Target ROAS for Hard Verticals: A Realistic Framework

Target ROAS sounds like a straightforward idea: tell Google what return on ad spend you need, and the algorithm bids accordingly. In e-commerce, this works. In regulated, high-CPC verticals — lending, mortgage, insurance, legal — most Target ROAS implementations are quietly broken, and the operators running them don't realize it for 60-90 days.

This is the working framework for setting and managing Target ROAS in hard verticals. It assumes you understand the basic mechanics; the failure modes I describe below show up after that.

Why Target ROAS breaks in regulated verticals

Three reasons, in order of how often they cause damage.

One. The conversion event Google sees isn't the event that pays you. In e-commerce, the conversion is the purchase, the value is the order total, and Target ROAS optimizes against revenue you actually received. In lending, the conversion Google sees is the application; the revenue happens 5-30 days later when the loan funds. In insurance, the conversion is the quote request; the revenue is the bound policy. In legal, the conversion is the intake call; the revenue is the signed retainer. The Target ROAS algorithm optimizes against the surface-level conversion and assumes its value is correct. In hard verticals, that assumption is wrong by a factor of 3-10x.

Two. Conversion values vary wildly within a single account by sub-product, geo, lead source, and time-of-day — variation that the algorithm has trouble learning fast enough. A funded auto loan in Texas has different unit economics than a funded auto loan in California. A purchase mortgage in a rising-rate market has different value than the same loan a quarter earlier. Target ROAS treats your historical conversion-value distribution as stable. In hard verticals, it's not.

Three. The conversion volume needed for Target ROAS to work is high. Google's official guidance is 50 conversions in the past 30 days per campaign for Target ROAS to bid intelligently. In regulated verticals where each conversion costs $80-300 and total deals per month per campaign might be 20-40, you don't hit that threshold. The algorithm bids on insufficient data and produces noise.

The combination of these three is why Target ROAS often underperforms manual or Target CPA bidding in hard verticals — not because Google's algorithm is bad, but because the inputs the algorithm needs are structurally not available in your business.

The unit-economics reality check

Before setting any Target ROAS, work out the math from the bottom up. Most operators set a tROAS based on what they want it to be, not what their economics actually allow.

The framework, by vertical:

Lending. Start with average commission per funded deal × your funded rate from clicks ÷ your target net margin. If your average commission is $4,000, your click-to-funded rate is 1.5%, and your target net margin is 25% (after all costs including operations and underwriting), then your maximum allowable cost per click is $4,000 × 1.5% × 75% = $45. Your target ROAS = revenue per click ÷ cost per click = $60 ÷ $45 = 133%. That's your tROAS ceiling. Set it lower than that and you're leaving margin; set it higher and you're losing money.

Mortgage. Same logic but with longer cycles. Average commission per funded loan × click-to-funded rate × (1 - operations cost ratio). The complication: the click-to-funded rate is the lagging indicator. You need 60+ days of attributed-back funded data to calibrate, which means in your first months you're guessing.

Insurance. Premium-to-commission economics. Average annual premium × your commission rate × retention multiplier (year 1 plus discounted future years) × click-to-bound rate. Insurance has the longest payback in this list — much of the value comes from year 2+ retention. Most operators understate their target ROAS because they only count year-1 premium.

Legal. Average case value × signed-rate from intakes × click-to-intake rate × (1 - operations cost ratio). PI has high case-value variance — a single mass-tort case can be worth 50x the average — so the math here is more about expected value over a portfolio of cases than per-conversion economics.

In every vertical the goal is the same: derive your tROAS from your true unit economics, not from a benchmark you read in a blog post.

Setting Target ROAS when you don't have downstream data

The honest situation for most operators in hard verticals: you're not going to have 50 funded loans per campaign per month for at least three months. So how do you start?

Three approaches, in order of preference:

Approach A: Target CPA on the surface conversion, manually manage ROAS in your own dashboard. For accounts under $50K/month spend, this often outperforms Target ROAS because the algorithm has insufficient data to bid well. You set Target CPA per the application/quote/intake event you can measure, and you track the funded/bound/signed metrics offline in your CRM. The bidding is dumber but the data feeding it is cleaner.

Approach B: Use Target ROAS with conversion-value adjustments. Tell Google that the application conversion is worth a fraction of the eventual deal — say 8% for lending (representing a 1.5% funded rate × $4K commission ÷ application count). Set your tROAS based on that adjusted value. The algorithm optimizes against a value that's closer to your true economics. The downside: you're still dependent on Google's volume-of-data assumption being met.

Approach C: Use Maximize Conversion Value with downstream conversion imports. This is the right answer once you have offline conversion imports working. You import funded deals (with actual commission values) back into Google Ads via the API. The algorithm bids against actual revenue. This works at moderate scale (50+ funded deals per month per campaign) and starts to dominate manual bidding around 100+ funded deals per month.

If you're under 50 funded deals per month per campaign, Approach A is almost always right. Above 100, Approach C starts to win. The middle range is where Target ROAS with adjusted values (Approach B) can work but requires close monitoring.

When Target ROAS actually works in hard verticals

Three specific situations where Target ROAS outperforms manual bidding even in regulated verticals:

Brand campaigns. Branded keywords have stable conversion rates and predictable values. tROAS works well here because the data is dense and stable. Set a high tROAS (say, 800%+) on brand and let Google bid down. You'll capture all your branded traffic at a low blended CPC.

Mature non-brand campaigns with offline conversion imports. Once you have 6+ months of funded-deal attribution flowing back to Google Ads, the algorithm has enough signal to bid intelligently. Target ROAS shifts from broken to better-than-manual.

Accounts with stable unit economics. If your funded-rate, commission-per-deal, and operations-cost ratios are stable quarter-over-quarter, Target ROAS converges to the right bid. Volatile accounts (rate-driven mortgage, rapidly changing insurance product mix, mass-tort case-value spikes) confuse the algorithm.

If your account doesn't fit one of those three profiles, manual or Target CPA is probably the right starting point.

The attribution window trap

A subtle failure mode that catches operators who switch to Target ROAS without thinking through the timing.

Google's attribution window for Target ROAS is configurable, but the default is 30 days. In lending and mortgage, your funded-deal data flows back over 5-60 days. If your attribution window is shorter than your funding cycle, the algorithm sees an artificially low ROAS and bids down — meaning you stop bidding on keywords that are actually profitable, because the funding hasn't shown up in the window yet.

The fix: extend your attribution window to match your funding cycle. For lending, set it to 30 days. For mortgage, 60-90 days. For insurance with bind-cycle delays, similar. For legal with intake-to-signed cycles of weeks, 30-60 days.

This is one of the most common reasons accounts switching to Target ROAS see performance crater for the first 60 days. The algorithm is starving the bid because it's measuring against a value horizon that's too short.

The conversion-value drift problem

Even with offline conversion imports working, conversion values drift over time. Rates move, commission structures change, lead-source mix shifts. Target ROAS uses your historical conversion-value distribution as the baseline. If that distribution drifts, the algorithm bids against stale economics.

The mitigation: re-calibrate your conversion values every 60-90 days. Pull your funded-deals data, recompute average commission, recompute click-to-funded rate, and update your conversion-value imports to reflect current reality.

Most operators don't do this. They set up offline conversion imports once, validate them, and move on. Six months later the values are stale by 20-40%, and the algorithm's bidding is mis-calibrated against current economics.

Worked example: lending operator switching to Target ROAS

A real-world flow we've walked customers through, sanitized:

Starting point. Personal loan operator, $80K/month spend, average funded commission $1,800, click-to-funded rate 1.2%. Running Target CPA at $35 against the application conversion. Funded-deal CPA is $290.

Step 1: Audit unit economics. Funded value $1,800. Funded rate from clicks 1.2%. Operating cost ratio 30% (collections, underwriting, ops). Target net margin 25%. Maximum allowable CPC = $1,800 × 1.2% × (1 - 30% - 25%) = $9.72. Hmm, current blended CPC is $14. Account is bleeding margin.

Step 2: Set up offline conversion imports. CRM webhook to Google Ads API on funded-deal events with actual commission values. Validate that funded deals are flowing back within 30 days.

Step 3: Switch to Maximize Conversion Value temporarily. Three weeks of bidding without a tROAS target while the algorithm learns the funded-deal value distribution.

Step 4: Set Target ROAS at the calibrated level. Computed tROAS ceiling from unit economics is 130%. Start at 110% (slightly conservative) for 30 days. Monitor.

Result at 90 days. Funded-deal CPA dropped to $190 (a 35% reduction). Volume of funded deals up 15% because the algorithm is bidding more aggressively on the keywords that produce funded deals and bidding down on application-only-noise keywords. Net account profitability up roughly 50%.

The whole change took 90 days from setup to validated improvement. Operators looking for a 30-day fix typically don't get the full benefit because the algorithm needs the data accumulation period.

What we'd do for your account specifically

The framework above is general; the application is account-specific. The audit we run on AiNeural customers in regulated verticals starts with: are your offline conversions correctly mapped, is your attribution window correct for your funding cycle, are your conversion values calibrated against current unit economics, and is your tROAS level derived from real economics or set by feel.

Most accounts we audit have at least two of those four wrong. Fixing them is 80% of the available improvement; choosing tROAS vs Target CPA is the other 20%.

If you want a free walk-through of your account against this framework — where the attribution math is broken, where the bidding is mis-calibrated, where the unit economics are bleeding margin — request a demo. 30 minutes on your account, no sales pitch.