Somewhere in your organization there is a rule that says work below a certain size does not need a business case. It might be fifty thousand dollars, or twenty days of effort, or a change that touches fewer than two systems. Whatever the number, its purpose was sensible: stop the portfolio board spending forty minutes approving a fortnight of work, and let small things move without ceremony.
Now count what has accumulated underneath that line. Not the cost of it, which is small by construction, but the hours of your scarcest specialists that it consumes in a year. In most portfolios I have looked at, the answer is somewhere between a quarter and a half of the constrained resource, spent on work that has never once been compared with anything else, by a process that was explicitly designed not to compare it.
That is the effect, and it needs a name, because an unnamed drain gets absorbed rather than managed. Call it the sub-threshold portfolio: the substantial body of work an organization runs without knowing it, made invisible by a governance rule that filters on the wrong quantity.
The rule is not badly set. It is denominated in the wrong currency. Thresholds are written in dollars, or days, or headcount, because those are what a finance function can audit. The thing that is actually scarce is hours of one specific overloaded resource, and nothing about being small in dollars makes a piece of work small in constraint-hours. As it turns out, the relationship runs the other way.
Small work is the most constraint-dense work you own
Ask why a request is small. Almost always it is small because everything expensive has already been stripped out of it. There is no procurement, no supplier, no training, no communications plan, no data migration, no project manager, no test campaign, no benefits case. What is left, the entire remaining content of the request, is a decision by someone senior, a design opinion, and a review.
Which is to say: what is left is the constraint.
Large initiatives are mostly not constraint work. A program with sixteen hundred person-days in it has an enormous body of activity that your architects and principal engineers never touch, and their involvement, though decisive, is a thin seam running through a very large rock. Small requests have no rock. They are seam.
So the practical relationship is the opposite of the intuitive one. Per dollar of effort, sub-threshold work draws several times more of the constrained resource than the governed portfolio does. Your threshold rule, read honestly, says this: work that is cheap in the currency we can measure may bypass the ranking, regardless of its price in the currency that limits us. And because that rule is applied hundreds of times a year,, by people acting entirely properly, it moves a large fraction of your delivery capacity outside the reach of any prioritization you do.
This is not an argument against small work, and it is emphatically not an argument against smaller projects. Cutting a ranked initiative into small increments is one of the best things you can do to a portfolio, because it shortens the distance between spending and earning. The sub-threshold portfolio is a different animal. Small batches are decided work delivered in slices. Sub-threshold requests are undecided work delivered because nobody had to decide. The first stays inside your ranking. The second was defined, at the moment the threshold was written, as work your ranking is not allowed to see.
What the exemption actually grants
Being below the line is not a lighter form of governance. Functionally it is a promotion, and it comes with four properties nobody intended to grant.
It is sequenced by relationship, not value. Above the line, work waits its turn against a rank order. Below it there is no rank order, so the queue resolves itself the way ungoverned queues always do: by who asked, how loudly, how recently, and how awkward they will be about it. Your organization therefore operates a fast lane whose entry criterion is not value, urgency or return. It is smallness. The one lane with no value test is also the lane that jumps the queue, which is the same economics as the expedite tax, applied a few hundred times a year and never called by that name.
It arrives in fragments. A governed initiative books constraint time in blocks. A small request arrives as an interruption, then another one four days later when the answer prompts a question, then a third at review. Each one lands on someone holding an hour of context in their head. This is the cost of multitasking in its purest form, and it means the nominal effort figure understates the true draw by a wide margin. Three and a half hours of work spread across five interruptions is not three and a half hours of cost. The shattered afternoons are the cost, and they land on the phantom capacity your plans have already spent.
It teaches the organization how to avoid you. A threshold is not only a filter, it is a published specification for how to get past the filter. Once people learn that fifty thousand dollars is the line, work arrives at forty-eight. A program that would not survive a ranked comparison is delivered as three unrelated small changes over eight months, each individually defensible, collectively a project that was never approved. Nobody experiences this as gaming. They experience it as getting things done, which is what you asked of them. The stricter the gate above the line, the harder the pressure below it, so tightening governance on large work reliably grows the sub-threshold portfolio.
It never closes. A project ends. A small change to a live system is permanent: it adds a fixture that must be regression tested, upgraded, patched, explained to the next person, and worked around by the next change. Every request below the line therefore leaves a tail of future demand, and that tail lands on the same specialists, because they are the only people who understand what was done. The sub-threshold portfolio is the only part of your portfolio that manufactures its own future workload. Left alone for a few years it does not settle at a level, it grows, which is why the specialists who tell you they have less time for real work every year are describing something arithmetic rather than a feeling.
A worked example
Numbers make the shape concrete. These are illustrative, and the arithmetic is the point.
A technology function of about two hundred people funnels its significant decisions through five principal engineers and architects. They are the constraint: nothing meaningful ships without their design input and their review. Each has about 28 genuinely productive hours a week once meetings, incidents and administration are removed, across 46 working weeks, so the constraint offers roughly 6,400 constraint-hours a year. That number, not headcount and not budget, is the delivery capacity of the organization.
The governed portfolio is 14 initiatives. Together they consume about 4,050 constraint-hours, roughly 290 each, and they return about $1,100 per constrained hour, which is the portfolio's value per constrained resource hour and the number every ranking decision is really made against.
Below the line, the service queue took 420 requests last year, about 35 a month. Average size, six days of total effort, comfortably under a twenty day threshold. Direct constraint time per request, 3.5 hours: a scoping conversation, a design call, a review, a sign-off. That is 1,470 hours of direct draw. Apply a conservative fragmentation factor of 1.6 for the interruption cost of work that arrives in pieces, and the real figure is about 2,350 constraint-hours.
Set the two side by side. The sub-threshold work is about a tenth of the organization's delivery effort and about 37 per cent of its constraint capacity. Per person-day of effort it draws roughly five times as much of the scarce resource as the governed portfolio does. More than a third of the thing that sets your delivery speed is committed by a route that exists specifically so that nobody has to think about it.
Now the part that decides what to do about it. Rank those 420 requests after the fact, on value delivered against constraint-hours consumed, and they do not form a blob. They form two populations.
- The best 90 of them consumed about 540 constraint-hours and delivered something like $1.6m: a rate of nearly $3,000 per constrained hour. That is not merely acceptable, it is the highest returning work in the building, comfortably ahead of the governed portfolio, and every hour of it was well spent.
- The remaining 330 consumed about 1,815 constraint-hours, more than a quarter of the constraint's year, and delivered around $500,000: a rate of roughly $275 per constrained hour, against a portfolio that pays $1,100 elsewhere.
Stop the lowest returning half of that second population, about 165 requests, and you forgo perhaps $250,000 of value while releasing roughly 900 constraint-hours. Redeploy those hours onto ranked work at $1,100 and they return about $990,000. Net, about $740,000 a year, from a decision that involves no hiring, no tooling and no reorganisation. For context, the fully loaded cost of all five constrained specialists is around $600,000. The throughput sitting inside the sub-threshold portfolio is worth more than the salary bill of the team it is taken from.
Notice also what the two populations mean together. The ungoverned lane contains both the best work in your portfolio and the worst, mixed indistinguishably, and the only thing that separated them was somebody looking afterwards. Abolishing the lane would have destroyed $1.6m of exceptional return. Leaving it alone costs three quarters of a million a year. The answer is neither.
Why single-project management is blind to this
Every instrument the organization uses is structurally incapable of showing this.
The portfolio report shows the 14 governed initiatives, because a portfolio report is a list of projects and none of this is a project. There is no artefact anywhere in the organization on which the 420 requests appear as a single line, and no owner whose job it is to hold that line. The largest single consumer of your constraint has no name, no sponsor, no budget and no reporting.
The finance view is worse than silent, it is actively reassuring. The sub-threshold work is cheap. Each item is under the line, in aggregate it is a tenth of the effort, and it is often absorbed into run costs rather than change budget, so the one function that could plausibly notice the aggregate is looking at the number that makes it look trivial. Cost is precisely the dimension on which this work is innocent.
And the loss lands nowhere in particular. The 900 hours are not missing from any one initiative. They are missing from all fourteen, a fortnight here, three weeks there, each shortfall small enough to be explained locally by a supplier, a specification change, or a delivery manager who needs to grip things harder. Fourteen plausible local explanations, and nobody sums the column. It is the same blindness that hides the price of a new project: a cost that falls on shared capacity is a cost no single-project view is built to see.
Underneath it all sits the Project Illusion in one of its most expensive forms. The organization believes its portfolio is the set of things it has approved. Its portfolio is actually the set of things that consume its constraint, and those two sets have diverged by more than a third without anybody deciding that they should.
What to do instead
1. Measure the aggregate once, in constraint-hours. You cannot manage this until it exists as a number. Take one quarter of the service queue, count the items, and get an honest estimate of constrained time per item including the interruptions, from the specialists rather than from the ticketing system. Multiply up. Then present it as a single line next to your governed initiatives: sub-threshold work, 2,350 constraint-hours, 37 per cent of capacity, no owner. That one line changes more behavior than any policy that follows it, because until it is written down everyone in the room is discussing a portfolio that does not exist.
2. Rewrite the threshold in the currency that binds you. The test should not be "under fifty thousand dollars". It should be "under four hours of constrained specialist time". Those two rules admit almost entirely different work, and only the second is measuring the thing that limits you. Items that are cheap in money but heavy on your architects are exactly what needs ranking, and today they are exactly what is exempt.
3. Give the lane a capacity, not a queue. Ungoverned demand expands to fill whatever it is given, so give it a bounded allocation instead of an open door: a fixed share of constraint capacity per period, say fifteen per cent, held deliberately and reviewed quarterly. Inside the allocation, keep the speed and the informality, because that is where the $3,000 an hour lives. Once the allocation is spent, the next request waits or gets ranked. This is starting less to finish more applied to the part of the portfolio that no work-in-progress limit currently touches, and it works for the same reason: a limit forces the comparison that abundance lets you avoid.
4. Rank within the lane, cheaply. The objection to governing small work is that a business case costs more than the work does, and that is true. It is an argument for a proportionate comparison, not for no comparison. Two questions, answerable in a minute, will separate your $3,000 an hour from your $275: what is this worth if it happens, and how much constrained time does it take? A single ordered list, refreshed weekly, with the specialist time estimated rather than measured, captures most of the available value. Precision is not what is missing here. Comparison is.
5. Route work off the constraint before you route it into a queue. A large share of sub-threshold requests reach a principal engineer because that is the habit, not because the judgment is genuinely scarce. Publish the decisions that do not need them, give those answers a shelf life so the same question is not re-answered four times a year, and let a named deputy handle the standard cases. Every request diverted here is capacity created next week at no cost, and it is the same move that makes any later elevation of the constraint cheaper, because the ramp then begins with some slack rather than none.
6. Charge the tail to the thing that created it. Estimate the ongoing constrained demand each small change leaves behind, and count it against the request rather than letting it disappear into next year's baseline. That does two useful things. It stops the portfolio treating permanent additions as one-off costs, and it builds the case for periodically spending constraint time on removal rather than addition, which is the only intervention that shrinks the sub-threshold portfolio rather than slowing its growth.
7. Give it an owner. Somebody has to hold the aggregate, defend the allocation, run the weekly ranking and report the number upward. This is not a service desk function, because a service desk optimises for responsiveness and this problem is created by responsiveness. It is portfolio economics, and it is the natural work of a Value Management Office: holding a scarce number that no single team can see, and forcing the comparison that nobody else is positioned to make.
The objection you are already forming
"You are describing bureaucracy. The small requests lane exists because our governance is slow and our customers need things to happen. If I start ranking two-day changes I will spend more on the ranking than on the work, my business partners will route around me within a month, and the responsiveness people actually value will be the thing I destroyed."
That risk is real, and it is why most attempts at this fail. But look carefully at what is being proposed, because it is not more governance on small items. Recommendation four is two questions and a list. Recommendation three keeps the lane, keeps the speed and keeps the informality, and simply gives it a size. Nothing here asks for a business case, a board slot or a stage gate on a two-day change, and if your implementation produces any of those, you have implemented the opposite of the argument.
The deeper answer is that you are not choosing between rationing and not rationing. You are already rationing, because the constraint is finite and 420 requests do not fit alongside 14 initiatives. The only question is what does the rationing. Today it is done invisibly, by interruption and social pressure, in a way that funds $275 an hour and $3,000 an hour identically. What is proposed is that it be done visibly, by a comparison that takes a minute, so that the responsiveness you are protecting gets spent on the requests that deserve it.
And responsiveness is worth protecting. That is the finding I would want you to take from these numbers, not the one about waste. The best ninety requests in that queue outperformed the entire approved portfolio by nearly three to one, and they did it because they were allowed to move without ceremony. An organization that reacts to this article by abolishing its small work lane will have destroyed its highest returning asset in order to fix its lowest. The point is not that below-the-line work is bad. It is that you currently cannot tell which of it is which, and you are spending more than a third of your delivery capacity on the assumption that it does not matter.
Somewhere in your organization that rule about work too small to govern is being applied right now, correctly, by someone doing their job properly. It has been applied a few hundred times this year. Nobody has added it up. The question worth asking at your next portfolio review is not which projects are behind. It is what proportion of your constrained capacity was committed this quarter by decisions your portfolio process was designed never to see.
This piece assumes you know which resource sets your delivery speed, which is where to start if you do not, and that you rank work by what it returns on that resource rather than by how big it is. If your aggregate turns out to be anything like the example, the fastest way to convert it into delivered value is the usual one: start less, and price what you do start in the currency that is actually scarce.
Leave a Reply