Lookalike audiences are the closest thing Meta gives you to a cheat code for prospecting. Feed the algorithm a list of your best users, and it goes hunting for people who resemble them. For most advertisers that means uploading a purchaser list and calling it a day.
For finance agencies, that approach quietly burns budget. A forex broker’s “purchase” is a first-time deposit (FTD) that lands days after the click. An iGaming operator’s most valuable players look nothing like the tyre-kickers who registered and never funded. If your seed list is built from the wrong signal, your lookalike inherits the noise, and you scale the wrong people across millions of impressions.
This guide covers how to build lookalike audiences that actually find depositors and high-value players in regulated, high-risk verticals, and how to keep them honest once spend climbs.
Why generic lookalike advice fails in finance verticals
The standard playbook assumes a clean, fast conversion: someone clicks, checks out, fires a purchase event, done. Forex, crypto, iGaming and signals funnels break every one of those assumptions.
- The conversion is delayed. An FTD can arrive 3 to 14 days after the ad click. By the time it happens, Meta’s attribution window may have closed, so the event never makes it back into the audience you’d seed from.
- Registration is not value. In high-risk niches, the gap between a registrant and a funded depositor is enormous. A lookalike seeded on registrations finds more registrants, not more revenue.
- Account bans fragment your data. Multi-Business-Manager setups, sudden account disables and re-spun pixels mean your conversion history is scattered. A seed list pulled from one disabled BM is half the data you think it is.
So before you touch the lookalike percentage slider, fix the seed. Everything downstream depends on it.
Seed audiences that reflect real money
A lookalike is only as good as the list behind it. The job is to feed Meta the smallest, cleanest signal of value, not the largest list of activity.
Seed on FTDs, not registrations
Your highest-signal seed is funded depositors. Export that cohort and build your primary lookalike from it. A 1% lookalike of 2,000 real depositors will out-convert a 1% lookalike of 50,000 registrants every time, because it is trained on people who actually opened their wallet.
If your depositors happen off-Pixel and out-of-window, reconcile them before you seed: import the missed conversions, then build the audience from the complete list rather than the partial one Meta captured on its own. A seed that is missing your best customers can only ever describe your second-best ones.
Segment by deposit value, not just deposit count
Not every depositor is worth the same. A whale who funds 5,000 EUR and a minimum-deposit churner both count as one FTD, but you only want more of the first. If you track deposit value, build a value-based lookalike from your top quartile of depositors. Meta’s value-based optimisation will skew the lookalike toward higher-value matches, which is exactly what you want in iGaming and crypto-exchange campaigns where lifetime value is wildly uneven.
Keep seeds fresh and per-region
Two practical rules that matter more in finance than elsewhere:
- Refresh monthly. Regulated verticals see fast shifts in who converts as offers, jurisdictions and compliance rules change. A seed from six months ago describes a market that may no longer exist.
- Build per country. A depositor in the UAE behaves nothing like one in Vietnam or Germany. Country-specific lookalikes outperform multi-country blends because the similarity is computed within a coherent population. For forex prop firms running across dozens of geos, this single change often moves cost-per-FTD more than any creative test.
Configuring the lookalike itself
Once the seed is right, configuration is mostly about matching audience size to your funnel depth.
Pick a percentage by signal quality
The tighter and higher-quality your seed, the smaller the percentage you can afford to run.
- 1% lookalike — use this when seeded on FTDs or high-value depositors. Smallest, most similar, lowest cost-per-FTD. This is your default for conversion campaigns.
- 2-3% lookalike — use when your seed is a mid-funnel event, or when 1% can’t spend your budget. Good balance of scale and quality.
- 5%+ lookalike — reserve for top-of-funnel awareness or markets where the audience is genuinely too small at 1%. Expect more registrants and fewer depositors per dollar.
Test 1%, 2% and 3% side by side with equal budgets, but judge them on cost-per-FTD, not cost-per-registration or cost-per-click. In finance verticals the cheapest click is frequently the most expensive deposit.
Always exclude existing customers
Build an exclusion audience of current depositors and apply it to every prospecting ad set. Showing acquisition ads to people who already funded is pure waste, and in high-CPM finance auctions that waste compounds fast. This is one place the generic advice and the finance advice agree completely.
Stitch fragmented data back together first
If your conversion history is spread across multiple Business Managers and a couple of disabled accounts, consolidate it before you seed. A client hierarchy that unifies Client, Brand and Ad Account lets you assemble a complete depositor list rather than seeding from whichever BM happens to still be live. The more complete the seed, the more reliable the lookalike.
Judging and scaling without fooling yourself
The danger with lookalikes is that they look great on the metrics Meta optimises for and quietly fail on the metric that pays your invoices.
Measure the right cost
Watch cost-per-FTD as your headline number. Meta’s in-platform CPA will report on whatever Pixel event you optimised for, which is often a registration or a landing-page view, not a funded account. If you only watch the in-platform number, you will scale audiences that produce cheap registrations and expensive deposits.
Because FTDs are delayed, give each lookalike a fair attribution window before judging it. Pausing a depositor-seeded 1% audience after four days because “no conversions” is the most common way agencies kill their best audience before its deposits land.
Triage before you scale
Before pouring budget into a winning lookalike, check that the underlying campaigns are healthy. Scaling an audience whose results were really driven by one fluke ad that is already fatiguing just buys you a more expensive version of a problem you already had. Scale into stability, not into noise.
Scale gradually and watch the balance
Increase budget 20-50% at a time and let performance settle between steps. For prepaid ad accounts common in high-risk niches, watch the account balance as you scale so a fast-moving lookalike doesn’t drain a prepaid balance overnight and stall the whole account.
Common mistakes specific to finance lookalikes
- Seeding on registrations. You get more registrants and a worse cost-per-FTD. Seed on funded depositors instead.
- Ignoring late, off-Pixel conversions. If delayed FTDs never reach Meta, your seed is missing your best users. Reconcile them before you build the audience.
- One global lookalike across all geos. Regulated markets diverge sharply by country. Build per region.
- Judging on click cost. The cheapest traffic in forex and iGaming is often the least likely to deposit.
- Killing delayed-conversion audiences early. Respect the FTD lag before declaring a lookalike dead.
Build lookalikes around the conversion that matters
Lookalike audiences are not magic. They are an amplifier. Feed them registrations and clicks and they amplify low-intent traffic. Feed them funded depositors and value-weighted cohorts, and they go find more of the people who actually grow your clients’ books.
The hard part isn’t the lookalike slider. It’s having clean, complete, value-based conversion data to seed from in the first place. That means getting late and off-Pixel deposits back into Meta, weighting by value, and assembling a complete depositor list even when your accounts are scattered across Business Managers.
Want to seed your lookalikes on real deposits instead of guesses? Start a free trial or book a demo.