Weekly GTM review · bolddesk

Run ID: 20261112-2134_weekly-gtm-review_bolddesk_e0d5 | Run by: Claude (unattended, via pbmaintain) | Period: last 28 days (2026-10-15..2026-11-11) vs previous 28 days (2026-09-17..2026-10-14), as set by the run parameters instead of the checklist's default Mon–Sun week.

Headline

Paid spend stayed nearly flat (-5.59%) but conversions fell 38.55% and CPA rose to $752.18. The cheap competitor-visitor display retargeting is now paused, and the budget went into high-CPC category search, including a US "Max Conversion" test that cost about $146.65 per click. Only 2 CRM-qualified leads were imported (4 previously), and Google Ads doesn't count them as primary Conversions, so bidding still optimizes for signups. Site sessions are up 41.76%, but much of the growth is Direct traffic and referral spam (aocr.org), not buyers.

Score

No score. The checklist does not define one, so overall_score is null.

Top 3 actions

1. Fix conversion goals so bidding sees qualified pipeline. Sales Qualified Lead and Paid Subscription imports show 0 in Conversions despite being primary. Make one signup action primary (not both Signup Completed and Signup Completed - Imported). Reconcile SQL counts with BoldSales. | Owner: Paid search lead | Due: 2026-11-19
2. Rebalance search budget. Stop or cap the US Helpdesk Max Conversion test ($5,865.99, 2 conversions). Move budget to India Helpdesk, which loses 38.1% of impression share to budget at about $281.84 per conversion. Decide whether competitor-visitor retargeting should return. | Owner: Paid search lead | Due: 2026-11-19 (needs approval, see below)
3. Clean GA4 data before reading growth. Exclude aocr.org referral spam (2,891 sessions, 0 key events), dev hosts and Tag Assistant test events. Investigate the +59% Direct jump. Add a contact-sales key event. | Owner: Web analytics owner | Due: 2026-11-19

Findings

Headline

Spend -5.59%, conversions -38.55% and cost per qualified opportunity +88.82%. Display retargeting is paused and spend shifted to high-CPC search. Non-brand organic clicks are up 12.11% while average position slipped.

Paid by campaign type (spend / clicks / conversions, this vs previous):

Notable changes

• Display retargeting (10 campaigns, now PAUSED): spend fell from $8,338.12 to $3,268.74 and conversions from 51 to 10. This explains most of the click (-42.0%) and conversion drop.
• US Helpdesk Max Conversion Testing: $5,865.99 for 40 clicks and 2 conversions. The manual-CPC control spent $5,079.32 for 98 clicks and 1 conversion. Neither is efficient. Zero-conversion terms in the test include 'customer service ai' at $633.50 for 2 clicks.
• India Helpdesk search: best Search performer (14 conversions, $3,945.78) but loses 38.1% of impression share to budget.
• Ads conversion mix: Signup Completed - Imported fell from 43 to 12, Demo Requested rose from 2 to 10, and Signup Initiated (secondary) fell from 185 to 95.
• GA4 /alternative/freshdesk and /alternative/zendesk: key events fell from 13 to 3 and from 9 to 1, as sessions dropped about 70%. This matches the paused competitor-visitor retargeting.
• GA4 /agent-builder and /support-agent: sessions rose from 198 to 1,030 and from 160 to 970, with 0 key events this period.
• GA4 /request-demo and /pricing: demo key events rose from 2 to 8 and /pricing views from 865 to 1,805. The requested_demo event rose from 9 to 24, but Tag Assistant test sessions carried 9 key events.
• GA4 referral: aocr.org brought 2,891 sessions with 0 key events and 17.8% engagement, which looks like referral spam. Direct rose from 13,914 to 22,175 with lower engagement.
• AI assistants: sessions rose from 148 to 266 (chatgpt.com 230), with 6 key events.
• GSC brand: impressions rose from 4,190 to 7,419, but CTR fell from 16.4% to 9.0% and clicks were flat (671 vs 687).
• GSC queries (partial view): 'bolddesk' gained 2,542 impressions and 'out of office message in teams' gained 1,371. 'customer success manager' lost 976. Most movers are informational blog queries.

Checklist results

Evidence for each item is in findings.json.

Not checked

• Bing Webmaster (W3.1 Bing half, W3.2 Bing, W3.3 Bing crawl issues): bolddesk.com is not in the Bing Webmaster account that the MCP can access (get_sites lists no bolddesk.com site). bing.nonbrand_clicks is null.
• W3.3 coverage, manual actions, Core Web Vitals: the Search Console MCP doesn't expose these reports. Only sitemap status was checked (6 sitemaps, 0 errors, 0 warnings).
• CRM reconciliation (W1.2): no BoldSales MCP was connected for this run. Qualified conversions come only from the Google Ads SQL import.
• Microsoft Ads: not connected. I don't know whether BoldDesk runs Microsoft Ads.
• W4.1 change: the latest AI monitoring run (20261112-1647) has no data, and the earlier run is the first baseline, so no change could be computed.
• W3.2 full ranking: limited to the top 60 queries by clicks per period. This is not a true impression-ranked top 10.

Proposed changes (need approval)

None of these has been applied. Each one needs a named human to approve it in chat.

1. Google Ads 5756012005, conversion goals: Sales Qualified Lead and Paid Subscription currently record 0 primary Conversions → include them in the account or campaign goals used for bidding. Signup Completed and Signup Completed - Imported are both primary → keep one primary and make the other secondary. Why: bidding currently can't see qualified pipeline and may double-count signups.
2. Campaign 'bolddesk - Search - US - Helpdesk - New - Manual CPC Max Conversion Testing - US Helpdesk': enabled, $5,865.99 per 28 days, 2 conversions → pause it, or end the test and add a max CPC cap. Why: about $146.65 per click against about $51.83 in the manual-CPC control.
3. Campaign 'bolddesk.com - Search - India - Helpdesk': budget-limited (38.1% of impression share lost to budget) → raise the daily budget, funded from item 2. Why: lowest cost per conversion among Search campaigns (about $281.84).
4. Helpdesk campaigns, negatives: add 'customer service ai' and 'support ai' as phrase negatives in the US Helpdesk and US Customer Support campaigns so those queries route to the AI Agent & Chatbot campaign. Why: $633.50 and $136.40 spent with no conversions.
5. Display competitor-visitor retargeting (10 campaigns, paused): keep them paused, or re-enable the best ones (US Freshdesk Keyword Visitors: 13 conversions on $682.29 in the previous period). This is a decision for the paid lead. Why: they carried 51 of the 83 conversions in the previous period.
6. GA4 property 312968368 (config change): add a referral exclusion or data filter for aocr.org and internal/dev hosts (localhost, *.workers.dev), and exclude Tag Assistant debug traffic from key events.
7. Website / support KB (from AI monitoring): update support KB 21603 to the current plans and credit pricing, and remove or source the ISO 27001 claim.

Data sources

• google-ads (responded): customer 5756012005 (_bolddesk.com, USD) via run_gaql: campaign metrics, conversion actions, search terms, ad policy status, budget-lost impression share and change history. Raw data: raw/google-ads-*.csv, raw/google-ads-account-and-conversions.json.
• google-analytics (responded): property 312968368 (BoldDesk - GA4) via run_report_tool. Raw data: raw/ga4-bolddesk.json.
• search-console (responded): https://www.bolddesk.com/ via search-analytics and list-sitemaps. Raw data: raw/search-console-bolddesk.json.
• bing-webmaster (responded, but no BoldDesk site): get_sites returned no bolddesk.com entry.
• AI monitoring log: reports/publish/bolddesk/ai-answer-monitoring/20261109-2041_ai-answer-monitoring_bolddesk_29e6/ (baseline) and 20261112-1647_…_48f6 (not checked).
• Choices made in this unattended run:
  • Used the 28-day window from the run parameters.
  • Brand queries are those matching the regex bold\s?desk.
  • "Trial sign-ups" means the GA4 email_verified key event.
  • Qualified conversions means the Google Ads "Sales Qualified Lead" import, all conversions.
  • Pricing views are counted for /pricing only.
  • Owners are given as roles, because the checklist names no people.
  • Remember to add this result to the team's shared results log.