The definition
Pre-spend creative intelligence is the practice of evaluating finished or near-finished ad creative with predictive models before media spend begins, producing a structured diagnostic (scores, benchmarks, and a verdict) while the creative can still be changed at near-zero cost. It exists for the people who decide which creative receives budget: performance marketers, agency creative strategists, and the budget holders they answer to. The decision it improves is singular and specific: should this asset receive spend as it is, be revised first, or be rebuilt?
That is the whole category in one paragraph. Everything else on this page explains why the category exists, and what it is not.
The Pre-Post Gap: why the category exists
Every ad goes through two moments of truth. The first is the creative review, when the team watches the cut and decides whether it is ready. The second is the performance dashboard, when the metrics come back after the budget has been spent.
The gap between these two moments is where the most expensive mistakes happen. We call it the Pre-Post Gap, and inside it three things are true simultaneously:
- The creative is still editable. The cut exists in a project file. Changes are possible.
- The budget is not yet committed. Media spend has not started. The cost of changing direction is near zero.
- There is no diagnostic signal. The team has opinions, but no structured data about how the creative is likely to perform.
The third point is the problem the category solves. Creative review relies on experience, taste, and gut feel. Those are valuable inputs, but they are not diagnostic. They cannot tell you which specific perception dimension is weak, how the creative compares to category norms, or what a target buyer's objection would be.
Post-launch data has the opposite flaw. It is precise, but it arrives after the money has moved. By the time CPA is rising or CTR is dropping, the options have narrowed to keep spending, pause, or start over. Pre-spend creative intelligence exists because neither opinion nor post-launch data delivers a signal inside the gap, where acting on it is still cheap. The economics of that window are covered separately; the short version is that most creative-driven waste concentrates there.
What it is not
The category is easiest to place by contrast with the four things it gets confused with.
Not A/B testing
A/B testing is a live experiment: two variants receive real spend, and the market picks the winner. It is the ground truth for comparing finished options, and nothing pre-spend replaces it. The differences are timing and cost. A/B testing happens after budget commits and charges media prices for its answer; pre-spend evaluation happens before, at near-zero cost, and tells you which variants are structurally sound enough to deserve a live test at all. One filters, the other proves. The full comparison is in AI creative scoring vs A/B testing.
Not a consumer panel
Panels and focus groups put creative in front of real people and collect stated reactions. They measure humans; pre-spend creative intelligence runs model predictions and scenario simulation. Synthetic buyer interviews, structured interviews with modeled buyer personas rather than recruited respondents, can surface likely objections in minutes, but they are hypothesis generators, not evidence about real consumers. When the stakes justify weeks and a research budget, a real panel remains the right tool. The trade-offs are laid out in SaliencyLab vs focus groups.
Not post-launch analytics
Analytics platforms tell you what happened: spend, clicks, conversions, by audience and placement. They are measurement, and they are indispensable. But they only speak after launch, and they describe outcomes without diagnosing the creative causes. Pre-spend evaluation is the other half of the loop: a forward-looking signal about the asset itself, before there is anything to measure. Teams that run both start with a diagnostic instead of a guess, then let live data correct it.
Not a generative creative tool
Generative tools make ads: variants, copy, resized formats. Pre-spend creative intelligence judges ads. The distinction matters more as generation gets cheaper, because producing twenty variants in an afternoon is now easy, and deciding which of the twenty deserves budget is the new bottleneck. Evaluation is unrelated to production volume; it is the quality gate that sits after any generator, human or AI.
What the signal looks like in practice
In SaliencyLab's implementation, RoastIQ returns a diagnostic across five fixed KPIs (Beat the Skip, Get Noticed, Brand Impact, Sell Proposition, Build Brand) with benchmark context and a verdict of Scale, Sharpen, or Rebuild, in about 90 seconds for images and under 3 minutes for video. A typical pass looks like this: a team uploads three candidate cuts, one comes back Rebuild with a weak opening, one comes back Sharpen with late brand entry, one comes back Scale. The Rebuild is recut before it costs anything, the Sharpen gets a targeted fix and a rescore, and the live test runs on two viable variants instead of three unequal ones. Reading the output well is its own skill: see how to read a RoastIQ report, and the full pre-launch workflow for where the diagnostic sits in the end-to-end process.
The evidence, stated plainly
A category that evaluates creative before spend must show its predictions correlate with something real, so here is the record. The scores are model predictions validated against public engagement and click-intent outcomes, cross-validated against more than 1,200 ads with public outcome data. Held-out, out-of-sample validation as of May 2026: Spearman correlation of +0.31 against TikTok engagement (n=700), +0.30 against TikTok CTR (n=691, 5-fold cross-validation), and +0.32 against YouTube view counts (n=403, 5-fold cross-validation), with a 6.5x top-vs-bottom quintile lift by predicted score. The scores do not predict sales, ROAS, attributed conversion, or brand recall; those are in-market business outcomes, and no pre-spend signal forecasts them honestly. The validation approach is documented in full in the methodology.
Moderate correlations, honestly labelled, are the right expectation for the category. They support ranking creative and catching structural weakness before spend. They do not support treating any score as a forecast of campaign results.
What this can and cannot tell you
- It can tell you whether an asset clears a structural bar before it receives budget, which perception dimension is weakest, and how the creative sits against a benchmark pool. Those are facts about the model's output, backed by the validation above.
- It should shape, not make, your decision. Treating a Rebuild verdict as a hard stop and a Sharpen as a revision prompt is our recommendation, grounded in the quintile lift but still a policy you should adapt.
- It cannot tell you whether the market wants the offer, what a competitor will do, or what your sales number will be. Attention heatmaps in this category are predictions of where viewers will look, not eye-tracking measurements. Synthetic buyer output is simulation, not consumer evidence. Live results remain the ground truth, and a team that stops checking predictions against them is using the category wrong.
The gap is the point
The Pre-Post Gap is not closed by better dashboards or bigger panels, because both operate outside the window where change is cheap. It is closed by a signal that arrives while the project file is still open. That is the entire premise of pre-spend creative intelligence: not a prediction of success, but a structured diagnostic while the ad is still editable.
If you want to see what that signal looks like on a real creative, the example report shows the full diagnostic, verdict and evidence included.