Harrow & Vane: organic collapse, brand bidding and Performance Max
Harrow & Vane is a UK garden retailer with 8,400 live SKUs and £26M of online revenue. After the March core update, organic sessions fell 34% and organic revenue 29% year on year, and neither has recovered. 62% of organic revenue lands on 1,900 category and product pages; 1,100 of those carry fewer than 60 words of unique copy, the rest being manufacturer feed text. The site also exposes 4,300 size and colour variant URLs, each self-canonicalised.
Paid search spent £840k last year. 31% went to brand terms at a reported ROAS of 12.6. A Performance Max campaign launched in January now absorbs 47% of non-brand spend, reports ROAS 5.2, and 38% of its attributed conversions are on queries containing the brand name. Blended marketing efficiency ratio is 3.4 against a floor of 3.0. The head of ecommerce wants to switch off brand bidding entirely and push the budget into Performance Max.
1. How would you diagnose the organic decline, and what would you do in the first 90 days?
Read the worked answer
Do not attribute the fall to the core update and stop there. A 34% drop concentrated in pages with feed-derived copy points to site quality and duplication as much as to the algorithm.
Diagnose in this order:
- Segment the drop by template - category, product, variant, blog. If variant and near-duplicate URLs fell hardest, the problem is duplication.
- Separate ranking loss from click loss. Falling impressions means positions were lost. Stable impressions with falling clicks points to SERP features: shopping units, AI Overviews, local packs.
- Check indexation and crawl: how many of the 8,400 products plus 4,300 variants are indexed, and what share of crawl budget is wasted on faceted and variant URLs.
- Compare top linked URLs against top traffic URLs. Feed-derived pages rarely earn links, which limits recovery.
90-day plan.
Weeks 1-4: consolidate variants onto a canonical parent page with selectable options, noindex internal search and low-value facets, rebuild sitemaps, and redirect internal links to the parent. This is cheap and removes the duplication signal.
Weeks 4-8: rewrite the roughly 200 pages that carry most organic revenue, adding genuine buying content - soil type, planting season, hardiness, size comparisons, care guidance. Depth on the pages that already convert beats breadth.
Weeks 8-12: merge and redirect thin categories that cannibalise each other, then measure by template.
Set expectations. Recovery is not guaranteed and is rarely quick. Treat organic as a six-to-nine month rebuild, track template-level trajectory at 8 and 16 weeks, and protect revenue with paid in the meantime rather than promising a date.
2. Evaluate the proposal to switch off brand bidding and move the budget into Performance Max. What would you do instead?
Read the worked answer
The proposal is understandable and, as stated, wrong.
A brand ROAS of 12.6 is inflated because brand queries are mostly people who already decided. Most of that revenue would arrive without the ad. That does not make the ad worthless - it means the correct question is what happens to total revenue and total profit when brand ads are absent. Three things move at once: competitors and marketplaces absorb some clicks, especially on mobile where the top slot is worth more; direct and organic pick up the rest; and the cheap brand clicks currently flatter blended ROAS, so removing them makes the non-brand number look worse, not better.
Test it rather than debate it. Pause brand in 30-40 matched regions for four to six weeks and compare total revenue and new-customer volume, not paid search revenue. Keep brand live in control regions. If total revenue holds and new customers hold, there is a defensible case for a defensive minimum: a handful of exact brand terms plus a small competitor set.
Fix the Performance Max measurement first. 38% of PMax conversions on brand queries means PMax is claiming demand it did not create and obscuring the non-brand picture. Apply account-level brand exclusions where the platform offers them, move brand to an exact-match campaign, and report new-customer acquisition separately from returning. Then compare PMax non-brand CPA against the search campaigns it is absorbing. If it cannot beat equivalent non-brand search CPA once brand queries are stripped out, that 47% shift is relabelling, not improvement.
Sequence: brand exclusions and reporting fix this month; geo pause test next month; reallocation only after both.