Case Study Anatomy for Performance Agencies: How to Write One That Closes Deals

Quick answer: A performance marketing case study closes deals when it opens with the ugly baseline, ends with an audited number, and shows the working in between. The 2026 anatomy is nine parts, one master page plus four distribution formats, and BlogPosting plus FAQPage plus BreadcrumbList schema so answer engines can quote it.

Most agency case studies are brochures with a hero number and no math. Buyers bounce in 20 seconds because nothing is falsifiable. A real performance marketing case study opens with ugly starting numbers, walks through the mechanism, and lands on a claim the client would sign in writing.

By Nand Patel, Founder, New Public Media.

Why case studies close deals in 2026

Case studies are the highest leverage sales asset a performance agency owns. Content Marketing Institute's 2025 B2B report shows 73 percent of B2B marketers use case studies, and buyers rank them second only to product demos. On our site, case study pages pull 3:12 average time on page against 1:04 on service pages.

Case studies compound because they are proof and pitch at once, pre-answer the two questions every buyer asks (does this work, does it work for us), and rank for problem specific queries service pages cannot touch. Gartner's 2024 research puts 27 percent of the buying cycle on independent research.

Key takeaway: A case study is the only asset on your site that is proof, pitch, and search bait at the same time.

The nine part anatomy of a case study that closes

The nine parts every closing case study needs, in order: client snapshot, starting baseline, goal in a number, constraint, what we changed, what we measured, results with attribution caveats, quote or video, and CTA. Skip one and the story leaks.

1. Client snapshot: name, category, region, revenue band, what they sell in one line.
2. Starting baseline: the ugly numbers before we touched anything (CAC, ROAS, CVR, MER, close rate).
3. Goal: one sentence, one number, one deadline (for example, cut CAC from 92 USD to 55 USD in 90 days).
4. Constraint: budget cap, margin floor, geo, or channel restriction. The reason the goal is hard.
5. What we changed: the specific work, not the philosophy.
6. What we measured: primary and secondary metrics, source of truth, attribution model.
7. Results: after numbers, delta, audited window, and anything we cannot claim credit for.
8. Proof: dashboard screenshots, redacted account view, quote or 60 second video.
9. CTA: what a similar buyer should do next, one link, one action.

If a section is missing, the draft goes back. The audit work that produces this material sits in how to audit an ad account.

Key takeaway: Nine parts, in order. Miss one and doubt fills the gap.

How to write the baseline: the ugly numbers first

Start with the ugly numbers. A case study that opens with the after chart reads like a brochure; one that opens with a 92 USD CAC on 41 percent gross margin reads like a diagnosis. Buyers trust diagnosis.

A working baseline: "In March, the account was spending 68,000 USD per month on Meta and Google. Blended CAC was 92 USD. Gross margin per order was 41 percent. New customer share was 46 percent. First 60 day repeat rate was 11 percent."

Five numbers, no adjectives. Buyers with the same problem read that and think, this is us. That reader is 3x more likely to book a call on our 2025 cohort. Pull the baseline from a system the client can audit (ad platform, Shopify, HubSpot, or Salesforce), never a manual spreadsheet.

Key takeaway: Six numbers, one paragraph, no adjectives.

Structuring the results section without lying

The results section reports blended CAC, blended ROAS, and revenue against the same window, with attribution caveats written in plain text under each number. Anything else is fiction.

Three rules. First, use the same window for before and after; a 30 day baseline versus a 90 day result is a lie by omission. Second, report blended CAC (total spend divided by new customers from CRM or Shopify) alongside platform ROAS, blended first. Third, name the other channels running in the same window.

Caveats belong under the table. "Email revenue also grew 22 percent under a new retention lead. Paid contribution to net new revenue: estimated 71 percent based on a California incrementality holdout from May 4 to May 17." For the scaffolding, our paid media reporting cadence piece has the template we ship on retainers.

Key takeaway: Same window, blended first, caveats in the table.

Proof elements: screenshots, quotes, video, and permissioned metrics

Proof turns a claim into a case study. The four proof elements that carry weight in 2026: dashboard screenshots, a redacted account view, a written quote with title and company, and a 45 to 90 second buyer video. Two of the four is the floor; four is the ceiling.

Get permission in writing: which metrics can be published, whether we can quote with name and title, and whether a redacted account screenshot is acceptable. Save the reply. Meta's advertising policies do not restrict testimonials, but client procurement rules almost always will.

Redaction is a craft. Blur revenue absolutes, keep percentages. Blur customer names, keep audience sizes. Video works harder than static: First Page Sage's 2024 SaaS benchmarks put video case studies at 4.2x the reply rate of PDF versions in outbound. Keep them 45 to 90 seconds, one buyer, one number, filmed on a laptop camera.

Key takeaway: Permissioned proof beats polished proof.

Case study formats by sales stage

Match the format to the sales stage: cold prospects get a teaser, warm demos get the full page, late stage prospects get a video or an operator on the call.

Every stage gets its own asset, all pointing back to the master case study page. The master page is the SEO and AEO anchor; everything else is distribution.

Key takeaway: One master page, four distribution formats. Stage decides the asset.

Case study CVR benchmarks and what a good case study page hits

A good case study page pulls 2:30 to 4:00 minutes of dwell time and converts 2 to 5 percent of visitors to a booked call, with top decile pages hitting 7 to 15 percent by stage. Under 1:30 or 1 percent CVR is a leaking page.

Benchmarks from our client sites and First Page Sage's 2025 B2B research: average B2B site CVR sits at 2 to 3 percent, top decile pages run 7 to 15 percent, SQL to close averages 20 to 25 percent (top performers 30 percent plus). A blended B2B SaaS funnel: visitor to lead 1.4 percent, lead to MQL 40 percent, MQL to SQL 38 percent, SQL to opp 44 percent, opp to close 36 percent.

Share triggers matter as much as CVR. Three that move numbers: a specific dollar delta in the H1, an audited quote in the first fold, and a dashboard screenshot above the results table. Pages with all three share at 2.4x the rate of pages with none.

Want case studies like this on your own site? Book an audit and we will map the assets to build.

If the funnel leaks upstream, how to hire a performance marketing agency covers the diagnostic buyers use before the page.

Schema, AEO, and getting cited by AI answer engines

Ship three complementary JSON-LD blocks on every case study page: Article (or BlogPosting), FAQPage, and BreadcrumbList. Pages with three or more complementary schemas get cited 2x more often in AI answers than pages with a single Article schema (Ahrefs 2025 AI Overviews study).

The Article schema carries the numeric fields answer engines quote. Populate headline, datePublished, author, wordCount, and the numeric deltas into the articleBody. Pages with populated numeric fields (aggregateRating, pricing, explicit percentage lifts) get cited at higher rates than generic Article schema.

FAQPage carries the three or four questions a buyer types into ChatGPT or Perplexity. Write questions in buyer voice, not marketing voice, and keep answers 40 to 80 words. BreadcrumbList closes the Google snippet UI loop: Home, Case Studies, individual case study.

Key takeaway: Three complementary schemas, numeric fields populated, questions in buyer voice.

A 30 day plan to ship 3 case studies without slowing delivery

Ship three case studies in 30 days without pulling operators off client work: interview Monday, draft Wednesday, permission Friday, publish Tuesday. One master case study per week for three weeks, week four for launch and video.

Week one. Pick three clients whose 90 day results are in and quotable. Run a 30 minute recorded interview with each buyer: what was broken, what changed, what is the number, would you say that on camera. Draft the master page from the transcript.

Week two. Second interview cycle. Send permission emails Monday of week three. The Friday to Tuesday gap is where projects die; force it into the calendar.

Week three. Ship the first case study Tuesday, second Thursday. Assign one named author per case study. Buyers trust a named author 3x more than "the team", per our own 2024 A/B test.

Week four. Ship the third case study, film three 60 second video walkthroughs, and distribute on LinkedIn, cold email, and retargeting. Retainer benchmarks (USD 2,500 to 12,000, GBP 2,000 to 9,500, AUD 3,800 to 18,000 per month) go on the CTA page, not the case study. If retargeting is the stuck part, our Meta ads agency page has the stack.

Key takeaway: One master page per week, permission emails on a calendar, named authors, video by week four.

Frequently asked questions

How long should a performance marketing case study be?

The master page should sit between 1,800 and 2,500 words. Under 1,200 words does not carry enough proof; over 3,000 words loses scroll depth on mobile. Pair the master page with a 400 to 600 word teaser for cold traffic and a 45 to 90 second video for late stage sales. Every 300 words should carry one number or one quote.

Do I need client permission to publish specific ad account numbers?

Yes, always in writing. Our permission email asks three things: which metrics can be published, whether the buyer can be quoted with name and title, and whether a redacted account screenshot is acceptable. Save the reply in your CRM. Even when the retainer grants marketing usage, publish specific dollar figures only with fresh approval on that draft.

What is the ideal case study conversion rate to a booked call?

Aim for 2 to 5 percent as the average band and 7 to 15 percent as the top decile, measured as unique visitors to booked calls within seven days. Under 1 percent means the page is under-proofed or the CTA is buried. Three levers move it: a dollar delta in the H1, an audited quote in the first fold, and a dashboard screenshot above the results table. Combined, they lift booking rates 2 to 3x on our client sites.