A fictional business. The figures are a modelled scenario, not a client result.
Case study · Commerce SEO
Levometar
Personal-finance comparison site · Sofia, Bulgaria
01 · The jam
Levometar's rate tables were loaded by script and updated manually, so Google saw empty pages and stale offers.
Levometar compares consumer loans, mortgages, deposits and cards from Bulgarian banks. Its editors knew the market, but the rates lived in a widget that filled itself in after the page loaded, from a spreadsheet someone updated when they had time. Search engines received headers and a spinner; visitors who did click through often met an offer the bank had already changed. Partners paid for clicks that never came.
Before · what Googlebot got
<table id="rates">
<thead>…APR · Cost…</thead>
<tbody>
<!-- filled by rates.js -->
</tbody>
</table>
<div class="spinner">
Зареждане…
</div>After · server-rendered
<table>
<caption>Loans of 10,000 lv,
60 months</caption>
<tr><td>Bank A</td>
<td>APR 8.9%</td>
<td>checked 06:00</td></tr>
…
</table>02 · Assay report
A rendering, log-file and template audit across 4,300 candidate URLs and 18 live bank feeds, scored by how many coins each fault kept out of the tubes.
RenderingCritical
Rate tables arrived by client-side XHR. The raw HTML of all 312 product templates held an empty <tbody>, so Googlebot indexed headers and a spinner.
FreshnessCritical
Offers were re-keyed by hand from bank PDFs. Median offer age was 41 days; 23% of listed APRs no longer matched the bank’s own page.
IndexationHigh
Only 180 of roughly 4,300 product, amount and term URLs were indexed. Filter parameters (?amount=, ?term=) canonicalised to the home page.
Structured dataHigh
No FinancialProduct, LoanOrCredit or DepositAccount markup. Rates, fees and providers were invisible to rich results.
TemplatesMedium
1,140 URLs shared the title “Кредити | Levometar”. Bank and product pages had no unique copy or representative examples.
PerformanceMedium
A 610 KB rate-widget bundle pushed mobile LCP to 4.8 s and blocked the table from painting.
LinkingMedium
Bank profile pages were orphaned. Nothing linked from a loan table to the provider or to neighbouring amounts and terms.
AttributionLow
Partner links ran through a three-hop JavaScript redirect, so outbound clicks were under-counted and slow to land.
03 · Rebuilding the sorter
Built on our Commerce SEO plan: every offer enters once, from the bank’s own feed, and the same record drives the table, the markup and the page that ranks for it.
Part 01 · Hopper
Nightly XML and JSON feeds from 27 partner banks, hourly for deposits. Each offer is checked against the bank’s published rate before it can drop; a mismatch holds the coin in the hopper and flags it.
27 feeds · 6,800 offers checked daily
Part 02 · Rail
Server-rendered HTML tables with a caption, a representative example, APR, total cost and a “checked” timestamp in every row. Sorting and filtering are progressive enhancement on top.
0 KB JS needed to read a rate
Part 03 · Assay stamp
LoanOrCredit and DepositAccount (both FinancialProduct types) with annualPercentageRate, interestRate, fees and provider on every offer, generated from the same feed record as the visible table.
100% schema/table parity
Part 04 · Sorting slots
Product × amount × term templates, published only when a combination has at least three live offers and enough unique copy. Thin combinations stay noindex until they fill.
3,900 pages passed the gate
04 · 11 turns of the crank
Render and log-file audit; feed contracts agreed with partner banks.
First 9 bank feeds wired in; validation rules against published rates.
Loan and deposit tables rendered on the server; widget bundle retired.
LoanOrCredit and DepositAccount markup shipped from the feed record.
Amount × term templates for consumer and mortgage loans behind the quality gate.
Provider profiles rebuilt and linked from every table row.
Hourly deposit feeds; term-deposit pages by duration and currency.
Credit cards and refinancing; city pages only where branches exist.
Template titles carry the best live APR and the check date.
Partner links reduced to a single tracked hop; click parity restored.
Pruned 210 thin combinations; feeds 18 to 27; handover runbook.
05 · The tubes, counted
Modelled scenario for a fictional business — not a client result. Rankings and traffic are never guaranteed; see our Terms.
Organic partner clicks
+246%
One coin ≈ +19%
Indexed offer pages
180 → 3,900
One coin ≈ 167 pages; grey = before
Rich result impressions
+410%
One coin ≈ +19%
Affiliate revenue
+138%
One coin ≈ +19%
Google.bg positions, month 0 vs month 11. Modelled monthly searches.
| Query | Searches | Before | After |
|---|---|---|---|
| потребителски кредитconsumer loan | 18,100 | 34 | 3 |
| кредит 10000 лв10,000 lev loan | 4,400 | — | 1 |
| депозити лихвиdeposit rates | 9,900 | 41 | 4 |
| ипотечен кредит сравнениеmortgage comparison | 2,900 | 22 | 2 |
| кредит за рефинансиранеrefinancing loan | 3,600 | — | 5 |
| срочен депозит 12 месеца12-month term deposit | 1,900 | — | 2 |
| кредитна карта без таксаno-fee credit card | 2,400 | 58 | 6 |
| бърз кредит онлайнquick online loan | 12,100 | 63 | 9 |
“—” means not in the top 100.
Month 0 · position 34
levometar.bg › krediti
Кредити | Levometar
Зареждане… Сравнете оферти. APR · Cost · Bank …
Month 11 · position 3 · rich result
levometar.bg › krediti › 10000-leva
Кредит 10 000 лв за 60 месеца — от 8,9% ГПР
| Bank A | 8,9% | 12 418 лв |
| Bank B | 9,4% | 12 571 лв |
| Bank C | 9,8% | 12 690 лв |
Checked today 06:00 · 14 offers
If your catalogue, rates or listings change faster than Google can see them, Commerce SEO turns the feed you already have into pages that rank.