they become real.

The senior decided what the instrument could distinguish. The juniors found out whether it did.

What the senior decided. Which six areas the indicators group into, which is the decision that makes the tool useful or useless, because it determines what the measurement can tell apart. Which indicators to discard, and there are always some that correlate with everything and separate nothing. And what the tool must not attempt to measure, which is the discipline that keeps an instrument from sliding back into being a questionnaire.

What the juniors did. More than thirty indicators defined, tested and scored across four pilot sites. Fifteen or more stakeholder interviews. The unglamorous work of checking whether an indicator that sounds meaningful actually varies between a site that improves and one that does not. Most of the value of the instrument comes from the ones that were thrown away.

What stayed behind. Two measurement tools, the six KPI areas, and the indicator set with its reasoning attached. The client can rerun it next year and the year after, which is the only way this particular kind of work pays.

Three things, and the second is a disagreement inside our own materials

Whether the three reported outcomes have been observed. The instrument and the baseline exist. A year of operating with them does not yet, at least not in anything we have seen.

Our own record and the published page do not agree. The page reports 20% higher idea to execution conversion, 30% faster bottleneck identification and 3x more visibility. Our internal engagement record instead reports more than 20 organisational cultural traits identified, two measurement tools built, and 2x visibility potential into engagement drivers. Those are not the same claims. The visibility figure differs in both the number and the hedge. We have used the published page throughout, because a published page is the source a reader can check, and we are reconciling the two.

Whether four pilot sites represent the estate. Four is enough to test whether an instrument discriminates. It is not enough to characterise a company of this size, and the case page does not claim it is.

Where every figure above comes from

From the published case study page. The 20% higher idea to execution conversion, the 30% faster identification of bottlenecks and the 3x more visibility into engagement drivers, quoted with the page’s own wording and questioned above rather than restated as fact. The client’s Vice President of Operations described the result as a structured way to measure improvement culture and translate it into actionable insights for managers.

From our internal engagement record. The €14 billion revenue scale, the four pilot sites, the fifteen or more stakeholders interviewed, the more than thirty cultural indicators, the six KPI areas and the two measurement tools built.

Deliberately not used. The second quote on the case page, which carries no name and no job title. An unattributed quote written in a client’s voice is not evidence and we will not lean on one. And nothing from the capacity block of our internal template, which is boilerplate printed on most pages regardless of engagement.

Clients are described by revenue scale and sector, never named.