Priced on what it manages.

Realvertising coordinates advertising across Google, Meta, TikTok, LinkedIn, Microsoft and other channels from one level above the ad networks. It reads your first-party business data — conversions, CRM records, collected revenue — and decides which channels deserve budget, which business outcome each should optimise for, which audiences to expand or suppress, and which creative and landing pages get tested. The networks keep running their own auctions; the Growth Brain decides how they work together. Every change is bounded by the Campaign Constitution, logged with its evidence, and reversible.

No. Google, Meta and TikTok run the auction — bid shaping, placement, per-impression delivery, audience matching inside their own inventory — and nothing here competes with that. Performance Max, Advantage+ and Smart+ keep running exactly as they do today; they are the execution layer the Growth Brain steers, and native network AI works better when it is fed real outcome data. What changes is the layer above them: which network gets the next euro, which business outcome each campaign is told to chase, and what happens when a network's reported ROAS does not match collected revenue.

Yes. You state the campaign intent — outcome, market, budget ceiling, constraints — and the platform builds campaign structure, audiences, budgets and ad variants for each channel through its channel adapters. Nothing reaches a live ad account until it passes the Campaign Constitution's checks, and approval gates can require a human on any account you choose. Every created object carries an idempotency key, so a retried build never leaves a duplicate campaign behind.

Yes, inside limits you set. The Campaign Constitution fixes a maximum daily movement, per-channel floors and ceilings, a maximum CAC, a minimum margin and the campaigns marked protected. The budget engine — deterministic code, not the model — computes the reallocation; the model decides that a reallocation is worth proposing and explains the reasoning. Anything outside your limits arrives as a proposal awaiting approval instead of an action.

Yes. It generates headlines, descriptions, image and video concepts and per-channel variant sets, drawing tone, brand rules and approved assets from your brand asset system and your content platform. Creative is produced against a hypothesis — an angle, an audience, a journey stage — and enters a test plan rather than an asset folder. When creative fatigue signals fire on a running variant, replacements are generated from the concepts still holding up.

It knows what your CRM knows. The CRM truth layer reads lead status, opportunity stage, deal size, win rate and close date, then returns qualified outcomes to the networks as offline conversions so their bidding stops chasing form fills. A campaign credited with cheap leads that never reach opportunity stage becomes visible as exactly that. Qualification rules stay in your CRM; this product consumes them and never invents a second scoring model.

Yes. Collected revenue from your finance system is joined to the journey by the attribution service, so campaigns are judged against recorded amounts rather than conversion counts. Value signals go back to each network's bidding model, which is what lets network-side value optimisation work on real figures instead of static placeholders. Revenue is credited to channels under the attribution rule you choose — credit, not proof of cause, which is what experiments exist for.

Yes, where your finance or accounting system supplies the inputs. The business metrics service computes contribution margin per product, order or customer from cost, discount and refund data, and Margin ROAS is reported next to revenue ROAS. Budget rules can then be written against margin: a minimum gross margin floor, a maximum CAC per segment, a pause rule for products selling below their floor. Realvertising does not define your margin — it reads the definition your finance system already uses.

Platform ROAS is what a network reports for conversions it can see and claim, on its own attribution window. True ROAS reconciles four layers kept separate — network truth, attributed truth, CRM truth and financial truth — and sets collected revenue against spend without flattening them into one figure. The gap between the two is the useful part: double-counted conversions, refunded orders, leads that never closed, and revenue no network claimed at all. Both numbers stay visible, because the platform figure still governs how that network's own bidding behaves.

Yes. The attribution service runs independent journey tracking across sessions, devices and channels, applying the model you select — first touch, last touch, linear, position-based or data-driven — over an attribution window you configure. Attribution distributes credit under a rule; it does not establish what would have happened without the channel. Every figure states the model and window that produced it, so two numbers computed under different rules are never compared as though they were the same measurement.

Yes — holdout tests and geo tests, run by the experiment service. A holdout keeps a defined share of an audience or a set of regions out of a campaign and compares outcomes against the exposed group, which is how incremental effect is estimated rather than credited. The Growth Brain proposes experiments where attribution and network reports disagree, and it will state plainly when a result is not yet separable from noise. Experiment results and attribution results stay in separate columns and are never merged into a single score.

The Campaign Constitution. It sets an absolute spend ceiling per account and per channel, a cap on how far any budget may move in a day, a maximum number of actions per run, a maximum CAC, a minimum margin, and the list of protected campaigns that may never be touched. The action service checks every proposed change against those limits before it reaches an ad account and rejects whatever falls outside them. Limits are account-level configuration that you own; the model cannot widen its own bounds.

Yes. Each action is stored as a decision object: what changed, the previous value, the new value, the evidence behind it, the rule that permitted it, who or what approved it, and when. Every action carries a reverse path, so a budget shift, a pause or an audience change can be rolled back from the log itself. Actions are idempotent by key, so a retry after a network timeout does not apply the change twice.

Yes. Your consent platform holds the configured state per user and region; Realvertising reads that state and enforces it, deciding which signals may be collected, which may be sent to which network, and which must be withheld. Where your policy allows it, consent-denied traffic is modelled in aggregate rather than tracked individually. This product enforces the policy you configure — it does not decide what the law requires of you, and it does not run a second consent engine beside yours.

No. One ad account is enough to begin: the platform reads network truth, builds independent journey attribution from your site, and reports where the two disagree. Connecting your CRM turns lead counts into opportunity and customer quality; connecting your finance system turns revenue into margin, and each connection adds a class of decision the earlier ones could not support. The interface shows which truth layers are live, so you always know which questions your current data can answer.

No. Generation is one capability inside a coordination system — headlines, descriptions, image and video concepts are real and included. The system exists for the decisions around them: which channels get budget, which outcome each should optimise for, which audiences to expand or suppress, which variant has fatigued, which landing page receives the traffic. A generator hands you more ads; this hands you an allocation of spend across networks that is answerable to collected revenue.

No. Reporting describes what already happened; this plans, proposes and executes changes in your live ad accounts through channel adapters, inside the Campaign Constitution. The four truth layers exist so that decisions have a defensible basis, not so that a dashboard has more tabs. You can read every number, but the output of the system is a set of actions with evidence attached, each one reversible.

It says so, and names what is missing. If margin is not connected, it reports revenue and marks margin unavailable rather than estimating it. If a sample is too small or a test has not reached a separable result, it reports the uncertainty and proposes the experiment or connection that would close the gap. The model never computes a business metric itself, so when the deterministic service cannot return a number, there is no number.

Fast enough to matter, slow enough to learn. Pacing and spend checks run continuously, while structural changes respect each network's learning phase so a campaign is not reset while its bidding model is still stabilising. The daily movement cap in your Campaign Constitution limits how far anything can travel in twenty-four hours, and minimum evidence thresholds stop a change from firing on one noisy day. Where evidence is thin, it proposes an experiment instead of a reallocation.

Yes. Workspaces separate brands, clients and markets, each with its own connected ad accounts, data connections, Campaign Constitution and approval rules. Agencies work across clients from one place while budgets, data and permissions stay isolated per workspace, and every action is attributed to the person or rule that authorised it. Cross-workspace roll-ups compare spend and outcomes without mixing the underlying accounts.

Your campaigns keep running exactly as they were. Realvertising never owns your ad accounts — it acts inside them through OAuth you granted and can revoke, and everything it created stays in place under your own account. Disconnecting stops new actions; it does not pause delivery, delete campaigns or claw back creative. The change log stays exportable so you keep the record of what was changed and why.

Anything your Campaign Constitution protects. Brand search campaigns, named markets, budget ceilings and any action type you have marked as approval-only are out of reach regardless of how confident a recommendation is — a rule always beats a score. Outside those, every executed action carries a before state, so a change can be reverted where the network API allows it.