RSA Copy Written From Sales Calls and Review Mining

We read at least twenty recorded sales calls and two hundred customer reviews before writing a headline. Copy at this level is a research output, not a writing task, and the research is most of the invoice. Where the words come from: Call recordings from Gong, Chorus or plain Zoom archives. Support tickets. Win-loss notes. The review corpus that matters for the category — comparison threads and vendor review sites for B2B software, retailer Q&A and Trustpilot for furniture, cancellation surveys for pet subscriptions. We are looking for two things: the objection stated in the buyer's own words, and the sentence that shows up in the first ninety seconds of calls that closed. Those two become headline one and the description that carries it. We also pull what your competitors are running right now from the Google Ads Transparency Center. A differentiation claim that three other advertisers are making in the same auction is not differentiation. How the ads are built: Fifteen headlines as a matrix of angle by proof, not fifteen rewordings of one promise. Every asset is written to survive being combined with any other, because that is what the system will do with it. Descriptions carry only claims the landing page can substantiate. If the page cannot, the claim does not ship. Pinning: we pin position one only where a legal line, a regulated claim or a brand requirement demands it. Pinning all three positions collapses the combination count and ends asset-level learning, which a surprising number of accounts do to themselves without noticing. In categories where the claim is the risk — pet health, financing offers on furniture, anything with a compliance reviewer — we pin, accept the cost, and say so in writing. Ad Strength is not an input to Ad Rank. We ship Average-rated ads deliberately and will not rewrite working copy to satisfy a meter. How we decide what won: Asset-level Low, Good and Best labels are relative rankings inside one ad, not performance data, and acting on them early is how accounts churn their strongest headline. We wait for asset impressions to accumulate — as a rule several thousand per asset — and read the combinations report alongside them. Structural tests run as campaign experiments at a fifty-fifty split, with the read date set by conversion volume rather than a two-week calendar habit. For most accounts we take on, that means roughly one hundred conversions per arm before anyone is entitled to an opinion. Landing page correspondence: The headline promise has to appear above the fold on the page it sends to. Where we can edit the page, we do. Where we cannot, we write to what is there and hand over the specific line the page needs, with the reason. What you get: A copy document with the rationale and source behind each asset, a Google Ads Editor CSV ready to import, an asset naming convention so that asset reporting is still readable a year from now, and ad customizers driven by a business data feed wherever price, stock or delivery window varies. What we will not do: Dynamic keyword insertion as a default. Keyword-stuffed first headlines. Volume programs — a hundred variants a month is churn dressed as testing and it starves every asset of the impressions needed to judge it. Superlatives your legal team has not seen. Copy for a landing page we have not been allowed to read. Who this is not for: Agencies buying white-label copy production. Companies that cannot give us recorded sales conversations or a review corpus; without source material this is guesswork with an invoice attached. And accounts where the offer itself is the problem — better copy makes that fail faster, and we would rather tell you now than in month three.