Enterprise Search — Live Demand Signal

What the market actually searched when we turned the ads on

Four weeks of live Google Search ads for Blockstream Enterprise — Brand and Unbranded — read purely as a demand and engagement signal. Not a performance review: a map of what institutions type, what they click, and where the real pull is. The Quant brief forecast this from Ahrefs; this is the auction confirming it.

May 5 – May 31 2026  ·  2 campaigns · 6,238 impressions · 559 reported search terms · $1,296 spend
  • Demand is overwhelmingly tokenization, not custody.62% of all impressions came from tokenization / RWA searches — "tokenization of assets," "rwa tokenization," "dtcc tokenization," "blackrock tokenization," "tokenized treasuries," "real estate tokenization." Every other theme combined is the remaining 38%. This maps to AMP (tokenized-asset issuance on Liquid), not the custody pitch the ads currently run.
  • Brand demand is a mirage.Only ~300 brand impressions in a month — almost nobody searches "blockstream enterprise." The 17% brand CTR looks great but it's "blockstream lugano" (the event) and "blockstream capital partners" (the investment arm) — not custody intent. Real enterprise-brand search is near zero.
  • The clearest custody engagement is on competitor names.Small volume, but the highest custody-side click rates we have: "copper crypto custody" 20%, "custodian crypto" 20%, "bitgo custody" 9.7%, "ledger enterprise" 4.8%. When someone searches a rival's custody product, our ad gets noticed.
  • The five ICP-targeted ad groups never served.Built May 13 — Qualified Custodian, RIA & Family Office, Corporate Treasury, Enterprise Multi-Sig, Competitor Conquest — all 0 impressions. Keywords were phrased the way we talk ("fireblocks qualified custodian") not the way buyers search ("fireblocks custody"), and a $30/day max-clicks budget funnelled everything into the one broad tokenization term.
  • This independently confirms the Quant brief.That analysis forecast — from Ahrefs estimates — that tokenization-issuer terms carry more demand than the other four ICP segments combined. The live auction agrees, from a completely separate data source.
Reported search-term impressions, May 5 – 31 (3,046 impressions across 559 terms)
Tokenization / RWA→ AMP issuance
1,888 · 62%
280 terms32 clicks · 1.7% CTR
Competitor names→ Custody conquest
447 · 15%
61 terms7 clicks · 1.6% CTR
Generic custody→ Custody & Treasury
422 · 14%
111 terms4 clicks · 0.9% CTR
Brand (Blockstream)→ mostly off-target
105 · 3%
19 terms18 clicks · 17.1% CTR
Wallet / multisig→ negligible
10
5 terms0 clicks
The volume sits in tokenization. The engagement sits in competitor-custody. The brand "demand" is the Lugano event. We are spending most of the budget on the highest-volume, lowest-pull theme — with an ad written for a different one.
Search termThemeImprClicksCTR
blockstream luganoBrand · off-target321031.2%
tokenization of assetsTokenization24852.0%
blockstream capital partnersBrand · off-target19421.1%
rwa tokenizationTokenization12132.5%
bitgo custodyCompetitor3139.7%
real estate tokenizationTokenization2428.3%
ledger enterpriseCompetitor4224.8%
copper crypto custodyCompetitor5120.0%
custodian cryptoCustody5120.0%
digital asset custodyCustody1317.7%
  • The forecast.The Quant brief concluded from Ahrefs that "tokenization-issuer terms carry the most search demand — larger than the other four segments combined," led by RWA / private-credit / tokenized-asset terms mapping to AMP.
  • The confirmation.Live auction data, four weeks, separate source: tokenization / RWA = 62% of impressions, 4× the next theme. The two datasets — estimated and actual — point the same way. That's about as strong as demand evidence gets.
  • What's new here.The Quant brief is a volume forecast. This adds the behavioural layer the auction can't be guessed: brand demand is near-zero and mostly off-target, competitor names out-engage generic custody, and our current keyword build can't actually reach the ICP segments.

Four moves, ranked by what we'd learn

01
Run tokenization as its own discovery test. It's 62% of demand — the highest-information bet available. Give it a dedicated ad group and copy written for RWA / tokenized-issuance intent (AMP), then watch CTR. If it lifts off 1.7%, there's genuine pull and a product story to build; if it doesn't, we stop funding it. Either way we learn something concrete.
02
Rebuild competitor-conquest the way buyers search. Replace "fireblocks qualified custodian" with the real query forms: "fireblocks custody," "bitgo custody," "copper custody," "cobo custody," "ledger enterprise." These already show the best custody CTRs — this tells us how much switching demand we can actually pull.
03
Give each test room to serve. $30/day across six ad groups, on max-clicks, guarantees the ICP groups get nothing. Separate the themes into their own budgets and lift bids/relevance so unbranded stops losing 62% of impressions to rank — otherwise every test starves.
04
Stop the brand leakage and clean the noise. Negative "lugano," "capital partners," "zurich" on brand; negative "stock," "etf," "trump," "news," "price," "jobs" on unbranded. Cheap, immediate, and makes every theme's signal readable.
SourceWhat we pulledNotes
Google Ads APICampaign, ad group, keyword, and search-term reports for the two live Enterprise Search campaignsCustomer 343-926-0341 · window May 5 – 31 2026 (launch to last settled day; Jun 1 excluded as incomplete)
Search-term view559 reported terms, 3,046 impressionsGoogle anonymises the low-volume long tail — these are ~half of the 6,238 total impressions, directionally representative
Impression-share metricsSearch IS, lost-to-budget, lost-to-rank, top & abs-top ISBrand IS 72.8% · Unbranded IS 17.4%, 61.9% lost to rank
Theme classificationEach search term bucketed by keyword match (tokenization, competitor, custody, brand, wallet)Buckets are analyst-assigned, not a Google taxonomy
Read as a demand and engagement signal only — conversions are deliberately out of scope here. Both campaigns currently run Maximize Clicks (no conversion goal), so click behaviour is the cleanest read available this window.