Creative concept portfolio sizing for a flat budget

Ask a UA team how many creative concepts they should be testing at any given time and most will answer with a production capacity number: however many the team can turn around in a sprint. That answer has nothing to do with how many concepts a given spend level can actually generate a statistically useful read on, which is the number that should be setting the portfolio size in the first place.

The question nobody answers with a number

Concept portfolio sizing gets treated as a creative resourcing decision when it is really a media-buying decision wearing a creative hat. A concept needs a minimum volume of spend and impressions before its performance data means anything, and that minimum does not shrink just because a team wants to test more ideas. Run too many concepts against a fixed budget and every individual concept gets an inconclusive read, which either wastes the whole testing cycle or, worse, produces a result that looks conclusive by chance and sends the team optimising toward a false signal. Run too few and the team leaves genuine creative headroom untested, mistaking a small, safe portfolio for discipline when it is actually under-testing.

As we set out in Detecting Creative Fatigue Before It Shows Up in CPI, the same statistical patience that catches fatigue early also governs how many concepts a budget can responsibly support at once. Both problems come from the same root cause: teams sizing creative decisions around production convenience rather than the spend available to test them properly.

How budget should set concept count, not the other way round

A workable rule of thumb starts from the minimum spend a platform's algorithm needs to exit learning phase and reach a stable delivery pattern for a single ad, then works backward. If a network typically needs $1,500 to $3,000 in spend per concept to clear learning phase and produce a usable early read, and a team has $30,000 available for a testing cycle, the arithmetic caps the portfolio at somewhere between ten and twenty concepts, not the twenty-five to thirty a production team might be capable of delivering. Anything beyond that cap is not additional testing. It is diluted testing that looks like more work for a worse answer.

A sizing framework

Apply this in order when planning the next testing cycle:

• Set the spend-per-concept floor first, based on the platform's typical learning-phase requirement, not an internal target. This number should come from your own account history, not a published benchmark, since learning-phase spend varies by vertical and by how competitive your specific auction is.
• Divide the testing budget by that floor to get a maximum portfolio size, then subtract 20% as a buffer for concepts that need a second push to clear learning phase properly.
• Reserve at least a third of the resulting slots for genuinely new concepts, not iterations on a currently running creative, since a portfolio made entirely of small variations on the same idea will converge on a local optimum rather than finding a new one.
• Cap iteration slots at two per surviving concept per cycle. A concept that has already earned a third iteration without a clear winner is consuming testing capacity that a new idea could use instead.
• Revisit the floor every quarter. Learning-phase spend requirements drift as auction dynamics change, and a floor set six months ago on a hypothetical $30,000 budget and a $2,000 concept floor might now need eight fewer or eight more concepts to fit the same spend.

Adjusting the framework across channels

The spend-per-concept floor is not a single number across an entire media plan, and treating it as one is a common mistake once a team starts applying this framework. A channel with a smaller, more competitive auction, such as a newer network still building out its advertiser base, may need less spend to reach a stable read than an established channel with a deep, mature auction where the algorithm has more inventory to explore before settling. Calculate the floor per channel using that channel's own account history, then size each channel's portion of the testing budget separately rather than applying one blended floor across a media plan that spans channels with meaningfully different auction dynamics. A single blended floor tends to overestimate how many concepts a smaller channel can support and underestimate how many a larger one can, which quietly skews a portfolio toward whichever channel the floor was calibrated against.

When to retire before adding

Portfolio sizing only works as a discipline if retirement is treated as seriously as addition. A team that adds three new concepts every cycle but rarely retires an underperforming one is not managing a sized portfolio, it is accumulating one, and the effective spend-per-concept floor quietly erodes as the same budget gets split across more surviving creatives than the sizing framework accounted for. As we covered in The creative concept retirement signals to watch for, a concept that has cleared learning phase and settled into a stable but mediocre delivery pattern for two consecutive cycles is a stronger retirement candidate than one still showing volatility, because volatility at least suggests untapped upside.

This channel-level view also changes how a team should respond when overall budget shifts. A flat total budget does not mean every channel's concept count should shrink proportionally; it means each channel's floor should be recalculated against whatever portion of the budget it retains, which can produce a portfolio that looks quite different in shape even when the total spend is unchanged from the previous cycle.

What this changes about the planning conversation

Sizing the portfolio to the budget, rather than to the production team's output, turns a recurring argument about how many concepts to greenlight into an arithmetic exercise the whole team can see and agree on before the cycle starts. It also makes the trade-off explicit: a bigger portfolio needs a bigger testing budget, and a flat budget means a disciplined cap on concept count is not a creative constraint imposed from outside, it is the honest ceiling the spend itself sets.

Related archive reading

These articles provide related context and remain subject to their stated review status.

• Setting a Creative Iteration Cadence for Live Mobile Games
• Q4 Creative Capacity Planning: How Much Is Enough