Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

SEO › Expert

Capstone: An Evidence-Led SEO Programme for a Classic Car Parts Retailer

The closing project of the expert SEO level. You write a year's programme for an invented classic car parts retailer whose crawlers are lost in its search pages, whose developers ship templates without requirements, whose last title rewrite was judged on a before-and-after chart and whose board wants a forecast for a new section of fitting guides. The programme sets the crawl evidence and parameter rules, the requirement process, how changes are measured and tested, a ranged forecast, the test the guides must pass and where they live.
Capstone projectExpert levelModule 3: Forecasts and scaled page programmes11 minUpdated

Brief

Grevalpe Motor Spares is an invented online retailer of parts for classic cars. Its catalogue runs to many thousands of part and category pages, and its search tool lets visitors narrow by make, model and year. It plans a section of fitting guides by car model, on a documentation platform separate from the shop.

Four things worry the new head of digital. The logs suggest crawlers spend most of their requests on sorted and filtered search pages while new parts wait. Developers ship template changes with no SEO requirement, and problems surface in the next audit. Last autumn's title rewrite was declared a success on a before-and-after chart. And the board wants a forecast for the fitting guides before it approves them, while the agency has offered one figure.

The advanced SEO level planned migrations, set indexing tiers and built an operating rhythm. This programme asks for evidence at every step: what crawlers did, what developers will build, what a change actually did and what a new section can honestly be expected to deliver.

Inputs and constraints

Illustrative example — not a real result

Grevalpe Motor Spares: a month of crawler requests and the parameter list

Requests with a Google crawler user agent: 2,240,000. Of these, 240,000 came from addresses that failed verification. Verified: 2,000,000. Search Console Crawl Stats for the month: 2,060,000.

TemplateURLsVerified requestsShare
Site search results with sort or filter parametersUnbounded1,020,00051%
Category pages with sort or filter parametersUnbounded380,00019%
Part pages46,000340,00017%
Clean category pages2,300120,0006%
Redirects, old part numbers through two hops80,0004%
Not found, discontinued parts still linked40,0002%
Articles and other pages20,0001%

14,500 part pages were not requested in the month; 11,800 of them were added in the last three months and are linked only from the fifth page of category pagination onwards.

ParameterFound inDistinct URLsWhat it does
sortCrawl and logs180,000Reorders search and category lists
make, model, yearCrawl and logs410,000Narrow search and category lists
pageCrawl and logs60,000Paginates
qCrawl and logs250,000Runs the site search
vatCrawl46,000Shows part prices with or without VAT
utm_source, utm_campaignCrawl9,000Tags on homepage banner links to categories
sessidLogs3,000Session identifier from the old checkout
refLogs6,000Tag on links from an owners' club forum
Illustrative example — not a real result

Grevalpe Motor Spares: releases, last autumn's title rewrite and search history

  • Releases every second Tuesday. Planned: the part page template gains a fitment table loaded by script after the page renders (late January); the search template is redesigned with filters as links (March); the fitting guides launch on the documentation vendor's platform at a guides subdomain (May). Code freeze for the second half of December. Seasonal index for non-brand clicks: 1.4 in April, 0.6 in December.
  • Last September, titles on 6,000 brake and suspension part pages were rewritten to lead with the car model. The agency's chart: brake and suspension clicks in October 18% above August.
Group, six-week windows from the crawl date, same weekdaysClicks beforeClicks afterAverage position
Brake and suspension, 6,000 pages, changed58,80061,2008.1 to 7.4
Electrical and engine, 8,000 pages, unchanged, same template78,40073,5008.6 to 8.7
  • Weekly ratio of changed to unchanged clicks: 0.71 to 0.79 across the twelve weeks before; 0.80 to 0.86 across the six weeks after. A "fits your car" link block went live on brake and suspension pages only, in the third week after the change. A sitemap fix in late September reached all part pages. 31 of 40 sampled results showed the new title.
Illustrative example — not a real result

Grevalpe Motor Spares: the fitting guide proposal and the site's own data

The agency proposes a guide for every part category on every model, 180 models × 40 categories = 7,200 guides, with steps generated from the catalogue's fitment data, launched in May at the vendor's default subdomain, and forecasts "120,000 extra visits a year" from a published click-through curve.

GroupCombinationsKeyword tool searches a monthCombinations with workshop fitting notes and photosSearches a month for those combinations
Top 20 models × 12 most-replaced parts24072,00015054,000
Top 20 models × other 28 parts5604,2004080
Other 160 models × 12 most-replaced parts1,92019,200302,400
Other 160 models × other 28 parts4,4801,10000
  • Results pages for "[model] [part] replacement" queries show specialist suppliers' guides, club pages, forum threads and videos.
  • Existing how-to articles: 40, with 168,000 non-brand clicks over the past year. Articles left unrefreshed lost 5% of their clicks against the year before. Model and part fitting queries already bring them about 400 clicks a month.
  • Click-through Grevalpe measured on its non-brand queries over the past half-year, by position band: top three 18%, fourth to sixth 7%, seventh to tenth 3%. For queries with a top-three Grevalpe result, keyword tool volumes come out at 1.2 times Search Console's impressions.
  • How-to articles from the past two years settled into a stable position three to six months after going live, four at the median.
  • Each organic click yields 0.9 sessions. How-to article sessions convert to orders at 0.8%, ranging from 0.6% to 1.0% quarter by quarter over the past year. Average order £85; gross margin 40%. The workshop can write 10 fitting notes with photos a month.
  • A reverse proxy can serve the vendor's pages under a fitting guides path on the main domain: a sprint of work and testing, with the vendor's canonical tags rewritten from the subdomain host.

Constraints:

  • The team is the head of digital, an SEO lead, a part-time analyst, the workshop team and developers on the fortnightly release train.
  • Log extracts contain IP addresses: handle them as the log file analysis tutorial sets out, agreed with the data protection owner.
  • The board will not accept a single figure; every forecast needs ranges and named assumptions, and no uplift may be invented.
  • The guide launch date can move; the documentation platform cannot be replaced this year.

Task

Write one programme document with these sections, in this order.

  1. Crawl evidence and the parameter register. Verify the crawler requests, read the month by template and size the waste and the neglected part pages, following the log file analysis tutorial. Classify every parameter and give each class a rule, a mechanism and an owner in a register, with the rule for parameters not yet invented, following the URL parameter audit. Say which model filter pages keep a place in the index under the tiers of the faceted navigation guide, and explain what the evidence means for Crawl Budget on a catalogue this size.
  2. Requirements before release. Write the requirements for the fitment table and the search redesign with the guide to writing SEO requirements for developers: URL rules, indexing and canonicals, what the rendered HTML must contain, metadata with fallbacks, markup, acceptance criteria with test URLs, and who checks what. Set the rule that no template ships without a signed requirement, with the SEO release checks after each release and the technical SEO audit checklist at each quarterly audit.
  3. Measuring changes. Re-judge last autumn's title rewrite with the tutorial on measuring an SEO change and write the result in one honest sentence. Design the next title change with the SEO title tag test and a fair test of the "fits your car" block with the internal linking test, each written in the experiment hypothesis template with its Holdout Group, metric, guardrails and decision rule.
  4. The forecast. Forecast the fitting guides with the SEO forecasting guide: the Baseline for existing articles, calibrated demand, Grevalpe's own click-through, net of the clicks articles already earn, timing from its ramp-up record and the season, low, central and high cases, and orders and margin. Answer the agency's figure, and design the monthly table that sets forecast against actual.
  5. The templated pages test. Put every combination through the programmatic SEO tests: distinct need, different substance, its state, the template, the pilot and tranches, sitemap and links, and how the workshop's notes expand the set.
  6. Where the section lives. Choose a subfolder or a subdomain with the subfolder and subdomain comparison, weighing linking, ownership, hosting and migration risk, with the reasons written down.
  7. Calendar and order. Place every change on the release calendar so each can be read on its own, outside the freeze, and record changes to templates, parameters and tracking in a measurement change log.
  8. Board summary. One page: what changes, in what order, the forecast range and its weakest assumption, and what would make you stop or roll back.

Evaluation criteria

CriterionWhat a strong answer showsCommon gap
Crawl evidenceReads logs by template, sizes crawl waste and neglected pages, and ties every parameter to a rule in a register with an ownerRobots rules written from a crawl tool alone
RequirementsWrites testable requirements for each template change before release, covering URLs, indexing, rendered content, metadata, markup and checksFixes raised as tickets after launch
Measuring changesJudges past and future changes against comparison pages and matched windows, and runs page-group tests with guardrailsBefore-and-after charts presented as proof
ForecastBuilds a baseline from the site's own history, sizes opportunity with its own click-through, gives ranges and tracks forecast against actualA single number from generic click curves
Templated pagesPasses the guide section through the demand and unique-data tests, stages its rollout and sets indexing and internal linksA page for every model whether or not anyone searches for it
Where it livesChooses subfolder or subdomain on linking, ownership, hosting and migration risk, with the reasons written downA subdomain chosen because the platform defaults to it

A strong programme is mostly a matter of order. The crawl evidence comes first, because the fitment table and the search redesign either fix the waste or entrench it, and their requirements should carry the register's rules. The fitment table is the riskiest release: content loaded by script after rendering is exactly what the requirement must pin down in the server response. The title rewrite is reported as a range with both caveats, the position shift and the link block in the window, rather than as a win or a failure, and the next change is randomised.

The guide forecast earns trust by what it refuses. A strong answer drops the generated steps, because a fitment list with a model name swapped in is the Doorway Page and scaled content pattern the programmatic SEO guide describes, publishes only the combinations with workshop notes and real demand, and lets the notes backlog decide each tranche. It notices that a May launch arrives just after the April peak, ramps while demand falls towards the December low and shows little until the next spring, and either moves the pilot earlier or says what that costs. And it puts the guides in a subfolder through the proxy, because they exist to send readers to the parts that fit.

Check your arithmetic against the reference figures after working them out yourself.

Illustrative example — not a real result

Reference figures for self-checking

  • Verification removed 240,000 ÷ 2,240,000 ≈ 11% of claimed crawler requests; verified requests sit close to Crawl Stats.
  • Waste: 51% + 19% + 4% + 2% = 76% of verified requests went to parameter URLs, redirect hops and dead links; part pages received 17%.
  • Requests per URL in the month: part pages 340,000 ÷ 46,000 ≈ 7.4; clean categories 120,000 ÷ 2,300 ≈ 52.
  • Neglect: 14,500 ÷ 46,000 ≈ 32% of part pages unrequested, most of them new and deep in pagination.
  • Title rewrite: ratio 58,800 ÷ 78,400 = 0.75 before and 61,200 ÷ 73,500 ≈ 0.83 after, about 11% higher. Against the pre-change range, the effect lies roughly between 0.83 ÷ 0.79 ≈ +5% and 0.83 ÷ 0.71 ≈ +17%, with position moving and the link block inside the window. The agency's 18% compares different months, not matched windows.
  • Indexable guides: 150 + 30 = 180, with demand of (54,000 + 2,400) ÷ 1.2 = 47,000 a month. The 40 noted guides with almost no searches suit noindexed pages linked from their parts; the other 90 top-model, most-replaced combinations wait for notes, nine months of workshop capacity.
  • Settled clicks a month, net of the 400 articles already earn: low 47,000 × 3% − 400 = 1,010; central 47,000 × 7% − 400 = 2,890; high 47,000 × 18% − 400 = 8,060.
  • Orders a month when settled: low 1,010 × 0.9 × 0.6% ≈ 5.5 (about £464 of sales); central 2,890 × 0.9 × 0.8% ≈ 20.8 (about £1,769 of sales, £707 of margin); high 8,060 × 0.9 × 0.8% ≈ 58 (about £4,933 of sales).
  • The agency's 120,000 visits a year is 10,000 a month, above even the high case's net clicks.
  • Baseline, central: 168,000 ÷ 12 × 0.95 ≈ 13,300 monthly clicks expected from the existing articles.

Submission checklist

0/12

When you have finished, compare your crawl section with the job board log and parameter exercise, your measurement section with the tyre retailer title exercise and your guide plan with the breed page programme exercise. The two checkpoints, on crawl evidence and on measuring and testing changes, show where to reread.

Pages this lesson applies

Templates and checklists used

Terms used on this page

Also related

Related