Skip to content
Get Ranked On TopIllustrative case studyCommerce SEO plan

A fictional business. The figures are a modelled scenario, not a client result.

RUANG RUMAH · Property listing portal, 180,000 listings · Jakarta

How a Jakarta property portal stopped watering dead listings and got its neighbourhoods crawled

+146%

organic enquiries, from the same crawl budget

Jakarta, Indonesia · 22 months · Home & real estate

Read the hillside as the portal: Googlebot’s daily visits are the water, each terrace a kind of page worth ranking. Water lost to dead listings and thin filters never reaches the paddies. Modelled data.

THE DRY SEASON

Plenty of water. Almost none of it reached the paddies.

Millions of thin filter pages and 41,300 expired listings soaked up three quarters of Google’s crawl, while new Jakarta neighbourhood pages waited weeks.

Ruang Rumah lists 180,000 homes, flats and boarding rooms across Jakarta. Every combination of area, price, bedrooms and sort order was a crawlable URL, so Google saw millions of near-identical pages. Listings that had sold months ago still answered as if they were live.

Googlebot came every day, about 240,000 requests of it. Three quarters ran off into pages nobody should land on, while the pages people actually search for, a neighbourhood like Tebet or Kelapa Gading, either did not exist or waited weeks for a first visit.

FIELD SURVEY · MONTH 0

  1. 41,300

    Expired listings still live

    Sold and let properties kept answering 200 OK with a full page, and stayed in the sitemap for months.

  2. 2.4M

    Filter URLs in the index

    Area × price × bedrooms × sort produced millions of URLs. About 2.4 million of them were indexed, most with a handful of listings each.

  3. 75%

    Crawl wasted

    Thirty days of server logs showed three requests in four going to expired listings, thin filters and sort or tracking parameters.

  4. 19

    Days to a first crawl

    The median new listing waited 19 days before Googlebot fetched it. In a market where good flats go in a week, that is too late.

  5. 0

    Neighbourhood pages

    Jakarta has 267 kelurahan. Ruang Rumah had no page for any of them; searches like “rumah dijual tebet” landed on a random filter URL.

  6. 16,200

    Duplicate listings

    The same unit posted by two or three agents, each copy competing with the others for the same query.

Where 30 days of Googlebot went7,200,000 requests, month 0

THE CENTREPIECE

Irrigate the terraces

Crawl budget is water. Block the channels that waste it and open the ones that lead somewhere useful. Water fills the terraces from the top; a terrace turns green as its pages get crawled and indexed.

Googlebot · 240,000 requests a day

54321

TERRACES, TOP TO BOTTOM

  1. District pages420 pages100% watered
  2. Active listings180,000 pages58% watered
  3. New listings, last 7 days16,800 pages0% watered
  4. Neighbourhood hubs1,602 pagesNot built yet
  5. Curated filter pages4,800 pagesNot built yet
BLOCK THE WASTED CHANNELS
OPEN NEW CHANNELS

Rice-terrace model of the crawl. 25% of Googlebot’s requests reach pages that matter; 1 of 5 terraces green.

Crawl reaching pages that matter

25%

60,000 requests a day reach the terraces

Lost to waste: 75%

Organic enquiries a month
9,400
First crawl of a new listing
19 days
Valuable pages indexed
105,420
Thin & expired URLs indexed
2,400,000

1 of 5 terraces green · Spare for recrawls: 0 a day

A simplified model of how crawl is shared out, not Google’s actual scheduler. Figures are modelled.

THE CHANNELS WE CUT

Four works on the hillside.

The engagement ran on our Commerce SEO plan: fewer pages for Google to wade through, more of the right ones, and a crawl that follows the market instead of the archive.

  1. 01 · FACETED NAVIGATION CONTROL

    Two facets, then a wall

    Area plus property type, or area plus bedrooms, stays indexable where people search for it. Anything deeper is noindex, follow; once it has dropped out, robots.txt stops the crawl. Sort and tracking parameters canonicalise to the clean URL.

    −36 pts of wasted crawl

  2. 02 · EXPIRED-LISTING LIFECYCLE

    A proper end for sold homes

    A listing that sells or lets shows a “no longer available” page with similar homes nearby for 30 days, leaves the sitemap the same day, then returns 410. Duplicate copies of one unit point to the primary listing.

    41,300 expired URLs retired

  3. 03 · NEIGHBOURHOOD HUB PAGES

    A terrace for every kelurahan

    1,602 hub pages, one per neighbourhood and property type, each with live listings, median price per m², the nearest MRT or KRL station and a short local guide. Hubs link to the newest listings first.

    1,602 hubs indexed

  4. 04 · CRAWL-BUDGET LOG ANALYSIS

    Reading the water

    Server logs, not guesses: every Googlebot request classified by page type each week, so we could see where the water went after each change and which terrace was still dry.

    Weekly crawl reports

SEASON BY SEASON

Twenty-two months down the hillside.

  1. Months 1–2

    Survey the slope

    Thirty days of server logs classified: 7.2 million Googlebot requests sorted by page type. The waste map drawn.

    26%
  2. Months 3–5

    Close the dead channels

    Expired listings get a 30-day “sold” page, then 410. Sitemaps rebuilt to list only live listings.

    47%
  3. Months 5–8

    Wall off the thin filters

    Indexing rules for faceted URLs: two facets maximum, only where people search. noindex first, robots.txt once they drop out.

    71%
  4. Months 8–12

    Build the neighbourhood terraces

    1,602 kelurahan hubs launched in waves by district, each linked from its district page and from every listing inside it.

    84%
  5. Months 12–15

    Open the fresh channel

    A new-listings sitemap with accurate lastmod and a “new this week” block on every hub. First crawl falls under a week.

    89%
  6. Months 16–22

    Tend and harvest

    Weekly log reviews, duplicate listings consolidated, curated filter pages added where demand appears. Enquiries keep climbing.

    92%
Share of the crawl reaching pages that matter, month by month
Month 0Month 2225%92%

HARVEST

The same water, a greener hillside.

+146%

Organic enquiries per month
9,40023,100Month 0Month 22

25% → 92%

Crawl reaching pages that matter
Before25%After92%

19 → 1

Days until a new listing is first crawled
191Month 0Month 22

2.4M → 61K

Thin & expired URLs in the index
2.4M61KMonth 0Month 22

Modelled scenario for a fictional business — not a client result. Rankings and traffic are never guaranteed; see our Terms.

SEARCHES, AS TYPED IN JAKARTA · POSITIONS

Positions for Indonesian property searches before and after the engagement
SearchMonthly searchesMonth 0Month 22Ranking page
rumah dijual jakarta selatan12,100175 · Up 12 placesDistrict pages
rumah dijual tebet4,400383 · Up 35 placesNeighbourhood hubs
rumah dijual kelapa gading3,600293 · Up 26 placesNeighbourhood hubs
apartemen disewakan kemang2,900444 · Up 40 placesNeighbourhood hubs
kost dekat mrt blok m2,400726 · Up 66 placesCurated filter pages
rumah sewa menteng1,900354 · Up 31 placesNeighbourhood hubs
apartemen dijual pluit1,600525 · Up 47 placesNeighbourhood hubs
apartemen 2 kamar kalibata1,300612 · Up 59 placesCurated filter pages

YOUR OWN HILLSIDE

Get a plan like this

If Google spends its time on pages you would rather it ignored, we can redirect the water to the ones that sell.

← Back to all case studies