# Smart-bidding doesn't read your business — it reads your conversions

> Target ROAS isn't broken — your conversion stream is. Here's why Smart Bidding optimises against the values you send Google, not your actual business outcomes.

**By Murtaza Rangwala** · **Published:** Apr 22, 2026 · **Read time:** 6 min read · **Category:** Bidding

Every other prospect call I take has the same opening: *"We tried Target ROAS for two months and it tanked our account."*

The conversation then goes one of two ways. Either the advertiser blames Google's algorithm and switches everything back to manual CPC. Or they double down, push the target higher, and slowly bleed budget while Google "learns."

Both reactions miss what's actually happening. **Smart Bidding isn't a personality. It isn't a strategy. It's an algorithm reading a single input — your conversion stream.**

If that stream is wrong, the algorithm is wrong. Not because Google is broken, but because Google is doing exactly what you told it to do.

!!!
**The TL;DR:** Target ROAS, Maximize Conversion Value, and every other "value-based" smart bidding strategy optimise against *the numbers you send to Google*, not against your actual business outcomes. Fix the input before blaming the output.
!!!

## What Smart Bidding actually optimises against

Here's the part most people misunderstand:

> Target ROAS doesn't optimise against revenue. It optimises against the **conversion value** you're sending it.

Those two things should be the same. In a lot of accounts I audit, they are not.

Google's algorithm sees:

- A **conversion event** (a "Purchase" tag fired)
- A **conversion value** attached to that event (a number, usually a `value` parameter)
- A **conversion action setting** that says whether to count this for bidding

That's it. That's the universe the algorithm operates in. It cannot see:

- Your gross margin on the order.
- Whether the customer refunded two weeks later.
- Whether this was a returning buyer or a first-time one.
- Whether the order was profitable for you.
- Whether the lead actually closed.

If your conversion value is "order total including delivery, taxes, and discounts," that's what the algorithm chases. Even if your real margin is 8% and varies wildly by product.

## The four most common ways the conversion stream is broken

In about 80% of the accounts I've looked at, at least one of these is happening:

### 1. All conversions weighted equally

Lead form submissions, newsletter signups, and PDF downloads are all counted as "conversions" with no value differentiation. Google then optimises for whatever is cheapest to generate, which is almost always the lowest-quality action.

The campaign looks fantastic in the dashboard. Conversion volume is up. Sales are flat.

### 2. Revenue is sent instead of profit

E-commerce stores plug in their Shopify or WooCommerce revenue feed and call it done. Google optimises to maximise gross revenue, so it pushes traffic toward:

- **High-ticket, low-margin items** (electronics, bundles, discounted SKUs).
- **First-time buyers with big discount codes**.
- **Promo-driven volume** that looks great in March and is gone by June.

Meanwhile your high-margin SKUs get less spend because their conversion values are smaller.

### 3. Refunds and cancellations are never sent back

Google fires a conversion when the order is placed. Most setups never fire a negative conversion or value adjustment when the order is refunded, cancelled, or charged back.

If your refund rate is 15-20% (normal for ticketing, fashion, beauty), Google is optimising on a 15-20% inflated picture of reality.

### 4. Offline conversions are missing entirely

This is the big one for lead gen, B2B, real estate, and ticketing with offline upsells. The form submission gets sent to Google. The actual sale, three weeks later, does not.

Google then optimises for whichever lead source produces the most *form submissions* — not the most *closed deals*. Junk leads from broad keywords often win. Quality leads lose budget.

## Why "the algorithm needs more data" is usually the wrong answer

When ROAS tanks, the default advice from agencies and Google reps is:

- "Give it more time to learn."
- "Increase the budget so it has more data."
- "Lower the target ROAS so it has more room."

Sometimes those are right. More often, they're treating a symptom.

If you're feeding Google garbage values, more data just teaches the algorithm to chase the wrong thing faster. The learning period isn't broken. The learning *material* is broken.

!!!
**The honest framing:** Smart Bidding is genuinely good at optimising. That's the problem. If you point it at the wrong target, it will hit the wrong target with impressive accuracy.
!!!

## What "good" conversion value actually looks like

The closer your conversion value is to your *real* business outcome, the better Smart Bidding works. A rough hierarchy, from worst to best:

- **Worst:** Every conversion = 1. (Counting form fills, sales, and newsletter signups identically.)
- **Bad:** Order total (revenue), no refunds adjusted.
- **Okay:** Revenue with refunds and cancellations pushed back as negative values.
- **Good:** Margin or profit, sent as the conversion value instead of revenue.
- **Better:** Margin + adjusted for refund probability by product category.
- **Best:** Predicted lifetime value (pLTV), informed by historical cohort data.

Most advertisers can move from "Bad" to "Good" in a week of dev work. That's usually enough to make Target ROAS go from "tanking the account" to "the best campaign you have."

## Use case: a live events ticketing company running Target ROAS

This is a composite based on patterns I've seen in multiple ticketing and live-events audits.

Picture a regional events company selling concert and theatre tickets. They run Performance Max and Search campaigns on Target ROAS at 600%.

**What they were sending to Google:**

- Conversion event: "Purchase" fires when checkout completes.
- Conversion value: `order_total` — the full cart amount including service fees, VAT, and add-ons.

**What was actually happening underneath:**

- Service fees were 12% of the order value but 100% of the company's actual margin.
- Roughly 11% of tickets were refunded or charged back (events cancelled, customers no-showed, plan changed).
- Bundled multi-ticket purchases by event organisers (B2B resellers) inflated AOV but generated near-zero margin.

**The result:**

- Target ROAS hit 600% on paper.
- Real ROAS on margin was closer to 80%.
- The algorithm was pouring spend into broad keywords that attracted resellers and bundle buyers, starving the higher-margin direct consumer keywords.

### What we changed

- Switched the conversion value from `order_total` to `(order_total - VAT - external fees) × historical_margin_factor_by_event_type`.
- Added offline conversion uploads for refunds, fired as negative values within 30 days of the original conversion.
- Excluded the B2B reseller checkout flow from the conversion action used for bidding.

### What happened over the next 6 weeks

- Reported ROAS in the dashboard *dropped* (because conversion values were now smaller).
- Real ROAS on margin *climbed* — from ~80% to ~340%.
- Spend redistributed away from high-volume low-margin keywords toward specific artist and event searches.
- Refund rate dropped because the algorithm stopped pushing impulse buys to the broadest audience.

We didn't change the bidding strategy. We didn't change the keywords. We didn't change the budget. We changed what we were telling Google to optimise for.

> Smart Bidding wasn't the problem. The brief we'd given it was the problem.

## How to audit your own conversion stream this week

If you run any value-based Smart Bidding strategy (Target ROAS, Maximize Conversion Value, tROAS in PMax), do this:

1. Open Google Ads > **Goals > Conversions**.
2. List every conversion action that has *"Include in 'Conversions'"* set to **Yes**.
3. For each one, ask:
   - What does this *actually* represent in the business?
   - Is the value field a real business outcome, or a vanity number?
   - Are refunds, cancellations, and chargebacks being adjusted back?
   - Is this conversion meaningfully different in quality from the others?
4. Pull the **Conversion lag** report (under Conversions > tools). If most of your conversions happen within an hour, you're probably missing the offline tail.
5. Cross-reference Google's reported revenue against your actual finance numbers for the same period. If they're more than 10% apart, the algorithm is operating on a different reality than you are.

Google explains the mechanics of value-based bidding in its [Smart Bidding help docs](https://support.google.com/google-ads/answer/7065882), and the conversion value rules feature is documented [here](https://support.google.com/google-ads/answer/10410266).

## The bottom line

Smart Bidding is not a magic strategy. It is a high-speed optimiser pointed at whatever target you give it.

- **Garbage values in** = optimised garbage out.
- **Real margin values in** = optimised real margin out.

The advertisers who say *"Target ROAS doesn't work for our business"* are almost always running it against a conversion stream that doesn't match their business. Fix the stream and the strategy starts working.

The algorithm isn't reading your business. It's reading your conversions. Make those two things mean the same thing.

---

**Sources and further reading:**

- [About Smart Bidding (Google Ads Help)](https://support.google.com/google-ads/answer/7065882)
- [Conversion value rules (Google Ads Help)](https://support.google.com/google-ads/answer/10410266)
- [About offline conversion imports (Google Ads Help)](https://support.google.com/google-ads/answer/2998031)

---

**Tags:** Smart bidding, tROAS, Conversion tracking, GA4

## About the author

Murtaza Rangwala is a senior independent Google Ads consultant. 8 years, 1,900+ campaigns shipped, $250M+ in client revenue generated. Independent practice capped at four concurrent clients.

- More posts: https://www.murtazarangwala.com/blog
- Book a 30-min call: https://calendly.com/murtaza_rangwala/30min
- Free Google Ads audit: https://www.murtazarangwala.com/#audit