Harness collective intelligence

Had a conversation with a peer in the creative industry last week that I have not been able to put down since. It kept circling back on me for days, so I did the thing that usually works for me: wrote it all the way out until my head finally went quiet.

I don’t know if any of this is right, though it’s what I believe right now. But if it helps someone wrestling with the same question in their own business, or gets a few people in the comments making the case for a completely different way to do it, it was worth setting loose.

Creative agencies should stop calling their work “art.”

Design gets considered a craft, sometimes even art, inside plenty of agencies. And there’s real subjectivity in it: two designers can look at the same brief and produce two very different, both technically “correct,” solutions. That subjectivity is often used as a shield, in two ways: “it’s creative work, you can’t measure it,” and “incentives kill creative culture.” I don’t buy either.

Art has no client, no deadline, no budget, and no obligation to move a business metric. Agency work has all four. The moment a client is paying for it, “subjective” stops meaning “unmeasurable.” It just means the measurement has to be built differently than a sales scorecard.

Quality can still be defined, even for creative work: did it solve the brief, does it perform (conversion, engagement, usability), how many revision cycles did it take, what did the client’s sentiment look like, how did it hold up under a creative director’s review, would the team put it in their own portfolio. None of those require a single “right answer.” They require agreed criteria, which is a very different thing from “no way to measure it.”

Once you can measure quality, a second question opens up: do you reward hitting that quality bar in less time or for less cost?

I think you should, but carefully. Rewarding raw speed alone just teaches people to cut corners. The reward should sit on the combination: quality held steady (or improved) while time, cost, or revision cycles come down. That’s margin the team actually earned, and it can be shared back: profit pools tied to realized margin, freed-up capacity redirected into R&D or passion projects, extra PTO, a training budget, more creative autonomy on the next brief.

This is where billing model matters, because it decides whether efficiency helps the team or quietly punishes them. Time-and-materials pays for hours, which means the faster and better your team gets, the less the engagement bills. You’re financially rewarded for staying slow.

But the deeper problem with time-based billing is the timesheet itself.

Most entries are reconstructed from memory at the end of a day or week, rounded, and allocated across projects under time pressure. The data is noisy before anyone even reviews it. People pad hours when they’re worried about looking slow, and misallocate them when they’re worried about looking idle on the “wrong” project. None of that is malicious, but it means the input was never as precise as it looks in a report.

Then there’s the cost of running the system on top of that shaky data: logging, chasing late entries, correcting misallocations, reviewing, auditing. That’s real time spent maintaining a dataset that was compromised at the point of entry, instead of going into the work itself. And the part that matters most: that data doesn’t stay in a spreadsheet. It feeds staffing decisions, project profitability analysis, pricing on the next proposal, sometimes performance reviews. Decisions built on it inherit all of its error, but the number they’re built on still looks precise, something like “139.86 hours” or “83.25% utilization,” which makes it feel objective in a way raw judgment never would. Bad data wearing false precision is often worse than no data at all.

Value-based pricing, tied to milestones rather than hours, sidesteps most of this by shrinking what actually needs to be tracked. Instead of hundreds of granular time entries, you’re tracking a handful of dates: when a milestone was agreed to start, when it was delivered, and the amount of time between milestones in your process. Then you average a cost per day the same way you’d average a cost per hour, just at a coarser, harder-to-fudge grain. A milestone date is binary: hit or missed. There’s very little room to quietly pad it, and almost no admin layer sitting on top of it. It also aligns revenue with what the client actually bought, the outcome, not the hours it took to get there.

Splitting those two, the time inside a milestone versus the gap between one milestone ending and the next starting, matters because they point to different problems. A milestone that runs long is usually about the work itself: scope, complexity, or skill. A gap between milestones that keeps stretching is usually about process: waiting on approvals, kickoffs that don’t happen promptly, handoffs that sit in a queue. Both show up as “slow” in a single blended number, but only one of them is about the team doing the work.

The one thing this needs to work: milestones have to be well-defined and insulated from delays outside the team’s control, with their own small pause/reset mechanism for blocked time. And it still has to run alongside a quality bar: a pure day-count, unpaired from quality, just recreates the “reward speed alone” trap in a different unit.

Take the milestone idea one step further and you can build a single unit that folds time and money into one number, one you never need a timesheet to produce.

Define one unit as a typical week of value delivered: the average weekly billing target for a resource at a given level, based on what the business needs that role to generate. That number gets set once, and revisited periodically. It’s a planning decision, not something anyone logs. From there, a resource’s units-per-week is just the value they delivered (pulled from the milestone data you’re already tracking) divided by that baseline. Hit your target exactly and you’re running at 1.0. Someone consistently at 0.6 is under it; someone consistently above 2.0 is generating double what the role is built for.

That gap is the useful part: not as a scorecard to hand someone, but as a question to ask. A senior person running at 1.8 most weeks might just be more experienced: faster decisions, fewer revisions, cleaner scoping. Or the baseline might be wrong for that level. Or that person keeps landing the highest-value revenue stream while someone equally strong is doing valuable but lower-priced work. Or one team has better support (ops, tooling, an account lead who shields them from noise) than another. None of that shows up in a timesheet, which just tells you two people both logged 40 hours and quietly treats an hour of senior strategic work and an hour of production work as identical inputs. They aren’t, and a shared value unit is what finally makes that visible.

It also settles the capacity-planning question without asking anyone to log an hour: total capacity is just headcount times 1.0 unit per person, adjusted for known time off. You’re not tracking what people did. You’re comparing what they delivered against a target denominated the same way the client pays.

That same milestone data gives you an even sharper planning number: days to deliver one unit. Instead of assuming everyone converts at a flat 5-day week, you track how many actual elapsed days it takes a given resource to produce one full unit of value: a real throughput rate, not an assumption. That tells you two things a flat weekly assumption can’t: how much lead time to build in before handing someone the next unit of work, and roughly when they’ll actually free up, so new work gets queued to land as capacity opens instead of being guessed at or piled on early.

Put three numbers together: units currently in progress, units already assigned to each resource, and that resource’s days-per-unit rate. Together they give you a real capacity model, roughly when each person clears their queue and goes idle, how much excess capacity exists across the team before that happens, and how long it would take to burn down. That’s what actually informs two decisions well: who a new unit of work should go to (whoever’s opening up soonest, or whoever’s throughput best fits that kind of work), and when hiring is actually justified: when the pipeline of incoming units keeps outpacing what the team can clear, not just when someone feels busy.

There’s a pricing implication here too, though it’s not automatic. If your team can deliver the quality a client actually wants, and that your own people are proud of, meaningfully faster than a competitor, that speed can be worth something: a brand identity in 2 months instead of a competitor’s 8, quality held equal, matters to a client who needs to move. But speed can just as easily work against you. Claim 2 days for that same brief and it reads the opposite way, like not enough thought went into it, which undermines the quality claim you’re trying to make in the first place. There’s a balance point, and it isn’t fixed; it’s set by what your ideal client actually values. Fast-and-good is only a pricing lever when your positioning, your product, and your ICP are already aligned around it. Speed on its own doesn’t sell itself.

The caveats are real. The baseline needs to be set per role or level, and probably per revenue stream too. Comparing a strategy lead’s target to a production designer’s doesn’t mean much until you’ve separated the two. It needs periodic recalibration as pricing and team composition shift. And it should be read as a rolling average, not a single week, since milestones don’t land evenly.

What I wouldn’t do, with any of this, is manage individual creative people to hours logged or ask them to account for their day. That’s how you get slow, padded, defensive time-tracking instead of good work. Some things are better left unmeasured at the individual level: the loose, non-linear time that produces the best creative ideas doesn’t show up well on a timesheet, and trying to account for it usually kills it.

The remote and hybrid shift made this same mistake more visible, not less. A lot of companies replaced office visibility with webcam-on mandates, mouse-movement or keystroke monitoring, and “active hours” software, a more invasive version of the same input-tracking instinct that made timesheets weak in the first place. It measures whether someone appears present, not whether the work got done, and it sends an explicit message: we don’t trust you unless we can see you.

That’s the opposite of what most companies say they want. “We don’t care where or when you work, as long as it gets done” is a values statement about trusting output over presence, and it only holds up if the company actually built a way to measure output. Without that, “flexible, trust-based culture” and “we track your mouse movements” end up running side by side, and people notice the gap between the stated value and the actual behavior faster than almost anything else a company does. Value delivered (milestones hit, work at the agreed quality bar, units against a shared baseline) is the version of measurement that’s actually compatible with remote and hybrid work, because it doesn’t care where or when the work happened. It only cares that it happened, and that it was good.

Utilization rate (rewards staying busy, not being efficient), billable hours as a standalone metric (rewards slowness), revision count in isolation (punishes rigor), and CSAT captured right at project handoff (a lagging, gameable snapshot). All four optimize for the next 30 days at the expense of the next three years. I’d rather build around retention and repeat engagement rate, margin per project, referral rate, and team retention. Slower-moving numbers, but ones that actually reward doing excellent work sustainably.

Time to answer both shields from the opening directly.

The first: you can’t incentivize creative people because their work is subjective. That’s really the “you can’t measure art” objection restated as a compensation problem instead of a measurement one. If quality can be defined, agreed criteria, not a single right answer, then it can be rewarded. Subjectivity was never the real obstacle. Trying to reward it with the same blunt tools built for sales is.

The second, and the one I hear more often: push productivity or efficiency incentives and you’ll damage a culture that values creativity. That treats creativity and productivity like opposite ends of a seesaw, more of one automatically costing you the other. That doesn’t hold up, and creativity doesn’t need protecting from accountability to stay a real value. What actually causes the damage is bad incentive design, not the goal of efficiency itself. Reward raw speed on its own and yes, the work gets worse, faster. But that’s a badly built incentive, not proof the two can’t coexist. Everything earlier in this piece, quality held steady while time or cost comes down, efficiency gains shared back as R&D time or profit pools, exists specifically so productivity and creativity pull in the same direction instead of against each other.

One more piece makes this work: the incentive doesn’t have to be identical to be fair.

Equal and equitable aren’t the same thing. A senior strategist and a junior production designer create value differently, at different rates, in different ways, so scoring them on the same metric at the same weight treats two different jobs as if they’re one. Equitable means the reward is matched to the role, the level, and what that person or team actually controls, not that everyone gets the same number. That holds at the individual level and at the team level. A creative team and an account team can run under two different incentive philosophies inside the same company and both be fair, as long as each one is actually matched to what that team is being asked to do.

A few things that hold up better than a straight bonus check, and most of them don’t require deep pockets:

• Profit-sharing tied to team or agency-wide margin, not individual project margin. Creative work is collaborative, and pitting individuals against each other on P&L usually damages the crit culture that produces the best work.
• Being selected to teach becomes the reward. GE ran leadership development this way for years at its Crotonville institute: senior leaders personally taught the curriculum, and getting tapped to teach it was a status marker in itself. The same works with a senior creative running a masterclass for the rest of the team.
• Peer-nominated recognition that doesn’t wait for a manager. Google’s peer bonus program lets any employee award a peer a modest, no-approval bonus with a specific reason attached. It keeps recognition current and specific instead of only flowing top-down at review time.
• Protected time to work on anything, in the spirit of Atlassian’s FedEx Days or Google’s 20% time. The reward is the removal of constraints, not an object.
• A protected budget for award submissions, and paid time to develop the work worth submitting.
• Career ladders built on craft and mentorship, not tenure or utilization.
• Recognition matched to the person, not a default. Some people are motivated by public credit, some by a manager’s undivided attention, some by a tangible gift tied to their actual interests: a gift card to their hobby, not a branded fleece. Knowing which one lands for each person costs nothing and works far better than defaulting to whatever’s easiest to buy.
• Near-zero-cost rituals that punch above their weight: a handwritten note, letting the team decide how it wants to celebrate a win, a small memento tied to a specific project, leadership simply showing up in person to say thanks.
• For senior creative leads specifically, longer-horizon incentives (equity, profit pool vesting, retention bonuses) tied to multi-year value, since that talent is expensive to lose and slow to replace.

None of that works without deciding what triggers it, and this is where a single KPI always breaks. I’d build it as a small scorecard across five categories, and pay out against the blend rather than any one number:

• Craft: CD/peer review scores, award entries and wins, portfolio-worthy output rate
• Business: margin per project, efficiency gains at held quality, value realized vs. price
• Client outcome: the actual result the work drove (conversion, engagement, brand lift), account retention
• Growth and behavior: visible personal development (new skills, certifications, habits actually built), decision ownership (did they own the call, learn from it, and adjust, regardless of how it landed), and impact on the people around them (mentoring, peer support, who they made better just by being on the team)
• Talent: team retention, referral hires

That’s close to why the Refraction Pathway, a framework I wrote about in my book EMBARK, gives growth and accountability their own full zones instead of folding them into craft or margin. None of it shows up on a client invoice, and it’s easy to leave off a scorecard built around craft and margin. But it’s usually the best predictor of who’s still doing great work three years from now.

It also forces a distinction worth sitting with, one Ben Horowitz gets at in The Hard Thing About Hard Things: not all risk is the same. A bad risk has no real upside case even before it plays out. A good risk carries real risk too, but the potential payoff justifies taking it. Someone can make a good, well-reasoned call, upside worth the downside, and still not have it land because of something outside their control. Someone else can take a bad risk and get lucky. A scorecard that only pays out on results can’t tell those two apart, and ends up teaching people to avoid risk and hide mistakes instead of owning them.

Weight it by what the business actually needs right now: an agency chasing growth might weight client outcome and referral hires higher, one protecting margin might weight efficiency higher, one rebuilding culture after turnover might weight growth and behavior and talent higher. The mix should move as the business’s stage and priorities move, but the categories should all stay represented. Drop any one of them and you’re back to optimizing the short term at the expense of the other four.

None of this means creative work isn’t craft, or that it doesn’t deserve real creative latitude. It means the agency still owes its client a business outcome, and “we can’t measure art” shouldn’t be the excuse for not being accountable to it. Get the environment right, clear criteria, real ownership, incentives that reward good decisions and not just lucky ones, and people stop needing to be watched to do great work. They just do it.