Refresh 50 Stale Product Pages, Lift a Client's Revenue 38% in 90 Days (Claude + Sheets)

The client was a mid-size DTC brand with 200 product pages that hadn't been touched in 18 months. Their organic traffic was bleeding — every quarter down another 6-8% — and the merchandising team was too busy shipping new SKUs to look back at the old ones. By month three of our engagement, we had 50 of those pages rewritten and watched revenue from the refreshed set climb 38% over the 90 days that followed. Here's exactly how we got there, and the spreadsheet pipeline that made the rewrite part actually scalable.

The audit that found the rot

Before writing a single new line of copy, I pulled all 200 product URLs into a Google Sheet and ran a four-signal audit. Each signal was binary, scored 0 or 1, so the math stayed simple:

• Last update date. Page-modified header or visible changelog. Anything older than 12 months scored 1 (stale).
• Outdated product facts. Specs, prices, dimensions, materials, or "compatible with X" claims that no longer matched the actual product. I caught these by spot-checking 20 pages manually first, then wrote regex patterns to flag likely offenders across the rest.
• Dead or wrong imagery. 404s, generic placeholders, lifestyle shots from a discontinued line, or images smaller than 600px wide.
• Thin copy. Under 80 words of body copy, or copy that was obviously a manufacturer-feed re-paste with no benefit framing.

Each page got a 0-4 score. The breakdown: 31 pages scored 4, 49 scored 3, 68 scored 2, 52 scored 0-1. The top 80 — call it the "rewrite first" pile — covered 73% of total traffic to the audited set and 81% of revenue. We rewrote 50 of those (the rest got a lighter pass: title tag + meta description + image swap only).

If you've ever wondered why "refresh old content" advice can feel underwhelming in practice, it's usually because the writer is doing a one-off rewrite on five pages and never finding the leverage. The trick is to make the audit cheap and the triage automatic. I now run this exact four-signal audit on every e-commerce engagement in week one.

The Claude + Sheets pipeline

The hard part is not the rewrite. The hard part is doing 50 rewrites without burning a week on them or letting quality drift by page 30. I built a Sheets pipeline with three tabs and a single Claude integration that handled the bulk work.

Tab 1 — Inputs. For each of the 50 URLs, I had columns for: current H1, current meta description, product specs (pulled from a separate feed), the actual product URL, the target keyword, and a 2-3 sentence "what's wrong with this page today" note. The last column was the only one I wrote by hand — and even that was templated. ("Outdated price: $X → $Y. Spec wrong: was 304 stainless, now 316. Image is from 2022 line. Body copy is 47 words, no benefit framing.")

Tab 2 — Claude outputs. A single formula per row:

=CLAUDE("You are a senior DTC copywriter. Rewrite the product page for "&B2&" using these specs: "&C2&". Current H1: "&D2&". Current meta: "&E2&". Issues: "&F2&". Output: 1) a 60-char SEO title, 2) a 155-char meta description, 3) a 90-110 word benefit-led body opening that leads with the use case, not the spec sheet, 4) five 8-12 word bullet points, benefit-first. Voice: confident, plain, no hype words like 'revolutionary' or 'ultimate.' Do not invent features not present in the spec feed.", A2:G2)

That single formula, dragged down 50 rows, produced all 50 rewrites in about 14 minutes. The total Claude spend for the batch was under $9.

Tab 3 — QA. A second pass ran Claude as a reviewer against my own output. For each row it answered: "Does the rewrite introduce any feature not in the spec feed? Is the meta description under 155 characters? Does the body open with a use case or a spec? Are any banned words used?" Anything that failed got a red flag, and I personally rewrote those 9 pages. The other 41 shipped with my edit pass — usually tightening one sentence or fixing a price.

What we shipped and what we measured

We pushed the 50 rewrites live over a Tuesday-Wednesday in week 9 of the engagement. The merchandising team handled the actual CMS work; my role ended at "approved copy in a Google Sheet with green QA cells."

I compared three windows — 30 days pre-refresh, 30 days post, and the matched 30-day window from the year before (to strip seasonality). Across the 50 refreshed pages:

• Organic sessions: +42% post-refresh vs. pre-refresh, +28% vs. the year-ago window.
• Average ranking position for the target keyword cluster: moved from 14.3 to 8.1. Eight pages that were sitting on page 2 jumped to page 1.
• Revenue attributable to the refreshed set: +38% post-refresh vs. pre-refresh, +22% vs. year-ago. This was the number the client cared about.
• Bounce rate on the refreshed set: down 11 percentage points. The copy rewrite was pulling more qualified clicks, not just more clicks.

The non-refreshed 150 pages showed almost no movement in the same window, which is the part of the result that actually convinced the client. It wasn't a site-wide algo shift or a traffic distribution quirk — it was the pages.

Three things I'd do differently next time

First, I'd run the audit on the entire catalog from day one, not the 200 the client flagged. The other 800 pages probably have the same rot; the 200 was just where they were already losing traffic visibly. Second, I'd integrate the price/spec feed into the Sheet directly with an IMPORTDATA() formula instead of copy-pasting, because we lost two days to a price mismatch that wasn't caught until launch. Third, I'd add a "rewrite confidence" column in Tab 1 — for products where the spec feed was thin, I'd mark them and do those by hand. I burned three Claude outputs on a product where the feed only had the SKU and product name; the result was generic, and I rewrote it from scratch anyway.

The part nobody talks about

The reason most "content refresh" projects stall isn't the writing. It's the audit and the triage. Once you have a 0-4 score per page and a short list of 50 with notes, the rewrite is the easy part. The audit is the bottleneck, and it's the part every team I've watched tries to skip. Don't skip it. Spend three days on a four-signal audit, and you'll write less copy, ship less, and lift more revenue than the team that "refreshed" 200 pages by hand over six weeks and produced mush.

The 38% lift wasn't magic. It was a clean audit, a tight triage, and one Claude formula dragged down 50 rows. The repeatable thing isn't the AI. The repeatable thing is the audit.