Write down the label you want, then work backwards. That is the standard instruction for building a target product profile, and as a starting move it is fine. You do need a target, because you cannot design a programme around a shrug, and "we'll see what the data says" is not a plan anyone funds.
But the advice quietly does a second thing it never owns up to. It tells you the TPP is the label you are aiming at. And the moment a document becomes the label you want, you acquire a stake in defending it. You keep it safe and consistent, and you stop going looking for the data that would prove it wrong. That is how the TPP ends up in a drawer, unopened on the day the decision it was written to inform finally lands on the table.
Call it the read-only TPP. You can open it, admire it, drop it into the board deck. You cannot write back to it. A TPP earns its keep only when it works the other way round: as a testable, revisable hypothesis you fully expect the data to change, rather than a first draft of a label you are already committed to reaching. A label is a thing you defend. A hypothesis is a thing you test.
Most teams do have a TPP: IQVIA reports that 83% of surveyed companies hold one for their assets [1], so the document almost always exists. What almost never happens is that anyone opens it when a real decision is on the table.
Look at where TPPs actually surface. In an analysis of FDA Summary Basis of Approval documents, Tyndall and colleagues found that just 91 of 2,138 NDAs and BLAs approved between 1999 and 2015 mentioned a labelled or quantitative TPP at all roughly one in twenty-three [2]. And where they did appear, they turned up late: TPPs were "most frequently introduced into the regulatory dialogue at a late stage of the process, usually at the time of the pre-NDA or BLA meeting or following NDA or BLA submission," and were "notably absent from discussion in meetings before submission and meetings at the end of phases I and II" [2].
That is a hard fact to face on a Monday. If your TPP first meets a regulator at the pre-NDA meeting, it never shaped a single trial it was written to inform. The Phase 2 that read out and the dose you carried into it were both settled while the document sat closed.
The timing data makes it sharper still. Wang and colleagues surveyed eleven companies: three had started drafting the TPP at the pre-clinical stage and five during Phase I [3]. So teams write it early, and yet of the companies that began pre-clinical, none built in an HTA or payer perspective until Phase I was already underway [3]. Drafted at the start, then left to age while the outside perspectives that should have stress-tested it arrived a phase too late. The document exists from day one and works as a decision tool at the end, if ever.
That gap is the whole problem.
Free download
The TPP Template
Target and minimally-acceptable profiles side by side, every claim linked to the evidence that has to support it.
Get the template →Let's be honest about what a TPP is on most programmes. It is an aspirational label, written months or years before there is any clinical data to justify one. The instant you frame it that way, as "what we want the label to say," you have committed to a claim you now have every incentive to protect, the same instinct that keeps an outdated biotech playbook running long after the data has moved on.
Watch what happens when the data complicates the claim. A read-only TPP has exactly two moves available: ignore the disappointing result and press on, or quietly soften the number, tell no one, and carry on as though nothing changed. Both corrode the programme from the inside. A read-write TPP has a third move the other two lack: revise the claim openly, record why, and let that revision drive the next decision.
I have watched the first failure type up close. Years ago I worked on an early-phase programme in a small indication where the TPP promised a treatment effect the emerging Phase 1 data was already, quietly, arguing against. The number had been set before the first patient was dosed, and somewhere along the way it had become the thing the programme was about. So when two cohorts came in soft, the instinct in the room was not to ask whether the target was still the right target. It was to find reasons the next cohort would rescue it. We spent those cohorts defending a number instead of interrogating it. Nobody had written down, in advance, what result would tell us the number was wrong. So no result ever did.
Here is the part teams get backwards. The regulators are more relaxed about revision than the sponsors are. FDA issued its TPP guidance as a draft back in 2007, and the Federal Register notice announcing it (docket FDA-2007-D-0256) is explicit that submitting a TPP is voluntary. "The TPP does not represent an implicit or explicit obligation on the sponsor's part to pursue all stated goals," the notice states, and it "does not constrain the sponsor to submit draft labeling ... that is identical to the TPP" [4]. That guidance was still draft when a 2017 renewal notice confirmed it, and no finalisation notice has surfaced since, but the framing has never wavered: the agency treats the TPP as revisable. The MHRA is more emphatic still, and its live ILAP Target Development Profile guidance, updated on 19 May 2026, states plainly: "We expect this roadmap to be a living document, updated along the development programme timelines and milestones as new knowledge is generated" [5]. The people you imagine you are protecting the number for are the ones telling you to change it.
The reframe costs nothing: it changes no template and adds no headcount. What it changes is what the document is allowed to do when the data arrives, and that is a shift of stance before it is a shift of format.
A TPP you can revise whenever you like is worthless. It becomes a temptation to "move the goal posts" : a document that retroactively rationalises whatever data you happened to get, so no readout is ever a miss and no target is ever failed. A fixed number, at least, holds the team to account. Make it endlessly revisable and you have built an instrument that cannot embarrass you, which is another way of saying an instrument that cannot inform you.
That objection is right about the risk and wrong about the cure. What separates a living hypothesis from goalpost-moving is a single discipline: pre-commitment. Revising a target because you crossed a threshold you named in advance is science. Relaxing a target after a bad readout, with no threshold ever written down, is exactly the failure the objection fears. And notice which kind of TPP invites that failure. The aspirational single-value label does, precisely because it never stated what result would count as falsification in the first place.
None of this needs a new document; it needs four changes to the one already on your desk.
1. Add a "what-would-change-my-mind" column. Beside every target attribute, write the specific result that would revise it and the decision that revision gates. Attribute, then revision trigger, then gated decision: if confirmed Phase 2 ORR in this population comes in below X%, we do not advance it into the registration-enabling trial. This is not exotic advice: a convergent body of work in early development has argued for years that these criteria should be fixed a priori: Wessels and colleagues on go/no-go criteria in Alzheimer's [7], Smoragiewicz and colleagues on biomarker-linked decision rules [8], Bratton on quantitative decision thresholds [9]. Yet Jiang and colleagues, reviewing how go/no-go calls are actually made, found they "have often been made based on evidence and knowledge available at the decision points, often without significant prior planning" [10]. The discipline is well understood; it is just rarely written down.
2. Give each attribute three values, not one. A minimum, a base case, and an aspiration. The single aspirational number is the label-in-waiting; the range is the hypothesis, because it already encodes where the line falls between a result you advance on and one you do not. Many teams still are not: IQVIA reports that while 83% hold a TPP, only 68% explore multiple TPPs per asset [1]. The minimum viable evidence framework is the same instinct applied to the whole evidence base: decide the floor before you fall through it.
3. Anchor revision to readouts, not the calendar. A simple prescription is to lock the TPP, before a Phase 2 readout to inform piv go. Tie each revision point to a value inflection, an end-of-phase meeting or a key readout, rather than to an annual diary review that lands whenever it lands. The question is never "is it that time of year again." It is "have we just learned something the document should answer to."
4. Keep a revision log. Record every change to an attribute with the data that drove it and the decision it changed. This is the piece that turns "revisable" into "accountable," and it is the direct answer to the goalpost-moving objection: a logged revision is a defensible one, an unlogged one is a quiet climbdown. It also happens to make the document legible to a regulator who, per FDA's own framing, already expects it to change.
Here is the test, and you can run it today. Point at your TPP and name one decision it changed in the last six months. If you can, it is a working hypothesis and it is doing its job. If you cannot, it is read-only: a photograph you took on day one and have been navigating by ever since, as though it were a live feed.
If I could put one thing into every TPP template, it would not be a better attribute taxonomy or a slicker layout. It would be the two columns this whole post keeps circling back to: the trigger that would revise each target, and the log of every time one did. Everything else is formatting.
Building the TPP in the first place is a separate craft, and we've covered it without the guesswork already. This piece only picks up from there, the authoring stance that decides whether the finished document drives a single decision, or just sits in the drawer being admired.
Get the monthly digest
The 5 things evidence leads need to know each month: regulatory moves, RWE developments and what they mean in practice. No pitch, one email a month.