For brands
For in-house teams, we connect your data, model your channels and work with you from the first brief through to weekly planning and optimisation. You can license the platform or choose to work with our team.
Response curves show which channels have moved beyond their most efficient range. The plan estimates the likely impact of reducing spend, with an uncertainty range.
Allocation is guided by marginal incremental ROAS rather than platform-reported ROAS, while experiments progressively replace modelled figures with measured results.
Figures show their source and recommendations show their confidence level. Weekly and monthly reports provide a clear basis for management discussion.
Attribution can overlap across channels, and reported ROAS can overstate the value of channels that are already saturated. Annual plans are often revisited only once, while traditional consultancy projects can take months and be outdated by the time they conclude.
Marketing measurement answers one question: what did each channel add to sales, and what would the next euro add? We answer it with three sources of evidence, and every figure says which one it came from. A media mix model fitted on your history, aggregate weekly spend by channel, revenue or orders and controls such as price, promotions and seasonality, typically over two to three years, gives each channel a response curve and its marginal incremental ROAS, the number that ranks channels for the next euro. Experiments you run, usually a geo-lift test, measure what a channel caused and calibrate the model. And the audited benchmark registry fills the gaps until your own evidence replaces it.
For cross channel measurement, that means online and offline channels in one model, read weekly: television, radio, out-of-home and print alongside search and social. Media measurement by platform still has its place, for optimising inside a channel, but its figures are read as attributed, not incremental. Cross-media measurement across the whole mix comes from the model, and unified marketing measurement is the arrangement of the three: experiments calibrate the model, the model allocates the budget across channels, platform figures serve the optimisation inside each one, and audited benchmarks fill the gaps until your own evidence replaces them.
Brand measurement uses the same model. Channels with long carryover show up in the adstock and the baseline trend, and the plan says how much of a channel's return is deferred rather than immediate; excess share of voice, share of voice minus share of market, is the check on whether the brand is spending enough relative to its category before any allocation is made. DOOH audience measurement is a case of its own: the reach of a digital out-of-home campaign is estimated, not observed, so the plan tags it as an estimate, and the incremental effect can be measured with a geo-lift test, which suits a medium bought by area.
Predictive marketing analytics, in this setting, means projecting scenarios on the fitted curves: each allocation gets a projected return with its band, computed on the model's posterior draws, and a lift whose band crosses zero is reported as not distinguishable from flat. The output is a marketing measurement plan: which figures are measured, which are projected, which are benchmarks, and which test to run first to replace the weakest projection with a measurement. That is marketing science applied to a media plan: it runs in the Media Planner, with our marketing scientists or your own team, and the model, the curves and the plan stay in your workspace.
A Bayesian model with response curves for each channel. Allocations can follow marginal incremental ROAS rather than an average return, while the model card and its assumptions remain available for review.
Review pacing, funnel progression and channel status against the plan, with each figure identified as measured or projected. A weekly report explains what has changed and the actions to consider.
Recommendations set out their hypothesis, proposed action, expected impact and confidence level. Experiments can be proposed from the first plan, helping replace modelled figures with measured results as evidence develops.
Where data is limited, audited benchmarks provide a starting point. Figures are clearly identified, and a first plan can be prepared within an hour.
A geo holdout, a conversion lift or an A/B test that replaces the weakest modelled number with a measured one.
As new evidence becomes available, the team can monitor, refine and re-plan. Over time, the model increasingly relies on your own data and the plan on the model.
Explore benchmark-based media plans and budget scenarios without connecting your advertising accounts.
Get the free preview →A monthly platform fee set by the number of brands: one, up to three or up to ten. Every plan includes onboarding, a monthly review with a marketing scientist and automated execution under human oversight.
Compare plans →A management fee applies to the media spend Kuwalyst manages, at a lower rate on each higher bracket. Custom marketing science and data work is scoped separately.
See media fees →Platform dashboards report attributed sales. Attribution overlaps across channels, and reported ROAS stays high on channels that are already saturated. Kuwalyst connects to your data sources, so the same platform figures are available per channel and aggregated, for online and offline channels alike. It then combines them with a media mix model and experiments, and tags every figure as measured, projected or benchmark, so budget decisions rest on evidence.
Yes. Benchmark-led plans start from your brief plus an audited registry of published benchmarks, every claim verified against primary sources. Each figure is tagged, and the plan states what to test first, so your own evidence replaces benchmarks over time.
That works well. Kuwalyst can operate alongside your agency, and many agencies use the same planner for their clients. In either case, the plan, model and supporting evidence remain in your workspace.
To fit a media mix model, we typically need two to three years of aggregate weekly data: spend by channel, including offline media; revenue or orders from your warehouse or e-commerce platform rather than a platform's attributed figure; and the controls that also move sales, such as price, promotions and seasonality. Without that history, we start from your brief and the audited benchmark registry, tag every figure, and prioritise the tests that will replace benchmarks with your own evidence.
A first plan within an hour: benchmark-led where your own evidence is limited, data-backed once the model is fitted on your history. Within weeks, we can build a Bayesian media mix model on that history, with a model card, response curves for each channel and a calibration plan. From there, monitoring is weekly: the plan is checked against its goals, the recommendations show their confidence, the experiments the plan proposes replace projections with measurements, and the model is refitted as the evidence builds.
The model itself, as artefacts you can inspect and reuse: the model card, the priors, the fitted parameters and response curves with their bands, the fit diagnostics and the calibration record. The plan, the models and the evidence stay in your workspace, whether your team runs the platform or ours does, and models your team or a partner already built can be loaded too.
I had always hesitated to step into the unfamiliar world of advertising, but Kuwalyst’s meticulous analysis and clear explanations gave me the confidence to take that first step.
Tell us where your media plan lives today—even a spreadsheet is a useful starting point. We will show you how planning, monitoring and optimisation could work in your situation.
Talk to us