11:10
amazon-ppc
Chris Rawlings argues that agentic AI (Claude) has, within just a couple of months, made his own previous flagship tutorial on manual spreadsheet/pivot-table analysis of Amazon PPC search term reports "completely irrelevant": simply uploading a search term report to Claude and asking for analysis — or running a purpose-built Claude skill — now produces faster, more visual, and more actionable insights that translate directly into higher Amazon profit.
His prior popular tutorial on search term report analysis via spreadsheets and pivot tables is now obsolete; he added a comment on that video telling people not to watch it anymore.
The "simplest possible attempt" (SPA) is the recommended starting posture: don't be intimidated by AI or feel you need a server farm or a dedicated Mac mini running agents 24/7 — just try the simplest possible prompt with your data.
Process to get the raw data: in the Amazon ad console, go to Sponsored ad reports, create a new report, choose Sponsored Products, set the maximum time range (last 60 days), generate and download the report.
The simplest first prompt is: 'analyze my Amazon search term report and tell me all of the best categories of insights you can give me from the data. Recommend reports, dashboards, or other outputs you think would be helpful,' then upload the report.
Claude automatically collates keyword data across campaigns, solving the exact problem that previously required a pivot table (a keyword's performance is normally split across separate campaign rows).
The first-pass output already included: top 30 keywords by revenue with ACoS, ROAS, and conversion rate; the top 25 most efficient keywords (filtered by a minimum order count for statistical significance, showing ACoS in the teens); and high spend/high ACoS keywords to negate or lower bids on.
You can tell Claude "save this as a skill," and in this case the instruction was to create a skill that generates the same search term dashboard directly from an uploaded search term report, so any team member can run it going forward.
Building a robust skill takes iteration, not a single perfect prompt: he started with a "dumb baby prompt," then added shopper intent, then a better layout, then company branding, then had team members (Martin and Mark) work through it further.
The final skill's dashboard includes: weekly report/match type efficiency, keyword insights, campaign insights, product-by-product insights, shopper intent insights, and a dedicated wasted-ad-spend page.
Shopper intent insights group keywords by underlying buyer motivation (e.g., for a necklace product: gifts for women, mother-daughter, milestone/birthday) and show click-through and conversion performance per intent group.
Identifying which shopper intent performs best/worst is meant to generate a hypothesis for a better primary product image, driving more traffic and sales from the same ad impressions — i.e., more profit without more PPC spend.
Running the skill requires at least the $17/month Claude plan; he says he is giving the skill itself away for free via a link in the video description.
He frames "utilizing and controlling AI" as the most important skill a person can have for the rest of their lives, positioning it as protection against forecasts that 90% of white-collar jobs will disappear by 2030.
He points to a companion video on using Claude for Amazon keyword research, which he says similarly replaced an older manual-research tutorial he had made just a couple of months earlier.
SPA (Simplest Possible Attempt) — A stated approach of starting an AI task with the dumbest, simplest possible prompt rather than over-engineering it upfront, framed as an antidote to feeling intimidated or paralyzed by agentic AI. Apply: Begin by simply uploading your raw search term report to Claude and asking it to analyze the data and recommend useful reports or dashboards, then iterate from there.
AI-driven Search Term Report analysis (replacing pivot tables) — Using Claude to directly analyze an Amazon Sponsored Products search term report and automatically collate the same keyword's performance across multiple campaigns, a task previously requiring a manual pivot table. Apply: Export the search term report for Sponsored Products over the max 60-day range from the Amazon ad console and upload it to Claude with an analysis prompt instead of building a pivot table.
Claude Skills (save/create a skill) — A Claude feature that lets a successful ad-hoc analysis be saved as a reusable, shareable automated tool, e.g. by instructing Claude to "create a skill that generates the search term dashboard.". Apply: After getting a good one-off analysis result, tell Claude to save the process as a skill so teammates can run the same dashboard themselves just by uploading a new report.
Iterative skill refinement process — The described method of building a robust Claude skill by starting with a basic prompt and progressively layering in more requirements (shopper intent, layout, branding) and team feedback rather than expecting a correct result on the first try. Apply: Draft a minimal skill first, test it, then add one feature at a time (e.g., shopper intent categorization, formatting, branding) and have teammates stress-test it before treating it as final.
ACoS (Advertising Cost of Sale) — A per-keyword efficiency metric reported by Claude's dashboard, used to identify both the most efficient keywords (ACoS in the teens) and the worst-performing ones (high spend, high ACoS). Apply: Use ACoS alongside a minimum order threshold to flag keywords worth increasing bids on versus keywords that should be negated or have bids lowered.
ROAS (Return on Ad Spend) — A per-keyword return metric included in the top-30-keywords-by-revenue table generated by Claude. Apply: Cross-reference ROAS with ACoS and conversion rate per keyword to judge overall keyword performance within the dashboard.
Statistical significance filtering via minimum order threshold — A filter Claude applied on its own that excludes keywords with too few orders (e.g., a single lucky 2% ACoS order) from the "most efficient keywords" list so only keywords with enough order volume are shown. Apply: When reviewing keyword efficiency, only trust keywords the tool marks as meeting a minimum order count rather than acting on low-ACoS keywords with only one or two orders.
High spend / high ACoS keyword flagging — A dashboard category listing keywords that are spending significant budget while returning a high (bad) ACoS, marking them as money-losing. Apply: Review this list to decide which keywords to negate entirely or lower bids on to cut wasted spend.
Shopper intent insights — A dashboard section that groups search keywords by the underlying purchase motivation behind them (e.g., for a necklace: gifts for women, mother-daughter, milestone/birthday) and reports click-through and conversion performance per intent group. Apply: Identify which shopper intent groups convert best or worst, then use that to form a hypothesis for a better primary product image or listing change that increases conversion at the same ad spend.
Wasted ad spend dashboard page — A dedicated page in the final skill's output that surfaces where and how ad spend is being wasted on the account. Apply: Check this page specifically when looking to cut costs, separate from the keyword-level ACoS analysis.
The video reframes profit growth versus revenue growth: capturing more clicks/conversions from the same number of ad impressions costs no additional PPC spend, so profit rises faster than revenue when shopper-intent-driven image changes improve conversion.
Claude's automatic cross-campaign keyword aggregation eliminates a specific manual-analysis skill (building a pivot table) from the workflow entirely, rather than just speeding up the same manual steps.
Claude applied a minimum-order threshold to the "most efficient keywords" list on its own initiative, filtering out a keyword that looks efficient (e.g., 2% ACoS) purely because it only had one order — a judgment call the presenter says a human analyst would otherwise have to make manually.
The recommended way to build a reliable custom AI tool is an explicit iteration loop — simplest prompt first, then add one specific insight category at a time (shopper intent, layout, branding), then get team feedback — rather than attempting to specify the perfect prompt upfront.
Turning a one-off analysis into a saved "skill" shifts the unit of value from "a person who knows how to run the analysis" to "a tool anyone on the team can run in minutes," compressing what he describes as roughly an hour of manual work into a couple of minutes.
«if you're watching this now and you do this right, you'll get a bigger Amazon payday than you would have otherwise.»
— 00:11
«We call it the simplest possible attempt, the SPA.»
— 00:25
«That's why a pivot table is needed, but you literally don't even need to know how to do that anymore because the AI is doing it for you.»
— 05:38
«Save this as a skill.»
— 06:51
«utilizing and controlling AI is now the most important skill a human being on Earth can have and it will be for the rest of our lives»
— 04:10
«And then most importantly, shopper intent insights.»
— 08:25
«This is literally like having a high-level Amazon PBC expert at your fingertips and it's free.»
— 09:50
Reception
Mostly warm, appreciative comments praising the Claude/Amazon PPC content, tempered by a few frustrated viewers who had trouble accessing the promised skill/report or felt misled.
A concrete, replicable workflow (from Amazon report export to prompt to saved Claude skill) that clearly demonstrates iterative skill-building in practice, though the "free" and "irrelevant within months" framing function as marketing hooks around the technical content rather than neutral description.
11:10