This is a guest post by Stanislav Galandzovskyi, an acquisition and growth marketing consultant working with fintech companies, brokers and prop firms and co-founder of Finforce One, a fintech marketing agency.
The conversation tends to go the same way. A firm has money now, from a round or a good quarter, and wants to know how quickly spend can go up. Fifteen minutes in I ask what share of registrations end up depositing. Usually there is a pause. Then either nobody knows, or the number that comes back is whatever the ad platform reported, which is a different number and almost always a friendlier one.
That pause is the whole argument. If you cannot answer from your own database, you are not ready to spend more, and the reason is arithmetic rather than caution.
The number nobody publishes
Cost per FTD is cost per lead divided by lead-to-FTD conversion. That is an identity, not a model. Your media team owns the first term. Onboarding owns the second, and onboarding is usually the one nobody has looked at in six months.
Across our client campaigns, lead-to-FTD lands somewhere between 2% and 4%, depending on market, offer and whether there is a competent desk behind the funnel. Hold cost per lead flat at $12 and watch what that spread does:
| Lead-to-FTD | Cost per FTD | Index vs 3% | FTDs from $100,000 |
|---|---|---|---|
| 1.5% | $800 | 2.00x | 125 |
| 2.0% | $600 | 1.50x | 167 |
| 2.5% | $480 | 1.20x | 208 |
| 3.0% | $400 | 1.00x | 250 |
| 3.5% | $343 | 0.86x | 292 |
| 4.0% | $300 | 0.75x | 333 |
Cost per lead held constant at $12. The conversion range comes from our own client campaign data; everything else in the table is arithmetic from those two inputs.
The same hundred thousand dollars buys 333 funded traders or it buys 125. Nothing about the media changed in between. No competitor outbid you, no auction got more expensive, and your creative is the same creative. The row I keep coming back to is the slip from 3% to 2%, small enough that most teams would not notice it for a month, and it raises cost per FTD by half.
The same hundred thousand dollars buys 333 funded traders or it buys 125. No competitor outbid you, no auction got more expensive, and your creative is the same creative.
There is no public benchmark for any of this. Listed brokers report funded or active clients, and some report a marketing cost per new client. What none of them publishes is the share of sign-ups that became funded accounts. The one number that would tell you most about a funnel is absent from audited disclosure across the sector. So when somebody quotes you an industry lead-to-FTD rate, ask where it came from.
Most of the friction is not yours
Founders arriving from e-commerce or SaaS want to delete steps. In a regulated brokerage you mostly cannot.
CFDs are complex instruments, which removes the execution-only route and makes an appropriateness assessment compulsory before the client trades. What happens after a client fails is narrower than people expect. MiFID II obliges the firm to warn. It does not oblige the firm to refuse. You can see the discretion written into the record-keeping rule itself: Article 56(2) of Delegated Regulation 2017/565 requires firms to record the warning given, whether the client asked to proceed anyway, and whether the firm accepted that request. The law assumes people will push past the warning and that firms will sometimes let them. That gap is where a lot of commercial pressure has historically gone.
The FCA reviewed 23 firms in June 2017 and found that where applicants failed the assessment but could easily click past the warning, a high proportion went on to trade. Read its remedy as a media buyer: applicants should not be asked to confirm they want to proceed as the very next step after the warning, and good practice includes a mandatory cooling-off period, or a separate communication the client must acknowledge before continuing. The regulator’s stated good practice is a delay in your funnel at exactly the point where you were losing people.
That has not softened. In a further multi-firm review in November 2025, the FCA noted some providers apply what it calls wealth bars: minimum salary and savings thresholds below which they will not accept a retail applicant even after that applicant has passed the test. Those firms are deliberately rejecting people who cleared every legal hurdle, and somebody paid to acquire each of them.
Cyprus tightened the deposit end too. CySEC’s Circular C721, issued 9 July 2025, caps deposits at €2,000 in aggregate before identity verification is complete, requires funds to arrive from a bank account in the customer’s own name, and gives the firm 15 days to finish verification or return the money. The detail people skip is that C721 relaxes the timing of verification only. Identification and the customer economic profile must be in place before any deposit is accepted at all.
| Where you lose people | What the rule requires | What is yours |
|---|---|---|
| Appropriateness assessment | Assessment compulsory, warning on failure compulsory, refusal discretionary | Question design, wording, retake logic, position in the sequence |
| Identity verification | Verification compulsory; identification and economic profile before any deposit | Method, device flow, document guidance, how a failed check gets recovered |
| Early deposit window, Cyprus | €2,000 cap, own-name funding, 15-day clock | Whether you rely on the derogation at all, and how you explain it |
| Risk warnings, leverage | Fixed by national product intervention measures | Comprehension, layering, timing |
| Deposit execution | Nothing compulsory | Rails, routing, retries, currency, what the failure screen says |
So the work is sequencing and comprehension. The fix is rarely a removed step. Usually it is moving the appropriateness test to a point where the applicant has already put in effort and is less willing to walk away. Sometimes it is nothing more than rewriting the document upload screen so people know what will be accepted before they photograph anything. And a failed verification needs somewhere to go other than a dead end.
Scaling on the wrong event is the expensive mistake
This is the one you cannot undo by pausing the campaign. Google’s documentation is specific. Target ROAS on Search and Shopping needs at least 15 conversions in the past 30 days at the conversion tracking level, with a learning period that can run to three weeks or one to two conversion cycles. Meta points to roughly 50 optimisation events per ad set in a seven-day window before learning settles. Both build a model against the event you nominate, and both are open about that.
Neither company writes anywhere that optimising on registrations will find you people who never fund. They do not have to. The system learns the label you hand it. Hand it registrations and it becomes excellent at locating human beings who finish registration forms. Whether those humans deposit was never the question you asked.
The documented answer is to send the funded event back as the optimisation signal, and that runs on a clock. Google will not import offline conversions uploaded more than 90 days after the associated last click, and for enhanced conversions for leads the window is 63 days. Since 15 June 2026 those uploads have also been migrating to the Data Manager API and blocked in the Google Ads API.
Put that against a funnel where registration to deposit takes three weeks and the verification queue adds another five days. You are not at the limit, but the signal reaching the platform is thin and late. Scale in that state and you buy a model trained on the wrong population, and you keep paying for it after the funnel is fixed.
The deposit belongs in the acquisition report
Most firms treat a failed deposit as a payments problem, reviewed by a different team in a different meeting. By then the media has been bought and paid for.
There is no honest public number here. Forex and CFD deposits sit in the highest merchant risk tier, but no reliable public figure separates issuer declines from acquirer declines from customers who gave up at an authentication screen, for brokerage merchants, in any region. The joint EBA and ECB reporting establishes that strong customer authentication works against fraud; it was not designed to measure completion. What I can say from our own campaigns is that when a market’s cost per FTD deteriorates without any change in lead quality, the deposit rail is the first place to look, and it is the answer often enough to check before the creative.
For prop firms, the purchase is not the conversion
A challenge purchase is a payment event and it flatters everyone to treat it as the finish line. Onboarding ends when the trader places a first trade on the paid account. Everything in that gap costs money without showing up anywhere: refund exposure, support load, and a customer who will tell other traders how it went.
Nobody publishes the interval between purchase and first trade, which means if you run a prop firm you are one query away from knowing something your competitors do not. Be careful with the rest of the prop numbers too. The often-quoted figure that roughly 7% of challenge buyers eventually get paid comes from industry press, not a regulator or an audited filing.
The regulatory picture is unsettled, and not in the direction most people assume. ESMA’s early-2026 statement on product intervention concerned crypto perpetuals, not evaluation challenges. Speaking to Finance Magnates in July 2026, CySEC chairman George Theocharides, who also chairs ESMA’s risk standing committee, said ESMA was not then in substantive discussions about retail prop trading, given the sector’s relatively limited size. A fast-growing sector whose regulator describes it as small is not a settled place to be scaling acquisition, and the position can change faster than a media plan can.
The workarounds are narrowing
The reflex when a funnel underperforms is to buy the volume back or put people on the phone. Both routes have narrowed.
Google’s financial services advertiser verification reached 24 further EEA markets with rolling enforcement from 23 July 2026, Cyprus and Malta among them. Advertisers get 30 days from notification to complete verification through Google’s external compliance partner before financial services ads are restricted, and agencies running those campaigns need verification on the in-scope accounts as well. Traffic you have already modelled and budgeted for is gated by a process, and the process runs on its own timetable.
The phone is harder still. Belgium bans retail CFD and rolling spot forex distribution through electronic trading platforms, and separately bans cold calling through external call centres. Spain’s CNMV resolution, applicable from 3 August 2023, bans CFD advertising to retail investors and the use of call centres to recruit them. France restricts unsolicited solicitation for products where the loss can exceed what the client put in. In the UK the unsolicited promotions rules sit in COBS 4.8 and the permanent CFD marketing restrictions in COBS 22.5. Get the citation right; the wrong one circulates constantly.
Enforcement is not hypothetical. In August 2024 the FCA fined a Cyprus-regulated CFD firm £276,100, reduced from £1.215 million for financial hardship, for failing to treat customers fairly and advising without permission. Telephone account managers had pressed customers to deposit, including customers who said they could not afford it. The FCA’s language there is worth remembering before anyone builds a conversion desk: it considers there are unlikely to be circumstances where a CFD provider should actively encourage, let alone pressure, a customer to deposit more.
Where this argument has limits
You cannot diagnose a funnel that has no traffic through it. The platforms document minimum event volumes before their systems stabilise, and ordinary testing practice needs a sample before a conversion difference is detectable. Below some floor you are reading noise and calling it insight. That is an argument for spending, not merely a risk to manage.
There is also a decent case against treating smoothness as the goal. In Signicat’s 2022 survey of 7,600 adults across 14 European markets, over half agreed that reducing fraud was an acceptable reason for a more complicated financial application. The wealth bar exists because some brokers concluded a slightly smaller funnel of better-capitalised clients was worth more than a larger one.
The limit is the difference between the volume you need to see a funnel and the volume you need to bet on one. A few hundred registrations in a single market, tracked server-side, will show you where the leak is. Tens of thousands will show you the same leak, after you have spent the money, trained your bidding on the wrong population and used up a first impression in a market you only enter once.
What to have in place before opening the budget
None of this is expensive. It is slower than launching a campaign, which is why it gets skipped.
- Lead-to-FTD read from your own database, not platform-reported conversions, and stable across four weeks
- Drop-off known separately for appropriateness, verification and deposit, per country
- Registration-to-deposit time held as a distribution, not an average that buries the tail
- Deposit success measured per rail and reviewed in the acquisition meeting, not the payments one
- The funded event sent back as the optimisation signal, comfortably inside the import window
- A documented recovery path for both appropriateness failures and verification failures
- One country, one language, one payment rail working before the second opens
| The arithmetic | Cost per FTD is cost per lead divided by lead-to-FTD conversion. Onboarding owns the denominator. A slip from 3% to 2% raises cost per FTD by half. |
| What you can change | Appropriateness testing, identity verification and risk warnings are compulsory. Sequencing, wording, comprehension and recovery paths are yours. |
| The lasting damage | Platforms optimise to the event you nominate. Scaling on registrations trains delivery toward people who register and never fund, and import windows put a clock on the fix. |
| The overlooked step | Deposit failure is an acquisition problem: the media was already paid for. No public decline benchmark exists for this vertical, so measure your own. |
| Prop specifics | Onboarding ends at the first trade on the paid account, not at the challenge purchase. Nobody publishes the interval between the two. |
| The narrowing workarounds | Google financial services verification reached 24 more EEA markets from 23 July 2026. Belgium, Spain and France restrict CFD advertising or cold calling. |
| The honest limit | Some volume is needed to diagnose a funnel at all. Buy that much, and no more, until it holds. |
Traffic is the most reversible decision in this business. Pause the campaign this afternoon, come back next month, same creative, roughly the same auction. A leaking funnel does not reverse like that: the money is spent, the bidding system has learned something wrong about who your customer is, and you have used up a first impression in a market you get to enter once. Fix the denominator. The traffic is not going anywhere.
who is who
Stanislav Galandzovskyi, is an acquisition and growth marketing consultant working with fintech companies, brokers and prop firms and co-founder of Finforce One, a fintech marketing agency. He has helped 30+ of them build acquisition systems that turn paid media into funded traders and first-time depositors, managed more than $3M in monthly ad spend, and run campaigns across Europe, the UK, the Middle East, Asia and Latin America, in 120+ countries overall. He has managed acquisition for NAGA, Zilch and Capital.com.