You Can't Run a Movie from a Single Frame: Why Static Plans Fail in Dynamic Industries

Most plans are obsolete the day they are bound, not from bad work but because the world will not hold still. The edge goes to the plan that moves as fast as the work.

Andy Mehrotra

Chief Executive Officer

T-36 months. The latest data is good, better than good, and for the first time the team lets itself say the words out loud: first-in-class. The mood in the room is the best it has been in a year.

The quarterly plan comes back from the launch consultancy that same week. 300 pages, elegantly bound, one copy for every seat at the table, heavy in a way that feels like proof. The figures have been checked twice, the dates line up, the resource curves run smooth from this quarter into the next. By any fair standard it is a good plan, and for about a day it describes the program everyone is so sure of.

Then the world does the one thing it can be relied on to do. A site that promised first patient in by March slips to May. A competitor pulls a readout forward. A supplier flags a delay it has sat on for a month. None of it makes the plan look wrong to anyone turning the pages. The binding still holds, the figures still add up. The plan reads exactly as true as it did the day it was bound. It has simply started describing a program that no longer exists.

No one reprints it. Regenerating 300 pages by hand is too high a price for a slip that feels small, so the book stays on the table and the organization manages the book. We pour real skill into producing a precise picture of a single moment, then spend the quarter treating the picture as the moving thing it could only ever photograph.

A plan is a photograph, a single frame of a thing that will not hold still. It freezes a moving subject, and the better the photograph, the more convincing the freeze. That is the trap inside competence. A blurry plan invites suspicion and gets questioned. A crisp one earns a trust it has not earned, because crispness measures the camera, not the stillness of whatever sits in front of it.

A plan's accuracy decays from the moment of signing. The only real questions are how fast, and whether anyone is watching the decay.

Few organizations can answer either one, because almost nothing in the planning machine is built to watch. It is built to produce the picture, frame it, and move to the next.

A Still of a Moving Subject

The most over-defined number in any plan is the date. First patient in, last patient out, database lock, submission, approval. Each is printed as a single day, a hard edge, a promise with a number attached. And each describes something that was moving the whole time it was being set down.

The proof is in how often the picture has to be retaken. Across the industry, the share of trial protocols that undergo at least one substantial amendment has climbed to 76%, up from 57% in 2015, and the average protocol now absorbs 3.3 of them, an increase of 60% over the same period. Each amendment is evidence that the study has outgrown an earlier version of its own plan. Some are forced by regulators, some by a change in strategy, some by what the data demanded. The cause varies. The pattern does not. The plan a team manages toward is rewritten again and again before the study ends, each revision replacing one the work had already outpaced.

Part of the reason a date hardens into a promise is that the systems holding it demand a single answer. The form has one box. It does not accept a range, a confidence level, or a note that reads ask again in three weeks. So the most likely date goes in the box, the qualifiers fall away, and a probability collapses into a fact the moment it is typed. Everyone downstream reads the fact and not the probability behind it, and begins making commitments against a precision that was never there.

This is the old difference between the map and the territory, worn smooth from use but still true. The map earns its keep precisely because it is smaller and stiller than the ground it stands for. The danger begins when the stillness starts to feel like accuracy, when a team defends the date on the page against the work in the lab because the page is legible and the work is not. A single-point date was never a fact. It was a range wearing the costume of one, and everyone in the room knew it on the day they signed.

The Number at the Center of the Frame

Pull back from the date and a larger, more expensive version of the same flaw comes into view. The number the whole plan exists to justify is the forecast, and the forecast may be the least reliable figure in the model, and the one with the longest reach.

The most thorough examination of this phenomenon analyzed more than 1,700 sell-side analyst forecasts across 260 launched drugs. Most consensus estimates were wrong, and not at the margins. A year before launch they missed actual peak sales by around 70%. Six years after launch, with real prescriptions and real revenue stacking up to correct them, they were still off by close to half. The number does not converge on the truth even when the truth is sitting in the sales ledger waiting to be read.

The misses are not abstract, and they are not only a matter of hindsight. Take the most watched market in the industry today, the GLP-1 obesity drugs. Working from the same public data, credible analysts put its value by the mid-2030s anywhere from roughly $95 billion to $190 billion. The gap between those two forecasts is the size of a major industry on its own. That is the number programs are being funded, staffed, and sequenced against right now, and the people paid to forecast it cannot agree within a factor of two. A snapshot assumes a subject that will hold still long enough to be captured. This one is sprinting.

Internal long-range forecasts are drawn up by different hands than the analysts', but they inherit the same trouble: they are estimates presented as calculations, the same gap between an estimate and a calculation that runs through the rest of development, here sitting at the center of the picture. And the error does not stay there. The plan funds, staffs, and sequences against that number, so a forecast wrong by half does not misjudge one asset, it bends the choices around it: pitched too high, it keeps a fading program resourced past the point of sense; pitched too low, it starves one that deserved more, and the company learns its mistake from a competitor's launch rather than its own model. The precision is real. The accuracy is a hope. The plan prints them in the same font.

Why We Keep Developing the Same Negative

If the plan is so reliably out of date, why does the organization keep developing the same negative every quarter and hang it on the wall as though it were news?

You would not plan your day around last month's weather forecast; you check the sky on your way out. Yet a program runs a full quarter off a read of the world taken at budget time, then acts surprised by a storm that was on the calendar all along, still following a plan, though not the one today would write.

The plan is wired to the wrong clock. It refreshes on the fiscal calendar while the work runs on its own. The budget cycle decides when the picture is retaken, and the budget cycle answers to the financial year, an interval with no relationship to when a program changes. A competitor does not read out on your planning schedule. A site does not enroll to your fiscal quarter. The plan and the events it tracks keep different time, and only one of them is allowed to call the meeting.

There is a deeper reason underneath the calendar, named long ago outside this industry: the annual budget behaves as a fixed performance contract, a promise locked in once a year and then used to measure the people who signed it. Reopening a contract feels like a renegotiation rather than an update, so once a plan is signed, changing it carries the weight of conceding the first version was wrong. The safe move is to defend the signed number and absorb the variance quietly. What was meant as a working hypothesis becomes a position to defend.

A planning lead at a large organization put it to me plainly. Everyone in the room knew the date had moved, she said, but the date on the page was the one we were measured against, so that was the date we defended, right up to the quarter it became impossible to defend. The plan had stopped being a tool for seeing the work and become a thing to be protected from it.

So the picture hardens. By the time a long-range plan is approved it already reflects conditions from months earlier, the cost of a cycle that at many organizations runs the better part of half a year. The planning systems were never built to move at the speed of the work, much as the organizations around them were never built as single integrated systems but grew function by function, each adding its own layer of plan on its own cycle. A plan that looks decided is easier to present than one that admits it is still moving, and the room rewards the version that looks decided. None of this is foolish. It is the sensible behavior of capable people working a system that asks them to manufacture certainty on a schedule, then treats the result as the truth until the next scheduled truth replaces it.

From Stills to Footage

The answer is not to take the picture more often. A monthly plan is still a slideshow, a run of stills shown fast enough to hide that nothing between the frames was ever captured. Speeding up a broken cadence buys a fresher photograph and the same blind spots in between.

You cannot run a movie from a single frame. The shift worth making is from the still to the footage, from a document refreshed on a calendar to a position continuously re-derived from the work itself. It changes what each number on the page is allowed to mean.

A date stops being a single day and becomes what it always was, a distribution. In place of first patient in by March, the honest version reports an earliest credible date, a most likely date, and a late-case date, along with the handful of activities driving the spread, and it moves as they move. The date is allowed to admit it is a range, which is the only thing that ever made it worth trusting.

A forecast stops being a fixed entry and becomes a live read, re-derived from current evidence rather than a template filled in once and frozen for the year. The resource picture beneath it follows the same logic, computed from the actual shape of the work as it stands today. And the resource question sharpens. A plan showing twelve funded people in a quarter says nothing about whether the specific few whose skills, availability, and dependencies the next critical milestone depends on will be free when the work reaches them.

And someone, or something, finally watches the decay. Variance against the plan becomes a continuous signal instead of a quarterly autopsy, so the gap between the plan and the program surfaces while there is still room to act on it, not at the review where the slip has already cost what it was going to cost. Tracking that drift by hand, across every program, was once weeks of labor that went stale the moment it was finished, so it was done once for the budget and shelved. That is the constraint that has lifted. Holding a plan as a living position, re-derived in the background as the work moves, is now something a system can sustain continuously, at the resolution where the work happens.

Picture the same March slip in that world. The site does not quietly fall behind until a quarterly review catches it. In week two, enrollment is tracking below the curve the plan assumed, and the date moves on the screen before anyone has to confess it. The program lead sees the slip beside the baseline, and sees what it pulls with it: the database lock that now drifts, the filing that rides on it, the specialist two programs were quietly counting on. The decision that used to wait for the next cycle is on the table while it is still cheap to make. Nothing was reopened. The plan told the truth in time to act on it.

None of this is a call to make drug development agile. You cannot sprint a Phase III, and a program that re-opens its plan every two weeks spends its quality and its credibility fast. The long horizon is right. Phases, milestones, and the dependencies between them belong in a plan built to look years ahead, and that plan can stay the baseline. What has to change is its relationship to reality. Beside the baseline runs a live read of the work: the current critical path, the uncertainty around the dates, the resource conflicts forming underneath, and the triggers that tell leaders when a decision is due. The baseline holds while execution adapts. Plan for the decade. Sense and adjust by the week.

A plan is not a promise to be kept. It is a hypothesis with a decay rate, and the job is not to defend it but to know, at any moment, how far it has drifted from the thing it describes.

The Picture and the People Outside It

Every month a plan describes a program that has already moved is a month the real work runs without a current map, and the bill arrives in two places at once. There is the revenue side, the exclusivity burning off the front of a patent life that was always going to be short, the competitor reaching the same milestone first because their picture sat closer to live than yours. The industry models this side, imperfectly, and at least argues about it in rooms with budgets.

The other cost stands outside the frame. Behind every slipped date is a person waiting on a medicine the science may already be capable of delivering, for whom the distance between a plan that moves with the work and one that lags it by a quarter is measured in months they cannot spare. Once a medicine has earned its way through the science, getting it to those patients on time is no longer only a scientific challenge. It is also a coordination one, and part of that coordination is how honestly the organization tracks its own moving target.

The picture on the table will always be a little out of date. That is what pictures are, and taking one is not the mistake. The mistake is treating it as the program: framing it, defending it, measuring against it, and looking up a quarter later to find the work has walked out of the picture while everyone was admiring the print.