Impact = value / effort - calculating and measuring what matters

Every roadmap meeting has a loudest voice, and in the absence of a better system, the loudest voice is the prioritisation framework. It's nobody's fault exactly - features arrive championed by whoever wants them, the arguments are all plausible, and something has to break the tie - so the tie gets broken by seniority, or by recency (whoever spoke to a customer last), or by sheer stamina. The roadmap that falls out of that process isn't a strategy, it's a negotiation transcript.

The fix we use is old, small and slightly unfashionable: impact equals value divided by effort. Score what a piece of work is worth, score what it costs, divide, and rank by the quotient. The formula fits in a tweet, and the whole craft is in how honestly you fill in the two numbers - which is what this piece is actually about.

Scoring the top half: value

Value is where scoring goes to die, because 'valuable' means something different to everyone at the table. We pin it down by splitting it into three questions that each get a simple 1-to-5 score: how much does this move the metric we've agreed matters (magnitude), how many customers or transactions does it touch (reach), and how sure are we of the first two answers (confidence). Multiply them together and you have a value score that's crude, comparable and surprisingly hard to argue with.

The confidence term is the one teams skip and the one that matters most, because it's where evidence earns its keep. A feature backed by interviews, funnel data and a competitor's public results scores a 4 or 5; a feature backed by 'the CEO mentioned it in the lift' scores a 1 or 2, and suddenly the difference between them is visible on paper rather than simmering in the room. Confidence scoring also creates a productive loophole - if you want your pet feature ranked higher, go gather evidence for it, which is exactly the behaviour a roadmap should reward.

Two disciplines keep the top half honest. Value gets scored against one metric (the quarter's agreed number, not a different metric per feature, or everything scores five out of five on something), and value gets scored by the group in the open, with the argument happening before the number is written down rather than after.

Scoring the bottom half: effort

Effort looks easier because engineering teams estimate all the time, and it hides a subtler trap: effort is not just build time. The honest denominator includes design time, the integration nobody has scoped, the data migration, the QA surface, the support burden after launch and the ongoing cost of owning one more thing - which is why we score effort as a team sport too, with engineering, design and operations all holding a pen.

We keep the scale deliberately coarse (T-shirt sizes converted to 1-to-5, roughly person-weeks on a log curve), because false precision is the enemy here. An estimate of '13.5 days' is a guess wearing a lab coat, and the formula only needs to know whether something is a snack, a meal or a banquet. Coarse scales also speed the meeting up enormously - you can score thirty items in an hour once nobody is pretending to know things to one decimal place.

A '13.5 day' estimate is a guess wearing a lab coat.

What the quotient changes

The first thing the division does is surface a category of work that opinion-ranked roadmaps systematically bury: the small, dull, high-return item. The checkout copy change, the form field removal, the email that recovers abandoned carts - modest value scores sitting on tiny effort scores, producing quotients that tower over the glamorous replatform everyone came to argue about. Ranked honestly, quick wins stop being a slogan and start being arithmetic.

The second thing it changes is the texture of the argument. Disagreements move from 'should we build this' (a status contest) to 'is this really a 4 on reach' (an evidence question), and evidence questions have the enormous social advantage of being answerable without anyone losing face. The meeting gets shorter, the quiet people get a mechanism through which their knowledge actually lands, and the final ranking arrives with its reasoning attached - which means it can survive contact with the stakeholder who wasn't in the room.

The third change is accountability with a memory. Scores are predictions, and predictions can be marked. When a shipped feature's actual impact comes in far below its scored value, the retro question isn't 'who do we blame' but 'which score was wrong and why' - and over a few quarters those post-mortems quietly calibrate the team, the way a golfer's scorecard calibrates a swing. Teams that never write predictions down never find out they're bad at predicting.

The traps, because there are always traps

• Precision theatre - decimal-point scores and weighted sub-criteria that make the spreadsheet feel scientific while the inputs are still vibes. Keep it coarse and keep it honest.
• Effort gaming - champions quietly shrinking the denominator on their favourite. The defence is scoring effort with the people who'll do the work, in the same room.
• Strategy blindness - the formula ranks the backlog you gave it, and some low-quotient work (paying down the platform, the compliance deadline) must ship anyway. The score informs the roadmap. It is not the roadmap.
• One-and-done scoring - a ranking from January is a fossil by June. Re-score quarterly, briefly, as the evidence and the metric move.

None of those traps is a flaw in the formula so much as a reminder of what it's for. Impact over effort doesn't think for you, it just forces the thinking into the open, one honest number at a time - and in our experience the roadmap that comes out the other side isn't only better ranked, it's better defended, because for once the answer to 'why is this first' is something other than the loudest voice in the room.