Search Engine Optimization Case Studies
Marketing5 Case StudiesUpdated Sep 2026
Search engine optimization sits where editorial judgment meets measurement, and these five cases show how teams make decisions when budgets are fixed, positioning drifts, traffic looks healthy but revenue does not, a complaint spreads across results pages, or a content library has grown faster than anyone reviewed it. Every case uses the same four lenses: objectives, evidence, constraints, and stakeholder impact. None of them ends with a tidy answer. The aim is to see how a team narrows its options, argues about tradeoffs, and decides what to do first when the deadline is close and the data is partial.
Case Study 1
Splitting a fixed launch budget across three channels
Wrenfield Home Goods, an online retailer of linen kitchen goods with 40 staff and a catalog of 1,300 products
The situation
Wrenfield had 62,000 to launch a spring linen collection over ten weeks. Priya, the marketing lead, could put the money into paid search on category and brand terms, into organic content aimed at long questions about linen care and washing, or into paid social toward a lookalike audience built from recent buyers. The two earlier launches gave her some history, but the landing pages had changed between them and the tagging was inconsistent, so comparisons were rough. Organic traffic to category pages had been flat for several months while two competitors pushed new guides into the top results. The team had to commit most of the budget before shipping began.
The challenge
The question was not which channel is best in general, since the team had no reliable evidence for that. It was which mix could reach the collection's buyers, cost less than the target of 38 per order, and produce numbers that finance would accept at the close. Merchandising wanted the collection visible on search results during the first two weeks. The warehouse could ship only about 180 orders a week during the launch, so a spike from paid social would create backorders and complaints.
Analysis
The business goal was a launch that paid back within the quarter, with cost per order under 38. Teams grounded in Search Engine Optimization usually start by separating what each channel is supposed to do: paid search captures people who already want linen, organic content builds a base that keeps earning after the budget ends, and paid social creates demand that often converts later and through other routes. The evidence was thin. Last season's data mixed branded and generic clicks, and the social audience had never been measured against a holdout group. That meant any split would be a judgment informed by partial numbers, and the plan needed to say so openly.
Operational limits shape the plan as much as marketing logic does. The warehouse ceiling of about 180 orders a week means paid demand has to be paced, which argues for search campaigns with daily caps and for social spend that ramps only after stock is confirmed. Finance cares about the payback window and wants a holdout test so that the lift can be separated from seasonal demand. Merchandising's concern is visibility, and the organic guides are the only channel that can keep the collection on results pages once the budget is gone. Customer service absorbs any gap between promised and actual dates, so messaging has to match warehouse capacity.
Actions taken
- Set paid search daily caps to match the warehouse ceiling of about 180 orders a week, and reserve the final two weeks of budget for whichever channel shows the lowest cost per order.
- Hold back about 10 percent of the social audience as a randomized holdout so finance can see incremental orders rather than credited ones.
- Publish two care guides in the first week and link them from category pages, so organic results keep building after paid spend stops.
- Give customer service a shared sheet of delivery windows that the warehouse updates each morning, and tie ad copy to those windows rather than to fixed promises.
Results
At close, paid search had returned the lowest cost per order at 31, under the target of 38, and it carried most early revenue. Paid social came in higher at 52 per order on credited conversions, but the holdout showed a fifth of its credited sales would have happened anyway, bringing its real cost nearer the target. Organic guides were slower but kept drawing searches for the collection after paid spend ended. The team did not claim a clean winner. The report included the holdout result and the warehouse cap on the second-week push, and finance accepted it.
Questions for discussion
- Which of the three channels would you have cut first if the warehouse ceiling had been half as high, and what evidence would you need before cutting it?
- How would you adjust the holdout design if the social audience were too small to split without losing statistical reach?
- What would you tell merchandising if organic guides performed well after launch but produced no measurable sales within the payback window?
Case Study 2
Repositioning a premium printer that customers see as cheap
Halvard Print Works, a family-run maker of printed packaging and labels with 120 employees and two plants
The situation
Halvard had spent three years describing itself as a premium partner for small food and beverage brands, with careful color control, short runs, and design help. Customer interviews and a review of site search logs told a different story. Buyers searched for words like cheap labels, bulk stickers, and fast printing, and the reviews that mentioned price praised speed and cost rather than finish. Sales were steady but the premium tier had stopped growing, and the account team was discounting to win orders it should have been pricing for quality. The brand team had a proposal ready, and a board member wanted it in market before the next trade show.
The challenge
The intended positioning was craftsmanship and design partnership. The perceived one was volume printing for thin margins. The proposal had to close that gap without alienating the cost-conscious buyers who still filled most of the plants' schedules, and it had to be believed by the sales force, who had heard repositioning promises before. The director wanted evidence, not a tagline, before committing the site rebuild budget.
Analysis
The aim was to lift average order value in the premium tier by moving buyers who already searched for quality terms toward a clearer offer. Evidence came in three layers: search logs, reviews that separated price from finish, and win-loss notes from sales. Each layer had its own weakness. Search logs reflect visitors already on the site, reviews skew toward strong feelings, and sales notes carry a stake. The useful part was that they agreed on one thing: buyers valued fast turnaround as a proof of reliability, not as a sign of low quality. That suggested the new position should not abandon speed. It should explain it.
Constraints were tight. The site rebuild had to fit within an existing content management system that the agency would not replace, and the press schedule was fixed for the trade show. Sales staff would be the first to hear any claim, so a message that promised design partnership they could not deliver would damage trust quickly. Plant managers were worried about any shift in mix toward short runs that could slow the lines. A test could run only on the premium landing pages, with the standard price pages left unchanged to protect volume customers. The proposition had to be testable and still keep the plants full.
- 1Name buyer languagePull the search terms and review phrases buyers actually use.
- 2Draft three claimsTie each claim to a term, and drop any sales cannot repeat.
- 3Run a split testSix weeks on premium pages, with a control that keeps the old copy.
- 4Decide before resultsSet the threshold and volume floor before the test starts.
- 5Brief sales and plantsShare the approved claims and the volume effects before public launch.
Actions taken
- Write the new proposition as three claims, each tied to a search term buyers used, and drop any claim that sales could not repeat in a customer call.
- Run a split test on the premium category pages for six weeks, with speed framed as reliability on one variant and left out on the control.
- Track assisted conversions and search impressions for the target terms weekly, and set the decision threshold before the test starts rather than after.
- Brief the sales team on the claims before any public launch, and give them a one-page answer for the price objection that does not disparage budget work.
Results
The six-week test favored the reliability framing. Premium pages with the new claims converted at a rate about 14 percent higher than control, and the gain held for buyers who arrived from the quality search terms. The lift was smaller for generic price searches, which the team had expected. Sales used the one-page answer in about two thirds of price objections and reported fewer discount requests. The board approved the full site rebuild, but with a narrower scope than the original proposal, and the price pages were kept as they were.
Questions for discussion
- If the reliability claim had won with buyers but lost with sales, which group would you have trusted more to decide the final wording, and why?
- How would you decide whether a repositioning that lifts premium conversion is worth a small drop in volume pricing?
- What would change your test design if the premium landing pages received too little traffic to reach a clear result within the planned window?
Case Study 3
Diagnosing falling orders while sessions hold steady
Corvane Outdoor Supply, an online retailer of camping and trail gear with 260 staff and two warehouses
The situation
Corvane's site drew about the same number of sessions each week as it had for most of the year, yet orders had fallen steadily for about two months. Conversion rate had dropped from roughly 2.9 percent to 2.2 percent, and the drop was concentrated on mobile devices. Search brought the largest share of visits, with a strong mix of category and product queries. Analytics showed a new attribution model had been adopted recently, which credited more sales to last clicks from direct visits and fewer to organic search. The marketing lead, Tomas, had been asked to explain the decline before the next planning meeting, and the team had three weeks to answer.
The challenge
Traffic was not the problem, so the team could not simply buy more of it. The real question was where the journey broke. Possible causes ranged from a checkout change and slower mobile pages to content that no longer matched the questions buyers asked, or an attribution change that made organic look weaker than it was. Each cause implied a different fix, and the engineering queue could take only two requests that month.
Analysis
Restoring orders to the level the marketing plan assumed was the target, and it had to happen without more paid traffic, which already looked saturated. Evidence was uneven. Session and order data were reliable, but the attribution change had been applied without a backfill, so channel comparisons before and after it were not comparable. Heatmap data covered only the desktop site, and the mobile sample was small. Support tickets offered qualitative signals, mostly about shipping cost and size guidance. Before any experiment was prioritized, the team needed to separate changes in the measurement from changes in buyer behavior, because the two could look identical in a dashboard.
Limits on engineering time decided the order of work. Engineering could take two requests that month, and a checkout change would need a regression pass across payment methods, which pushed it toward the end of the window. Mobile pages shared a template with the wholesale portal, so any speed fix touched a second group of customers. Merchandising and the warehouse each had requests that competed for the same release window. Finance would judge the team on cost per order, so an attribution correction could change the story it told upward without changing any buyer's behavior. Each group needed to know what was measured and what was inferred.
- Days one to twoConfirm session and order data match, and freeze the attribution model.
- Day threeBackfill organic and direct credit so both periods read on one basis.
- Week oneBreak mobile checkout drop-off down by step and device version.
- Week twoShip the template fix and the shipping cost display test.
- Week threeCompare recovery against the backfilled baseline before choosing next work.
Actions taken
- Rebuild the channel comparison with a backfilled view of the old attribution model, so the two periods can be read on the same basis before any budget decision is made.
- Pull mobile checkout sessions for the past six weeks and compare each step's drop-off against desktop, starting with shipping cost display and address entry.
- Use one of the two engineering requests to fix the slowest mobile product template, and keep the wholesale portal in the test group so its performance is watched.
- Tag the top twenty support complaints by theme each week, and hold the category refresh until the shipping cost finding is confirmed or ruled out.
Results
The attribution backfill showed that organic search had been undercounted by a clear margin, though not enough to explain the full drop in orders. Mobile checkout showed the largest single loss at the shipping cost step, where the price appeared only after address entry for a portion of visitors. The template fix shortened mobile product pages, and the team saw a modest recovery in mobile conversion in the following weeks. Conversion did not return to its earlier level, so the team kept the shipping display change in the next experiment and reported the remaining gap as unresolved rather than as explained.
Questions for discussion
- If the backfill had shown organic search was not undercounted at all, which of the remaining explanations would you test first, and what would it cost to test it?
- How would you explain a partial recovery to a finance lead who wants a single cause for the decline?
- What evidence would convince you that a mobile template change, rather than a checkout display change, drove the lift?
Case Study 4
Managing search results during a public complaint thread
Fennmoor Kitchen Co., a direct-to-consumer cookware brand with about 90 staff and one main distribution center
The situation
A customer posted a short video showing a Fennmoor skillet with a coating that flaked after a few weeks of normal use. Within a day, the clip was shared by a small consumer forum and then by two larger accounts that focus on product reviews. Comments asked whether the batch was faulty, whether the company knew, and whether refunds were being denied. Fennmoor's branded search results started to show forum threads and an unofficial review page above its own product pages. The social team had a holding statement, but the customer service queue had grown to nearly four times its normal size, and no one had yet pulled the batch records.
The challenge
The company had to answer a factual question it did not yet have answered, while the public conversation ran faster than any investigation could. Leadership wanted a statement within the day. Legal wanted nothing said about the batch until the records were checked. Meanwhile the search results would keep shaping how new buyers judged the brand, and whatever was said on the company's own pages would be indexed and quoted.
Analysis
Protecting repeat purchases came first, and the defect question had to stay a defect question rather than become a brand crisis. Verified facts were scarce at the start. The company had sales records and a returns log, but the batch code was visible only in the video and was obscured. Customer service had 14 tickets that described similar flaking, which was a signal but not a count of affected units. The most useful evidence would come from retained samples and the coating supplier's certificate, both of which required a day or two to pull. Until then, the team could say what it had seen and what it was checking.
Every step was shaped by a limit. Customer service could not answer questions about batches until the warehouse confirmed lot numbers. Legal needed every public line checked against records, which meant the holding statement could not be upgraded until the sample results arrived. Retailers who sold the skillet had their own listings and their own customer contacts, so Fennmoor needed to brief them without appearing to shift blame. Buyers who already owned the pan needed a clear path to a replacement, and the people most affected were those who had bought in the past month and might not have seen any update.
Actions taken
- Pull retained samples and the coating supplier's certificate for the lot in the video, and state publicly only what those records confirm and what remains under review.
- Replace the holding statement with a short update that names the product, gives the lot check timeline, and offers a direct replacement path for any owner of the affected lot.
- Send retailers a one-page brief with the same facts as the public update, and ask them to route complaints to a single intake form rather than improvising answers.
- Publish a dedicated page for the product's care and coating questions, and optimize its title and headings for the branded queries now pulling in forum threads.
Results
The sample check came back in the third day and found the coating defect in a small share of one production run, which was narrower than the forum threads had suggested and broader than the company had hoped. Fennmoor posted the finding, offered replacements to owners of that run, and named the coating supplier's corrective step. Soon after, the branded results page carried the company's care page and the update above the forum threads, and ticket volume returned close to normal. Repeat purchases dipped slightly below plan, which the team read as a cost of the episode, not lasting damage.
Questions for discussion
- If the sample check had been inconclusive, how would you balance replacing units for owners against waiting for a result you could publish?
- Which stakeholder would you have briefed first, and what would you have said to them before the public update went live?
- What recovery indicator would you have set for the search results if the forum threads kept ranking after the care page was live?
Case Study 5
Auditing a credit union guide library with uneven results
Kestrel Ridge Credit Union, a member-owned lender with 310 staff, four branches, and a public library of about 1,100 guides
The situation
Kestrel Ridge's public content library had grown over several years, mostly through guides written by branch staff and product managers. Roughly 1,100 pages covered loans, savings, fraud, and retirement planning. A quarterly review showed a small group of pages drawing most organic visits, while many others had almost no traffic and some gave rates or eligibility details that no longer matched current products. Editorial standards were written down but applied unevenly, and approvals ran through three separate reviewers with no shared checklist. The digital lead, Anil, had been asked to set priorities for refreshes before the next budget round.
The challenge
The portfolio had too many pages to review one by one, and the team did not know which mix of topics, formats, and search intent was actually working. Some formats, such as calculators and short comparison tables, seemed to attract the right visitors, while long narrative guides were harder to read. Refresh work had to be ranked by business value rather than by who complained loudest.
Analysis
Members needed to reach the right product pages, calls about outdated rates had to fall, and compliance should not have to flag the library. The data had gaps. Analytics showed which pages drew organic visits, and the call center logged topics behind follow-up calls. The weakness was that search intent had never been tagged. A page could rank for a question that a member asked in a branch, yet the team had no record of whether the page answered it. Without that tagging, any ranking of refresh priorities would rest on traffic alone, which favors pages that were already popular.
Capacity was the real constraint. Three reviewers signed off on each page, and the average approval took long enough that a corrected rate could stay live for weeks. Branch staff relied on the guides when members walked in, so they wanted changes quickly. Compliance cared most about pages that quoted rates or eligibility, and it had no capacity to review the whole library. Product managers owned the numbers in their pages but not the wording, which meant accuracy and tone often came from different people. The governance design therefore had to cut review time for low-risk pages without weakening the check on the pages that could create liability.
- Audit the library Tag each page by topic, format, search intent, and current accuracy.
- Prioritize refreshes Rank by organic visits, call volume, and rate or eligibility risk.
- Approve by tier Route low-risk pages to one reviewer, and rate or eligibility pages to compliance.
- Retire or refresh Retire pages idle for two cycles and log replacements for branch staff.
Actions taken
- Tag every page with one topic, one format, and one search intent before any refresh is scheduled, using the existing analytics export as the starting list.
- Score refresh priority using organic visits, call volume on the topic, and rate-sensitive content, so that pages with money or eligibility risk move to the top.
- Create a single approval checklist with a two-tier path, where pages without rates or eligibility claims need one reviewer and anything quoting current terms goes to compliance.
- Set a quarterly review that retires pages with no traffic and no call topic for two cycles, and log each retirement so branch staff can point members to a replacement.
Results
After the first full cycle, the team had tagged most of the library and cut the approval path for roughly half of the pages in the refresh queue. Eleven pages with outdated rates were corrected within a single review period, and the call center reported fewer follow-up calls on those topics. Retirement removed a little over 200 pages, and organic visits to the remaining guides held steady, which the team took as a sign the library had been carrying dead weight. Compliance now reviews a shorter list each quarter, though the checklist needs a revision after two months.
Questions for discussion
- Which tag would you have been least willing to skip during the first audit, and what would you have accepted as a proxy if the call log were unavailable?
- How would you handle a high-traffic guide that compliance wants retired but branch staff say members rely on daily?
- What would you need to see in the second quarter to decide that the two-tier approval path was too loose for pages quoting current terms?
Learning Resources
Marketing CASE STUDIES
China, India, United States, Indonesia, Pakistan, Nigeria, Brazil, Bangladesh, Russia, Japan, Mexico, Philippines, Vietnam, Egypt, Turkey, Germany, United Kingdom, France, Italy, Canada, South Korea, Australia, Spain.