Terra-Incom
Customs clearance · Yandex.Direct · STAG restructure
Terra-Incom (ООО «Терра-Инком», Minsk, Belarus) provides electronic customs clearance, import/export declarations, and foreign trade (VED) consulting. Primary clients: companies needing regular or one-time customs processing for cross-border goods.
Target: CPL below ₽3,000, 75+ leads/month. Starting state (January–February 2024): 20 leads/month at ₽5,000 CPA. Every basic structural element of the Yandex campaign was misconfigured simultaneously.
30+ keywords in one ad group — algorithm had no intent signal to optimize toward
Zero negative keywords — ads showing on irrelevant queries, wasting budget
Yandex.Metrica not tracking form submission clicks — CPA calculated on phantom conversions
RYaN audience undivided: new visitors and retarget mixed in one placement — algorithm confused
Form submission goal set as primary conversion in Metrica
Call tracking connected — calls tracked separately from form leads
UTM labels on all sources — lead → channel → keyword → manager mapped
One intent cluster per ad group (STAG = Single Theme Ad Group)
Keywords in phrase match + broad modifiers — no pure broad match
Campaigns split: Russia-wide vs Minsk-only · by service type
New audience — cold, no prior site interaction
Retarget — site visitors 7–30 days, no conversion
CRM lookalike — modeled on existing client profiles
Auto-pause: any ad with CPA > ₽3,000 after 300 impressions — paused automatically
ROI optimization on RYaN — offline CRM conversions uploaded for smarter bidding
Weekly performance review with pause/scale decisions per group
3 UTP variants tested in search ads. Winner selected by combined CTR + CR (not CTR alone).
When 30 different keywords share one ad group, the algorithm receives one performance signal for all of them. It cannot tell which keywords drive quality leads and which drive irrelevant traffic. STAG (Single Theme Ad Group) structure separates the signals: each group's CPA is independent and actionable. The algorithm can now optimize per-intent rather than averaging across incompatible query types.
Before the audit, Yandex.Metrica was not tracking form submissions — the primary conversion event for a B2B lead generation campaign. The algorithm was optimizing toward clicks and page views. Auto-strategies optimize toward whatever goal is defined. If that goal is not a real conversion event, the algorithm finds users who click and view — not users who submit a request. Fixing measurement was the first action, before any campaign change.
'Customs clearance all-in — estimate in 1 minute' outperformed every other variant because it removes two primary B2B purchase blockers in one sentence: complexity ('all-in' = we handle everything) and uncertainty ('estimate in 1 minute' = no commitment required). The headline is not a feature statement. It's an objection removal. The other three variants stated features; this one stated the outcome of the interaction.
Spending ₽70,000 more per month and getting 55 additional leads means the incremental CPL on that additional spend was ₽1,273 — far below the ₽2,266 average. This happens when a campaign is structurally sound and the limiting factor is budget, not quality. A broken campaign does not improve proportionally with more spend. A correctly structured campaign can.
In 3 months: leads grew from 20 to 75/month (+275%). CPA dropped from ₽5,000 to ₽2,266 (−54%). Site conversion rate grew from 1.8% to 4.9%. ROMI improved from 180% to 487% — a 2.7× increase. Budget grew 70% from ₽100K to ₽170K, with the incremental spend performing at sub-₽1,300 CPL.
Winner UTP: «Таможенное оформление под ключ — расчёт за 1 минуту»
Result: 20→75 leads · CPA −54% · CR 1.8%→4.9% · ROMI 180%→487%
Client name is a pseudonym. Report ID is internal. All figures are actual Yandex.Direct and Metrica data for the January–May 2024 period.