The Financial Services Guide to Customer Acquisition

The six things that make it hard

Financial services acquisition runs into the same six obstacles whatever the product. They are worth naming, because each chapter in this guide is a response to one of them.

1. The buyer is comparing on price, in public. A term deposit, a home loan, a credit card or a brokerage account is shopped on a comparison site before your page is ever opened. If your rate is not competitive, no campaign fixes that. If it is, the campaign has to prove it at the moment of decision.
2. The category is dominated by balance sheets bigger than yours. The Big Four in Australia, the high-street banks in the United Kingdom and the captive lenders at the dealership advertise nationally, at category-typical cost, all year. A mutual or a specialist cannot outspend them; it has to out-target them.
3. The conversion happens after the click, and often after a human. A home loan application becomes a loan at the landing page and the broker handover, weeks later. A trading account is worthless until it is funded. Attribution that stops at the form is attribution of the wrong event.
4. The margins are thin and regulated. Interchange on a low-rate card is capped, and in Australia the Reserve Bank lowers the caps from 1 October 2026. KPMG’s 2025 Mutuals Industry Review put the mutual sector’s net interest margin at 2.28%. Acquisition cost has to sit inside those numbers, which means the ceiling is set by the product, not by the media plan.
5. The rules are real, and they reach the ad. An Australian ad for a fixed-term consumer loan, such as a home, car or personal loan, needs a comparison rate. A UK financial promotion has to be clear, fair and not misleading, and since July 2023 the Consumer Duty has asked firms to check that their communications support good outcomes. A New Zealand lender advertises inside the Credit Contracts and Consumer Finance Act and its Responsible Lending Code, with the Financial Markets Authority as regulator since 1 July 2026. Every page, ad and email is written inside those lines.
6. Trust is the product. A mutual’s community standing, a bank’s award, a platform’s regulator licence: these are the reasons a rate-shopper chooses you over the cheapest listing. They only work if the customer can see and verify them at the moment they decide.

Why it is worth doing on the numbers

When the economics are modelled properly, financial services can justify more acquisition spend than a cost-per-click view suggests, because the customer keeps paying. A term deposit acquired for Teachers Mutual Bank was valued on the margin its balance earns while it is held, not on the cost of the click that opened it. Modelled at the KPMG-disclosed mutual sector net interest margin of 2.03% and a three-year average holding period, a single six-week campaign returned 5,090% on its investment. The acquisition cost was paid once; the margin was modelled across the three-year hold.

• 5,090%Return on a single six-week term deposit campaign, modelled as A$5.7M of multi-year net interest margin (Teachers Mutual Bank)Case study →
• A$94MModelled home lending originated across three connected flights at a blended A$716 per application (Teachers Mutual Bank)Case study →
• A$12Cost per credit card application, against a Finance and Insurance Search benchmark of about A$110 (Teachers Mutual Bank)Case study →

What follows is the order we now build it: the website first, then organic and AI search, then paid search by product line, then the audiences, the amplification, the conversion flow, the nurturing that turns an application into money, and the attribution and economics that decide what to do next. Chapter one starts with a trading platform where the work began with the website.

A trading platform that started with the site

Rakuten Securities, a global forex and metals trading platform, had an acquisition problem: sign-ups cost A$100 each, and too few of them went on to fund their accounts. The previous programme had been run by the group’s own marketing arm. The engagement began as a website redesign, and the website was the first priority.

The website was overhauled, all the creative was redesigned and built from the ground up, and the marketing was restructured as a two-stage funnel: acquire the sign-up first, fund the account second. Campaigns were written for active traders, people who already understood terms like tight spreads and leverage. The media changed with it: long-tail and competitor search terms instead of broad targeting, audiences cut into micro-niches by interest, platform, intent stage and geography across Google, Facebook, YouTube, LinkedIn, Twitter and programmatic, and continuous creative testing across every format.

• 80%Reduction in cost per sign-up, from A$100 to A$20, after the site, the creative, the funnel and the media were all rebuilt (Rakuten Securities)Case study →
• 60%Of sign-ups went on to fund their accounts (Rakuten Securities)Case study →
• 5×Funded accounts against the original forecast (Rakuten Securities)Case study →

What a financial services website has to do

• One page per product, with the rate and the fee on it. A page called “Savings” competes with every bank on earth. A page called “Term deposits for self-managed super funds”, with the current rate, the minimum balance and the term options in the first screen, competes with almost nobody and answers the question the searcher typed.
• The number the law requires, where the law requires it. In Australia the advertised rate on a fixed-term consumer loan carries its comparison rate and the prescribed warning. In the United Kingdom a consumer credit promotion that quotes a rate or a cost of credit carries a representative example, representative APR included. Put them where a regulator would look for them, and where a customer would, which is the same place.
• Calculators and eligibility before the application. A repayment calculator, a borrowing-power estimate, a deposit-growth illustration. They do two jobs: they keep the ineligible from applying and wasting your assessor’s time, and they let the eligible see the number before they commit.
• An application in as few steps as the risk allows. Identity and consent checks belong late in the flow, not on the first screen. Save-and-resume is not a luxury on a home loan form; it is where many applications would otherwise die.
• The people and the licence. Australian financial services licence and Australian credit licence numbers, the FCA firm reference, the number from New Zealand’s Financial Service Providers Register. Names and photographs of the people who answer the phone. Anonymous institutions are hard to trust and, as chapter two explains, hard for an answer engine to recommend.
• The proof. Awards, audited growth, member numbers, the years in business. Teachers Mutual Bank’s credit card carried Money Magazine’s Cheapest Credit Cards recognition through its campaign window, and that third-party line corroborated the price story at the moment of decision.

Speed is a ranking factor and a trust factor

Google measures how real visitors experience your site and uses it in ranking. The three Core Web Vitals are Largest Contentful Paint (the main content should appear within 2.5 seconds), Interaction to Next Paint (the page should respond to a tap within 200 milliseconds) and Cumulative Layout Shift (nothing should jump around after it loads; a score under 0.1). Institutional sites often fail the first one on a phone, because a content platform and a tag manager can ship a megabyte of script before a rate is visible.

The commercial cost is simpler than the technical one. A rate-shopper on a train opens two lenders from a comparison site. One paints in a second, the other in six. The second lender does not get a second chance, and never finds out. Our free website audit measures six things on any site, from the bytes it actually sends rather than from a score: whether the final URL is secure, whether the homepage title says what the institution does, whether anything on the page tells Google where it is, whether the page tells a phone how to size itself, whether there is a description for search results to show, and how heavy the document and its render-blocking assets are.

Writing inside the rules

Compliance is not a reason to say less; it is a reason to say precise things. In Australia, ASIC’s Regulatory Guide 234, reissued in June 2026, is its guidance for anyone promoting financial products, advice or credit on staying clear of false or misleading claims; the National Credit Code requires a comparison rate when a fixed-term consumer loan, such as a home, car or personal loan, is advertised; content that gives general financial product advice to retail clients carries a general advice warning; the design and distribution obligations give the products they cover a target market determination the marketing has to respect; and a licensee that earns commission cannot call itself independent. In the United Kingdom, every financial promotion has to be clear, fair and not misleading, whatever the medium, and the Consumer Duty, in force since 31 July 2023, asks firms to test their communications and monitor whether they support good outcomes, not merely to avoid bad ones. In New Zealand, the Financial Markets Conduct Act’s fair dealing provisions cover financial products and services, credit contracts included, and consumer credit is advertised inside the Credit Contracts and Consumer Finance Act and its Responsible Lending Code; the Financial Markets Authority has regulated the Act since 1 July 2026.

The practical version: state the rate, the fee, the eligibility and the audience the product was designed for, in the words the regulator uses. Use customer outcomes with permission and with the caveats the licence requires. Have your compliance team read the site once, and keep a note of what they approved. Then never write a page, an ad or an email that could not pass the same reading.

Where the rate-shopper starts

The queries in financial services sort into three kinds. Head terms (“home loan rates”, “best savings account”) are held by comparison sites, the largest banks and, increasingly, by an AI Overview that summarises the comparison sites. Product-and-segment terms (“term deposit rates for SMSF”, “home loan for teachers”, “low rate credit card no annual fee”) are where a mutual or a specialist can rank, because the page that answers them is specific and the competition is thin. Question terms (“fixed or variable in a falling rate environment”, “what does a mortgage broker cost”) are where the AI engines do most of their answering, and where an institution earns the trust the product pages convert.

The plan for organic search is therefore not to chase the head. It is to build one page for every product-and-segment query you can honestly serve, one guide for every question that comes before the purchase, and the entity signals that let a machine say your name with confidence.

Pages that earn the click

• Product pages that answer, in order. The rate and the fee, who it is for, what it costs to leave, how to apply, and the four questions people actually ask. The answer to each question in one plain paragraph under its own heading, so a search engine and an assistant can lift it whole.
• Guides for the question before the purchase. Fixed versus variable. Offset versus redraw. What funding a trading account involves. Whether a roboadvisor is advice.
• Calculators that are pages. A repayment calculator with its own URL, its own title and its own explanation can rank for “repayment calculator” queries and hands a warm visitor to the product page beside it.
• The member story. A mutual’s reason to exist is its members. Who the bank is for, how many members it serves and what it returns to them is the story, and it belongs on the site in plain numbers.

Being the answer, not just a result

An answer engine, whether Google’s AI Overview, ChatGPT, Gemini or Perplexity, names an institution when it can corroborate it: the same name, product and claims across the regulator’s register, the comparison sites, the awards, the reviews and the institution’s own site. For financial products the engines lean unusually hard on those external sources, because the cost of being wrong is high. That makes the work concrete. Every listing has to carry the same legal name, the same licence number and the same product names. Every award has to be visible on the site and on the awarding body’s site. Every person named on the site has to be findable on a professional network.

On the page itself, the answer-engine work is a direct answer at the top of every product and guide page (two or three sentences that state the fact, the number and the condition), question-and-answer sections marked up as FAQ structured data, organisation and product structured data that carries the licence, and named authors with their own profile pages. Google says its AI Overviews and AI Mode need no special markup, so treat this as making the page easy to quote, not as a switch that gets it quoted.

State by state, brand from generic

The home loan programme ran three connected flights across the portfolio: the largest, multi-channel Australian Dream flight; a Classic Home Loans flight weighted toward retargeting; and a tightly bounded six-week Fixed Home Loans flight at a prime rate. The search architecture was the same each time. Every state and territory was campaigned separately, with brand and generic keyword sets split in each, so a searcher in Perth met keywords and bids set for Western Australia rather than a national average. The acquisition cost had to stay materially under A$1,000 per application for the unit economics to hold.

Across the three flights the whole channel mix, search with Facebook, retargeting, display and selective comparison-site and LinkedIn placements, produced 103,211 clicks, 33.5 million impressions and 342 verified home loan applications at a blended A$716 per application. The Fixed Home Loans flight recorded a 22.17% Google Search click-through rate over 33,009 impressions, 7.6 times the Finance and Insurance Search benchmark of 2.91%; Australian Dream ran at 16.37% over 232,153 search impressions. At the bank’s documented 66% application-to-loan conversion that is roughly 226 settled loans and, at the documented Australian average new home loan size in the window, approximately A$94 million in modelled originations.

• 22.17%Google Search click-through rate on the Fixed Home Loans flight, 7.6× the Finance and Insurance benchmark of 2.91%Case study →
• A$716Blended cost per verified home loan application across three flights and every channel, inside the sub-A$1,000 ceilingCase study →
• 342Verified home loan applications, modelled to about 226 settled loans at the documented 66% conversionCase study →

Where the product’s price is the ad

The credit card product was deliberately low-rate, which constrained its economics: less revolver interest, more reliance on interchange, which is capped under the Reserve Bank’s standard. Acquisition cost had to sit far below the category. The search layer targeted low-rate and balance-transfer keyword sets where the product’s actual price advantage could win the click, with comparison-channel placements used selectively and a Facebook prospecting layer in the education-sector audience. Applications were acquired at approximately A$12 each against an approximate A$110 Finance and Insurance Search benchmark. The Money Magazine Cheapest Credit Cards recognition the product held through the window corroborated the price story at the moment of decision. A competitive price, a keyword set concentrated on the intent that price wins, and a landing page that did not leak: the gap came from all three together.

Consumer-direct, before the showroom

Car loans are captured at the point of sale by dealer finance and nationally by the largest banks. A mutual goes after the consumer-direct segment instead: the buyer who has chosen the car and is shopping the loan, before the dealer’s finance partner meets them at the showroom. A six-week consumer-direct flight routed straight to the product page delivered 5,975,708 impressions, 13,038 clicks at A$2.94 and 66 verified applications at A$580.78 each; the case study reports the flight as a whole, not by channel, so what transfers is the targeting and the routing. At an industry-typical 66% application-to-settlement conversion that is about 44 settled loans and, at the Australian Bureau of Statistics average new car loan of about A$24,000, approximately A$1.05 million in modelled lending, a modelled multi-year margin return of about 231% with a sensitivity band the case study states rather than hides.

Rules of the account

• Phrase and exact match only, negatives first. Broad match on “home loan” buys the entire mortgage-broker industry’s curiosity. The search-terms report is read weekly and every term you would not pay for becomes a negative before it costs another click.
• One landing page per ad group. The low-rate card ad lands on the low-rate card page with the rate in the first screen, not on the credit cards hub.
• Brand campaigns kept separate. Brand traffic converts at a rate that flatters everything it is mixed with. Report it alone so the generic campaigns are judged on their own economics.
• The legal number in the ad, and a verified account. A comparison rate in an Australian ad for a fixed-term consumer loan; a representative example, with its representative APR, in a UK consumer credit ad that quotes a rate or a cost. Google also shows financial services ads in Australia, New Zealand and the United Kingdom only from advertisers it has verified, so start that verification well before the first flight.
• Manual bids until the account can learn. Smart bidding needs volume; a mutual’s ad group does not have it on day one. Manual cost-per-click caps for the first thirty days, then a target cost per application once an ad group has thirty conversions to learn from.
• The conversion is the application, and later the settlement. Never the click, never the calculator use. Chapter eight covers how the settled loan gets back into the account.

Three audiences, one product

The term deposit campaign ran three parallel audience strategies, each with its own creative, intent profile and channel allocation. Prospecting reached in-market term deposit shoppers and the community-trust audience, the teachers and education-sector segments the bank exists for, across Google Ads, Facebook, programmatic display and Instagram. Remarketing went back to recent visitors who had landed on the term deposit pages without converting, on Facebook and the Google Display Network, with the product’s benefits and its offer dates. Closing went after the highest-intent segments, drop-offs, multi-page visitors and current customers identified from CRM data, with strong calls to action and date-specific deals on Facebook and display. Working media ran at a blended A$2.43 per click and scaled materially through the engagement as performance gave the business case to keep investing.

• 24,722Estimated monthly prospecting clicks across Google Ads, Facebook, programmatic display and Instagram at a blended A$2.43 per clickCase study →
• 5,556Estimated monthly remarketing clicks against recent site visitors on Facebook and the Google Display NetworkCase study →
• 2,847Estimated monthly closing clicks against drop-offs, multi-page visitors and current customers from CRM dataCase study →

The three-stage shape, prospect, remarket, close, is the one to copy. Each stage has a different audience, a different message and a different acceptable cost, and reporting them together hides which one is working.

Micro-segmentation for a trading audience

For Rakuten Securities the audiences were cut into micro-niches by interest, platform, intent stage and geography, then targeted across Google, Facebook, YouTube, LinkedIn, Twitter and programmatic. Creative was written for active traders who knew what tight spreads and leverage meant, because the brief was traffic that went on to fund an account, not volume. The Google Ads strategy moved from broad targeting to long-tail and competitor terms, and Chinese-language ads were introduced in Australia; on those Chinese-language variants, cost per click fell 42%. Facebook conversion rates rose 3.4 times after the audiences were refined. The lesson is not the platforms; it is that the segmentation aimed at the traders who would fund, not at whoever would click.

Timing the market

JB Markets, a securities and derivatives firm serving wholesale and retail traders, needed net new trader sign-ups at a cost per lead under A$50. Two gated offers ran in parallel: an ASX report delivered both through a landing page and through Facebook lead ads, with the two funnels tested against each other; and a Bitcoin Futures campaign timed to the CME’s launch of the contract, when news coverage and visible price action were drawing prospective traders to the category. The ads ran on Facebook, Google Display and remarketing. Leads arrived at A$40 in the first week, A$10 under target, and at A$20 once the A/B testing data accumulated. Two timely offers, and testing that kept running after the first week, brought the cost 60% under what the plan had budgeted.

Who not to attract

A product’s target market determination, or its equivalent outside Australia, is the most useful audience brief in the building. The speculator who will never fund, the rate-hopper who leaves at the first maturity, the applicant outside the lending criteria: each costs an assessment and returns nothing. Exclude them in the targeting, say who the product is for on the page, and ask the qualifying question before the application, not after it.

Paid social carries weight when it is attributed

On the home loan programme, Facebook prospecting against custom audiences produced 95 home loan conversions at A$384 each, outperforming search on both volume and cost on the same flight. AdRoll retargeting across the long consideration window produced 85 at A$528. Both were measured rather than estimated because tracking had been rebuilt on Campaign Manager 360, and that is what let the channel mix change across the second and third flights. Social and retargeting are not awareness spend in financial services. They are acquisition channels that happen to sit earlier in the visit.

• 95Home loan conversions from Facebook at A$384 each, beating search on volume and cost on the same flightCase study →
• 85Home loan conversions from retargeting at A$528 each across the consideration windowCase study →
• 97%Increase in video view-through rate on the Rakuten Securities programmeCase study →

Premium environments without premium cost

For the launch of Apple Pay, Macquarie Bank’s display campaign was constrained to an Apple-mandated list of 48 premium publications, from the BBC and the Sydney Morning Herald to Forbes, Reuters and The Economist. The placement strategy concentrated spend on the publications where the bank’s audience was already present rather than spreading across the whole list, and the static brand assets were animated into HTML5 creative fit for those environments. The campaign delivered at a cost per thousand of A$2.52 and a cost per click of A$3.14, on premium-publication inventory that typically carries a cost per thousand many multiples higher. Disciplined placement, with brand-environment compliance built into the workflow rather than bolted on, absorbed the constraint without inflating the cost.

Traditional and digital in concert

Owners Advisory by Macquarie was a newly launched roboadvisor for do-it-yourself and self-managed super fund investors, in a category the market did not yet understand. The programme ran traditional media alongside performance digital under one team, on the premise that for an emerging product buyer education is the precondition to acquisition. Search, social and display carried the high-intent capture; traditional media carried the category. Sessions rose from 4,453 a month at kickoff to a peak of 9,703, a 2.18 times lift, with users up 2.76 times, and held at 8,726 in month four. For an emerging product category, the awareness layer compounding into engaged sessions was the number that mattered.

The application is the landing page

• The rate, the fee and the next step in the first screen. A visitor from a comparison site is checking that the page says what the listing said. If it does not, they are gone before the scroll.
• Eligibility before effort. Two questions that tell the ineligible not to bother save your assessors a queue and save the applicant a rejection.
• Steps sized to the product. A term deposit can open in minutes. A home loan application is a conversation across days; build save-and-resume and a named human into it.
• Identity and consent late, not first. Verification belongs after the applicant has decided, not as the price of finding out.
• The disclosure block where the regulator would look. Comparison rate, target market, general advice warning, licence numbers. Present, legible, and not in the way.
• A human option on every screen. A phone number that is answered and a booking link. The applicant who would rather talk is often the one with the largest balance.

Lead form or landing page? Test it

JB Markets’ gated ASX report was delivered two ways at once, through a landing-page form and through Facebook’s native lead form, and the two were tested against each other as the campaign accumulated data. That test, with the continuous A/B testing across the Facebook, Google Display and remarketing ads around it, is what moved cost per lead from A$40 in week one to A$20. Native lead forms convert more cheaply and qualify less; landing pages convert fewer people and tell you more about them. Which one wins depends on what your sales team does with a lead next, so run both and let the funded-account number decide.

Funded, not just signed up

The Rakuten Securities restructure split acquisition into two stages with two measurements: the sign-up and the funded account. Sixty percent of sign-ups funded, above expectations, and that number was there to report because funding was measured as a stage of its own. A brokerage, a lender or a bank has the same second stage under a different name: funded, settled, activated, first deposit. Name it, measure it, and report conversion against it rather than against the form.

The gap between application and money

At the bank’s documented rate, one home loan application in three does not become a loan. The car loan model applies the same 66% as an industry-typical rate. On the trading platform, four sign-ups in ten did not fund. The reasons repeat: documents not supplied, a competing offer, a broker who went quiet, a buyer captured at the showroom, an applicant who simply forgot. None of these is a media problem, and no amount of media spend fixes them. A nurturing sequence does, and it is the cheapest conversion lift in the funnel because the acquisition cost has already been paid.

Sequences that stay inside the rules

• Service messages, not promotions, wherever the law allows it. “Your application is missing a payslip” is a service message. “Rates have moved, apply now” is a marketing one, and it needs consent under the Spam Act in Australia, the Unsolicited Electronic Messages Act in New Zealand and the privacy and electronic communications rules in the United Kingdom.
• Short, specific, and stopped when the job is done. Three touches over ten days for a deposit; a longer, gentler cadence for a home loan that is waiting on a valuation. Every message answers one question or asks for one thing.
• A named person. The sequence comes from the lending specialist or the account manager, with a direct number, not from a no-reply address.
• The same compliance reading as the site. The comparison rate, the target market and the general advice warning travel with the message.

Capacity is part of the funnel

JB Markets’ campaign had to be paused. Not because it failed, but because at A$20 a lead it produced more volume than the internal sales team could service, and a lead that waits is a lead that goes elsewhere. The pause held until the firm streamlined its process to absorb what the campaign could deliver. Plan the handover before the first flight: how many qualified applications a week the team can process, who owns the ones that arrive on a Friday afternoon, and what the campaign does when the queue is full. A budget the operation cannot absorb is not a growth plan.

The rollover audience

The cheapest deposit a bank acquires is the one it already holds. In the term deposit system, the existing member rolling a deposit over had a search keyword set of their own, beside the comparison-shopper and brand-defensive sets, and the campaign’s closing stage went back to current customers identified from CRM data. Every product has an equivalent moment, the fixed rate expiring, the card anniversary, the account that has not traded in ninety days. Build the sequence for it before you buy a single new customer.

Rebuild the attribution before the first flight

The home loan, term deposit, credit card and car loan systems shared one attribution layer, so every channel’s contribution to every product could be reconciled and the channel mix reset between flights on what had actually worked rather than what was assumed to have worked. It is the reason Facebook’s 95 home loan conversions and retargeting’s 85 were measured rather than estimated, and it was what the bank had asked for: attribution clean enough that the next year’s plan could be modelled rather than guessed. Rebuilding attribution is unglamorous work: a tag plan, a conversion taxonomy, consented cookies, server-side events where the browser blocks them, and a reconciliation against the core banking system every month.

Define the conversion by the product

The platform is told about the second column and optimises for it. The business is run on the third. Chapter eight’s job is to make the fourth column real, so the platform learns from the number that matters.

Close the loop offline

Google Ads and the social platforms accept conversions after the fact, matched by the click identifier stored at the first visit. A settled loan imported three weeks after the click teaches the bidding which keywords produce loans rather than applications, and it is one of the strongest levers on cost per funded customer once the account has volume. Consent comes first: once it is matched to an applicant, the click identifier is personal information under Australia’s Privacy Act and New Zealand’s Privacy Act 2020, and personal data under the UK GDPR, so the consent banner, the privacy policy and the data-sharing terms have to say what is being done with it.

The five numbers to report monthly

1. Cost per application, by product line and by channel, with brand reported separately.
2. Application-to-funded (or settled, or activated) rate, and the days it takes.
3. Cost per funded customer, the number the ceiling in chapter nine is set against.
4. Modelled margin acquired this month, on the product’s own economics, with the assumptions stated.
5. Share of conversions attributed beyond last click, so the channels that start the journey are funded as well as the one that ends it.

Run it on your numbers

Set the ceiling from the margin, not the market

The ceiling on acquisition cost is the modelled margin per customer, multiplied by the holding period, divided by the payback you require. For a term deposit: the average balance acquired, times the net interest margin, times the average years held. Teachers Mutual Bank’s campaign modelled the balances at the KPMG-disclosed mutual sector margin of 2.03% over a three-year average hold, which produced A$5.7 million of modelled margin from a single six-week flight and a 5,090% return on the campaign investment. That arithmetic was the brief: the campaign had to clear its own margin hurdle, not just a cost per acquisition.

• 2.03%The mutual sector net interest margin the deposit balances were modelled at (KPMG, year to June 2017)Case study →
• A$5.7MModelled multi-year net interest margin from one six-week term deposit flightCase study →
• 5,090%Return on that single six-week term deposit campaignCase study →

Worked models from the case studies

Sensitivity, honestly

A model is only as useful as the range it admits. The car loan return of about 231% carries a stated band of 102% to 378%, reflecting the ABS principal range of A$22,000 to A$26,000 and a net-margin band of 8% to 16% of principal across the three years. The home loan and deposit case studies state no band: the home loan originations depend on the documented 66% conversion and the national average loan size holding, and the deposit margin depends on balances staying for the modelled three years at the modelled margin. State the bands, update them from the settled numbers every quarter, and let the ceiling move with them. A return quoted without its assumptions is a number nobody should spend against.

Growth you can audit

The test of a modelled return is whether it shows up in the accounts. In 2016–17, Teachers Mutual Bank’s audited retail deposit book grew A$743 million, from A$4.55 billion to A$5.29 billion (16.3%), and its total loan book grew 19.2% to A$5.48 billion. The asset base grew from A$5.54 billion to A$6.68 billion. Those are bank-wide figures, and the year included the bank’s November 2016 merger with the Fire Brigades Employees’ Credit Union, which brought more than A$215 million in assets, so they are not the acquisition systems’ result alone. The case studies say the systems contributed to that growth; they do not claim all of it. The point is that the campaign models and the audited growth pointed the same way.

• A$743MRetail deposit book growth at Teachers Mutual Bank in 2016–17, a year that included its November 2016 credit-union mergerCase study →
• 19.2%Loan book growth to A$5.48B at Teachers Mutual Bank in 2016–17, the merger yearCase study →
• 231%Modelled multi-year margin return on the car loan flight, band 102% to 378%Case study →

What to do, in order

1. Model the customer’s value first. Margin, holding period, payback. Set the ceiling on acquisition cost from that, per product line, and write it down.
2. Rebuild the attribution. One layer across every product and channel, the conversion defined as the money event, the loop closed offline with consent.
3. Fix the website. One page per product with the rate on it, calculators and eligibility before the application, speed on a phone, the licence and the people, the disclosures where they belong.
4. Earn the long tail and the answer. Product-and-segment pages, the guides for the question before the purchase, and the entity work that lets an assistant say your name.
5. Buy the decision precisely. Phrase and exact match, state by state, brand apart from generic, one page per ad group, manual bids until the account can learn.
6. Cut the audiences three ways and amplify. Prospect, remarket, close; social and retargeting funded on attributed conversions; premium environments bought with discipline.
7. Convert to the money event, then nurture to it. Applications that do not leak, a test between forms, a sequence inside the rules, a team sized to the volume.
8. Report five numbers, and move the ceiling with the settled results.

The engagements in this guide are where this order comes from. If you would like to see what the same order would look like for your products, the fastest way is a short call with the person who edited it.