# Turnarounds *What it actually takes* --- I've spent a good part of my career as MD of businesses that had stopped working - some I bought, some I was handed. These are the things that mattered most. Details are disguised and numbers rounded. ## Know your window The window is usually set by cash, but not always. Software businesses die slowly and can stay profitable for years while they do, so the clock there gets set by other things: an owner's patience, an exit horizon, a big customer renewal. How long you have decides how drastic you need to be. ## Find the problem yourself Have people walk you through the products and the customer base, not the strategy. Sit through the product as a buyer would, and ask each person separately what they'd fix if the business were theirs. The answers won't match, and the gaps between them are most of the diagnosis. In one business, a week of this showed us the central problem before any analysis did: products at a few hundred pounds a year were being sold and managed alongside systems at tens of thousands, as if they were the same thing. Everyone could see it, but nobody had felt it was theirs to call. That last part is the usual shape of the problem. The people close to a broken thing can mostly see it; they don't feel empowered to reopen it. They weren't in the room when it was decided, or the person who owned it has gone, or raising it now would read as second-guessing a colleague. So it carries on under its own momentum, and everyone's private reservations stay private. Which is also why your first months are worth more than any months after. You're the only person in the building with no stake in the previous decisions, so you're the only one who can re-run one without it reading as criticism. That licence expires: within a few months you'll have positions and allies of your own, and you'll be part of the momentum. Spend it while you have it, on the two or three decisions where nobody can give you the reasoning, only the conclusion. And keep checking the diagnosis against the ground, because I've got this wrong too: I once treated a shrinking business as a sales and marketing problem for eighteen months, partly because the people closest to the product believed in it, and partly because a sales problem was the diagnosis I preferred, quicker and cheaper to fix. An afternoon in the product settled it. ## Change the structure before you judge the people When results disappoint, everyone starts with the people. The problem is usually the structure they're working in ([[Systems first]]). In one situation we split a management team, inside the first two weeks, that had been stretched across businesses with nothing in common, and the people who took what looked like narrower jobs were mostly relieved. There was a cost: one side, seen on its own, was short of management, and filling that gap became the next problem. Give the reorganised teams time and space, too, because the productivity that comes back can surprise you. One team we disbanded outright, because years of steady customer losses said it wasn't working; another that looked just as unprofitable we refocused instead, because underneath the numbers it was propping up the development team and giving away days baked into old contract pricing. Repositioned, and protected from work nobody was paying for, it swung to a couple of hundred thousand pounds of annual contribution. The two looked the same in the monthly pack. ## Pick the two or three decisions that matter Most of what moved these businesses came from two or three decisions, and half the job was refusing to spend energy anywhere else. In practice that meant picking one competitor to take on rather than the whole market, so the story and the reference case only had to work once before we could carry them elsewhere. We stopped the marketing, events and retention spend in a declining business, on the bet that nothing would move: sales didn't move, churn didn't either, and the money went where it made a difference. We simplified six products into two with clear price points. And a team pointed at one business, with its own plan, gets more done than the same people spread across three. The other half of focus is the stuff already in flight. A struggling business is rarely short of initiatives; it's carrying too many, and most of them are stuck behind decisions nobody has made. Before you start anything, count what's already started. Ask each team to list every active commitment and the number that comes back will be a multiple of what anyone guessed, which on its own explains why everything feels late: a queue that long means everything in it waits, however hard people work. Most of what you park, nobody misses - and the few things with a loud sponsor come back quickly enough to prove they mattered. Making the work visible doesn't need a system. In one business we weren't getting paid quickly enough, and part of the reason was that the paperwork the customer needed to pay us against simply wasn't getting finished: it felt like an end-of-the-day job, and nobody had connected it to the cash cycle. We put a row of physical boxes in the office and moved each job's paperwork through them, one box per stage - not started, being worked on, ready for checking, ready to submit. Time to submission went from over ninety days to under thirty. Nothing about the work itself changed; it just became visible, in boxes everyone walked past instead of piles on people's desks. ## Put the plan in the budget A plan only becomes real when it's in numbers people are held to, and the budget is where that happens. Expect to send the first cut back: it will arrive with everything still going backwards, and the work is in the assumptions - when development lands, when a deployment becomes something you can sell, what a missed window costs. When we modelled a cross-sell wedge into an adjacent customer base, we shaped it as a five-year adoption curve rather than the straight ramp the first draft had, segment by segment, some segments assumed to adopt at a fraction of the rate of others. In a business that sells on an annual cycle, a missed implementation window doesn't cost a month, it costs a year, and you can't sequence a plan without knowing it. What goes back is assumptions, not totals. Where a business's future is in doubt, build the case, don't assert it. For one that was being written off, we costed the technology rebuild to know how long the customer base needed to hold, put renewal probabilities against the customers and looked at the range of outcomes rather than the average, and separated the spend that protects existing customers from the spend that has to earn its case. It had been keeping customers on one-year contracts because nobody would commit to its future; when we committed, they did too. Two years on it makes more profit than it did when it was being written off. ## Set the rhythm, then hold your nerve Keep the machinery simple: short weekly sessions that actually decide things, a written update at the end of the week, the numbers and the product plan reviewed separately so each gets proper time, and a plan on a page for each business - the long-term goal, this year's milestones, the projects that get there, the metrics that prove it ([[Instrumentation]]). Holding your nerve is harder, because the right moves routinely make the reported numbers worse before they make them better, and the pressure to reverse them is strongest exactly then. Exit unprofitable contracts and revenue falls; move engineers onto root causes and the roadmap slows; tell your biggest customers the truth about timelines and the satisfaction scores dip. Every one of those is the right call, and every one looks identical to a new leader making things worse. What gets you through it honestly is knowing the difference between a system correcting and a plan failing, because both look the same in the board pack. The headline numbers are lagging: they're still reporting the consequences of decisions taken months ago. The signals that answer the question sit upstream, closer to the work - the defect rate, retention in the customers you kept, how long support tickets take to resolve. When the plan is working, those turn first while the headlines keep falling, and the gap between the two turns is where you earn your money. When they all fall together, the plan isn't working, and the honest move is to look again rather than hold harder. Name the upstream signals when you build the case, not after the numbers dip; if the ones you named are moving, hold, and if they aren't, look again. And be slow to declare victory: one of the businesses above still isn't finished, and I'm not counting it yet. ---