17 February 20267 min read
Two deals that were going to close in March both slip to June, and a third that was written off signs in a fortnight. Every agency knows this pattern, and most respond by trying to forecast better: weighted pipelines, more disciplined stage definitions, a monthly review that argues about probabilities.
That work is worth doing and it will not solve the problem, because the variance is not a measurement error. Client budget cycles, procurement, and the personal circumstances of one champion inside a client organisation genuinely determine when a deal lands, and none of them are visible to you. The useful move is not a better prediction. It is a staffing shape that does not punish you for being wrong in either direction.
Both errors cost, and they cost differently
Understaffed, you turn work away, or you accept it and deliver it badly. The first is a revenue loss you can count. The second is a reputational loss you cannot, and it tends to arrive with rework attached.
Overstaffed, you carry bench cost. That is real money, and it is also corrosive in a way spreadsheets miss: idle senior people get restless and leave, and the ones who stay get put on work that does not need them, which trains the organisation to pad estimates.
The asymmetry matters. Bench cost is immediate, visible and survivable. Delivering badly to a client you fought to win is slower to show up and much harder to undo. Plan accordingly: a small deliberate bench is a reasonable purchase, and a large accidental one is a management failure.
Core and flex
The shape most agencies converge on is a permanent core sized to the work they are confident of, plus flexible capacity for everything above it. What belongs in each is the actual decision.
- Core, permanent: client-facing leadership, architecture and technical direction, anyone holding deep domain knowledge of a long-running account, and the people who set and enforce your standards.
- Flex, partner-supplied: implementation capacity on well-specified work, second and third people on a team that already has a lead, specialist skills you need a few times a year, and coverage outside your own hours.
The line is roughly the same as the differentiating-work line. Keep permanent whatever compounds: relationships, domain understanding, standards. Buy flexibly whatever is legible from outside and can be handed over with a specification.
Sizing the core is the uncomfortable part. The usual mistake is sizing it to the good quarters, because that is what the business feels like when the decision is being made. Sizing to something closer to a reliable floor, the work you would still have if the pipeline went quiet, is what makes the flex layer meaningful rather than decorative.
A partner is not capacity until you have used them
Agencies routinely believe they have flexible capacity because they have a signed framework agreement with somebody. They do not. An unused partner is a phone number, and calling it in a crisis produces two to four weeks of onboarding at exactly the moment you have none, plus the discovery that the people they described are on someone else's project.
Capacity you have never used is not capacity. It is a phone number and an optimistic assumption.
Keeping a partner genuinely ready costs something, and the cost is the point. Give them a small piece of continuous work rather than only surge work: a maintenance stream, a test suite, an internal tool. It keeps two or three people familiar with your standards, your repositories and your review process, and it means the surge is an expansion of an existing team rather than a cold start.
The alternative is a paid retainer that reserves named people, which is cleaner but is bench cost by another name. Continuous small work is usually the better trade because you get output for the money.
The overlap constraint on flex
Flexible capacity is directed capacity, at least at first, because a team you brought in this month does not yet know enough to work unsupervised. That makes shared working hours a practical requirement rather than a preference. A partner who overlaps most of your day can absorb direction in real time, which is what compresses the ramp; a partner ten hours away is workable on well-specified continuous work and painful on a surge, because a surge is exactly when specifications are thinnest.
For European agencies this is the practical argument for near-shore partners. A team an hour or two ahead of Berlin or Amsterdam shares the whole working day, which means the first fortnight of questions costs hours instead of days.
Decide in advance what you will do at the edges
The last piece is a set of rules agreed when nobody is panicking. What do you do when a deal lands and you cannot staff it: delay the start with the client's agreement, staff it partly and phase it, or decline? What do you do when utilisation drops below a level you have named: internal product work, training, or a reduction in flex spend? Which accounts are protected from being restaffed under pressure?
Written down, these are ordinary policies. Decided in the week the situation happens, they become arguments between sales and delivery, and the resolution usually favours whoever is loudest rather than whoever is right.