The culture diagnostic, which is not a survey.

We will not run an engagement survey for you, and this is the page where we explain why at length rather than in a footnote. A survey is an instrument for a population you cannot speak to individually. At forty to four hundred people you can speak to everybody who matters in six weeks, which is faster than most survey vendors take to return a dashboard. What the survey adds is a number that lets a leadership team discuss the problem without naming anyone. That comfort is the product. We do not sell it.

What the six weeks contain

Week one — documents before people

We start on paper, because arriving with opinions already formed by interviews is how consultants get captured by whoever talks first. We read the last two appraisal cycles in full: ratings, comment boxes, the overturns, the moderation notes if any survive. We read every exit interview form from the last twelve months and, more usefully, we ask for the resignation emails themselves. We pull the org chart as it exists in Darwinbox or Keka or greytHR and compare it to the org chart the founder draws on a whiteboard when asked. Those two are never the same, and the difference is the first finding roughly half the time.

Weeks two and three — the leavers

We call people who resigned in the last six months. Not all of them agree; usually six or seven in ten do, which is more than enough. They have no stake left, and after two years in a WhatsApp group with their old team they know precisely what is still broken. We read those transcripts together in one sitting with names removed. Patterns surface in about twenty minutes and they almost never point at "culture". They point at three or four specific managers, one broken handover between sales and delivery, or a promotion decision in the last April cycle that everybody watched.

Weeks three to five — the people still here

Thirty to forty conversations, forty-five minutes each, on your premises. We deliberately over-sample two groups: people with eight to eighteen months' tenure, who can still see the company clearly and have not yet stopped mentioning it; and first-time managers, who carry the most load and get the least support. We sit in on real one-on-ones with permission — typically eight to twelve of them — and in at least one skip-level and one weekly review. Watching a manager run a review teaches us more than asking a manager to describe how they run a review.

Week six — findings, argued out loud

Four findings. Not twelve. Each one is written as a sentence a line manager would recognise, with the evidence beneath it, the people involved named, and a proposed owner. We read them out to the founder and the leadership team in a two-hour session and we expect to be argued with; findings that survive that argument are the ones worth acting on. Anything that does not survive gets struck out in the room and does not appear in the written record.

What you get, physically

• A findings memo of eight to twelve pages. Prose, not slides. No maturity model, no radar chart, no red-amber-green grid.
• An anonymised thematic read of every exit conversation from the last year, which most clients have never had in one document.
• A manager-by-manager picture, shared verbally with the founder and CHRO only, never circulated.
• A one-page sequence of what to do in the next ninety days, with names and dates against each item.
• All raw interview notes destroyed at handover. We say this to every interviewee at the start, and we keep it.

Patterns show up in about twenty minutes. Almost always they point at three or four managers, not at a culture.

When this is the wrong engagement

If your leadership team has already decided what the answer is and wants evidence for it, this will be an expensive six weeks and an unpleasant last meeting. If a co-founder relationship is the actual problem, a diagnostic will find that in week two and there is nothing we can do with the finding except say it. And if you are under about thirty-five people, the founder already knows all four findings and is avoiding one of them; a diagnostic buys permission, not information, and there are cheaper ways to buy permission.