Tracking & results

Performance and Autopilot

Understand what Adwave Autopilot does, how campaign results are measured, and how undelivered media is credited.

From an ad to a measured result

1. 01Ad exposure

   A person sees or clicks an ad.

2. 02Website visit

   Your business’s tag records a visit.

3. 03Conversion

   A configured action records a result.

4. 04Reporting

   Credit and assists appear separately.

A reported conversion is not proof of an incremental sale. Lift experiments answer a different question.

On this page

What we don't promise

Adwave does not guarantee a particular ROAS, lead count, or sales result. Your offer, website, market, and creative all matter. Use early results to assess the campaign before increasing its budget.

In-app launch view ranges (when inventory data is available) are estimates. They are not a guarantee of impressions, and under-delivery credits still follow ads that ran.

Measured on your site

Results come from your website tracking: cost per conversion once wins count, and return on ad spend (ROAS) once you set what a win is worth or send real order amounts.

Platform dashboards (Meta, Google, streaming) report their own view of delivery. Adwave's ROAS and CPA use your Adwave tracking snippet (site visits and conversions on your website). Browser privacy settings and consent choices can still limit measurement.

Cost per conversion (CPA) works as soon as conversions count. Return on ad spend (ROAS) needs a dollar value on those wins. When we cannot measure (no tag, or conversions with no value), we show a dash instead of 0.00×. Autopilot still optimizes for cost per conversion until you set values.

Autopilot on

Audience activation and funnel tuning run every hour. Creative, journey-stage budget, segment, and bidding decisions run daily from your site data, with a larger weekly journey mix pass and monthly holdout readout. On live campaigns, Autopilot stops testing weak interest segments, brings in replacement audiences, adjusts demographic filters when delivery stalls, and works to improve delivery without pausing the campaign. Every decision is logged so you can see what changed.

Budget as a ceiling

Your daily budget is the maximum media charge for a day. Campaigns draw from your prepaid wallet and pause when funds run out. Wallet & billing covers how undelivered budget is credited back.

Create before you spend

Analysis and campaign ads for businesses come first, included with the campaign. Direct business signups get $60 starter credit in the wallet for media and extra ads. Agency workspaces skip that credit and fund the agency wallet at published unit prices. Paid agency generation and optional extra ads can charge the wallet before a campaign runs.

You see what changed

A weekly email digest plus an in-app Autopilot feed show what Autopilot did.

After launch, Home shows an Autopilot feed of recent decisions, and each campaign's Activity tab counts every check-in: "Checked 312 times this week, 4 changes", with each day closing on how many looks ended in nothing to change. Every Monday you also get a weekly email digest of spend, conversions, and cost per conversion: plain language.

Assisted conversions and Lift

Last-touch attribution determines attributed results and ROAS. Daily campaign charges come from your prepaid budget, with credits for undelivered media. Assisted conversions on the Performance tab show when another channel on the same campaign touched someone earlier (a click, household view, or person-resolved view) before a different channel got last-touch credit. Assists are reporting-only: they do not move money. Funnel campaigns show Mobile web and TV as separate platform cards (from real line-item delivery), not one combined open-inventory row. Channel cards are what the campaign buys; the Devices & locations and publisher sections show where ads showed. Desktop may appear as an observed device even though it is not offered as a channel on the picker.

Path breakdowns can overlap; the headline assist count is the distinct union, not a sum. Person-resolved assists need overnight identity matching, so coverage can lag. Soft-credit scores used by Autopilot look at a recent window and can nudge journey-stage budgets; they never change last-touch ROAS or wallet charges. Platform cards show all-time assists.

The Lift scorecard is a separate causal check (holdout, overlap, effects on other channels). When a holdout read is still inconclusive, we say so and do not reshuffle holdout states mid-flight. Start a new campaign if you need a fresh causal map.

How rates are defined, on every tile and in every sentence:

• Click rate = clicks ÷ views, both reported by the ad platforms.
• Frequency = views ÷ people reached, both from the same source (the first-party pixel). When the pixel saw less than half of the delivery the platforms report, the tile reads "Not measurable" with the reason, never a number.
• Conversion rate = wins ÷ site visits from ads. When wins outnumber attributed visits, the funnel shows Views to Wins and asks for the click id on your landing links.

Signals: what the platform learned

Anything that needs you lands on one page, Signals in the sidebar: tracking health, connection problems, a campaign running behind its forecast, a jump in cost per win, a short wallet runway, and the decisions Autopilot held back. Each row says what changed, against what, when we first saw it and how many times, with a What to do link into the page that fixes it.

A signal is one row per fact: a repeat refreshes the row instead of adding another, and a fact that clears on its own (a detector that stops firing for three days, a wallet that is funded again) closes the row for you. Mark as handled closes it by hand. Warnings and critical rows also open the weekly email under Needs a look. The Autopilot status uses the last completed check, so you can tell when checks are overdue.

Windows

Every number names the days it covers. Home, the campaigns list, a business HQ and a campaign Summary all default to the last 30 days; the 7 and 14 day views and All time are written into the address explicitly so a shared link shows the same window. Comparisons name the prior dates ("vs Aug 1 to Aug 30") rather than "the prior period", and a return sentence always ends with its window, "last 30 days".

What's next

• Install tracking so site conversions can power CPA, and set what a win is worth when you want ROAS.
• Wallet & billing for ceilings, credits, and auto-refill.
• Campaigns for budgets, flights, and the live Performance tab.