Over the past few weeks, I have spoken with people acrossseveral CFD brokers. Different companies, different markets, but the complaintwas remarkably consistent: campaigns that worked in May and June weredelivering fewer first-time depositors by the end of Q3. Cost per acquisitionand overall customer acquisition costs were rising, and management wantedanswers.London's trading industry is coming home!Some brokers questioned their marketing teams. Others lookedtoward new agencies. I understand the instinct. Acquisition budgets areexpensive, and declining results demand action. But I think the industry isasking the wrong question.Instead of asking who failed to deliver the FTDs, perhaps weshould ask why brokers with years of client and trading records still rely soheavily on yesterday’s marketing assumptions.The Algorithms Really Did ChangeThis year, the advertising giants continued their shifttoward AI-driven optimization. Googlebegan automatically upgrading eligible Search campaigns using certain oldersettings to AI Max in September. TikTok revised its Smart+ system during Q3,combining automated and manual controls. LinkedIn expanded automated creativetesting, while Meta continued developing models that learn across its advertisingsurfaces.Not every change happened in Q3, and none proves thatalgorithms caused every broker’s declining performance. Competition, creativefatigue and changes in client demand also matter. Financial advertising facesadditional restrictions: Google permits eligible CFD campaigns only underlicensing and certification rules, while Meta has long prohibited direct CFDpromotions.Nevertheless, one thing is clear. The systems deciding whosees an advertisement are evolving rapidly. Running October’s campaigns onMay’s assumptions is risky.But the real failure may not be that brokers cannot controlthese algorithms. It may be that they have neglected the one advantage they docontrol: their own data.Sitting on a Gold Mine, Marketing with a GuessA CFDbroker can know far more about its clients than any advertising agencylooking at a campaign dashboard. It can see where funded clients came from, howlong they took to deposit, where onboarding failed, which instruments theytraded, how often they returned and what happened after their first deposit.Does a particular market attract traders interestedprimarily in gold, while another brings more activity in indices? Do certaincampaigns attract registrants who never pass verification? Does one sourcegenerate inexpensive FTDs who disappear after a week, while another bringsclients who remain engaged for months?These are questions for data, not instinct.Yet in my experience, marketing discussions too often startwith familiar beliefs: gold works in this region, forex performs in that one,this audience is younger, this creative converted last quarter. Industryexperience matters. I have relied on it throughout my career. But experienceshould generate hypotheses, not substitute for evidence.I cannot claim that most brokers fail to use big datawithout a comprehensive industry study. What I can say is that the gap betweenthe information a broker collects and the information its marketing teamactually uses deserves far more attention.A Registration Is Not a Business OutcomeHere is where the disconnect becomes expensive. An adplatform may report excellent results because it finds people likely to clickor register. But brokers are not paid for clicks, and a registration is not afunded account.Even FTD numbers alone do not tell the whole story. Anacquisition channel must be assessed against the cost of attracting eligibleclients, the quality of the onboarding process, retention and sustainablerevenue. Full customer acquisition cost includes more than media spend.If a broker sends an advertising system only registrationevents, it risks teaching the system to find registrations. Google’senhanced conversions for leads and offline conversion capabilities aredesigned to connect advertising activity with later outcomes; its 2026 changesto conversion data infrastructure make accurate implementation even morerelevant.That does not mean uploading clients’ trading histories intoan advertising platform. It means connecting permitted, reliable conversionsignals to campaigns and conducting deeper analysis inside the brokerage.The more AI automates advertising, the more important thequality of those signals becomes.The Solution Is Not Another AgencyBefore spending another million on acquisition, I wouldstart with a simple exercise: put marketing, CRM, platform activity,onboarding, deposits and revenue data into one analytical view. Not anotherdashboard full of impressions and clicks. A view that follows a client fromfirst touch to funded account and beyond.Then compare cohorts. Look at acquisition source, market,onboarding completion, instrument preferences in aggregate, engagement andretention. Identify where conversion dropped between June and October. Separaterising media prices from weaker registration-to-deposit conversion. Testwhether the problem appeared across every channel or only certain campaigns.If aggregate data shows strong interest in gold, testrelevant, compliant educational content rather than generic advertising. Onlythen should a broker change budgets, audiences, creative strategies oragencies.Big data does not mean buying an expensive piece of softwareand calling it transformation. It means building the ability to ask betterquestions and test the answers. Sometimes the result will prove the marketingteam underperformed. Sometimes it will expose weak tracking, an onboardingbottleneck or a product that no longer matches the audience.And because these are financial products, the analysis needsboundaries. Use aggregated insights where possible, respect data protectionlaws and platform policies, and never use trading losses or signs ofvulnerability to target people with more speculative advertising. ESMA hasrepeatedly stressed the need for appropriate target markets and investorprotection for complex derivatives.Understanding customers is not the same as exploiting them.The Advantage Brokers Already OwnGoogle, Meta, TikTok and LinkedIn will keep changing theirsystems. No marketing director can prevent that. What brokers can change istheir dependence on assumptions.When FTDs fall and acquisition costs rise, replacing amarketing team may feel decisive. But if the next team receives the sameincomplete data and the same flawed measurement, why should anyone expect adifferent result?Our industry has spent years collecting extraordinaryamounts of information about how clients arrive, trade and stay. Thecompetitive advantage is no longer simply having that data. It is turning itinto decisions.Before asking who to fire, I would ask one question: do we actually know our clients, or do wejust know how much we paid to acquire them?This article was written by Badea Alexandru Gabriel at www.financemagnates.com.