Introducing Bigabid Alta:
Scale Retargeting and Exceed ROAS Targets with Neural Networks

How Bigabid Alta, our platform based on neural-network models, lifts performance across core ROAS metrics, and what it delivers for our advertisers.

Bigabid Research | 6 min read | Nov 2026

ON THIS PAGE

What changed

What Bigabid Alta delivers

Reaching the users others pass on

The results

What it means for your account

Built around your goals

After three years of development, Bigabid Alta is here: a platform based on neural-network models that evaluate every auction in real time and keep learning as the market moves. This article covers what changed, what Bigabid Alta delivers, and what it means for your account.

What changed

This year Bigabid Alta went live across our retargeting accounts. It replaces the traditional ML models we ran before, and it changes how every bid gets decided.

The difference is in how decisions get made. Our traditional ML models recalibrated monthly, creating pricing gaps on non-standard inventory. Bigabid Alta’s neural networks evaluate every auction dynamically in real time. They learn the pricing directly from the data, across far more signals at once, and they keep relearning. Every impression is valued on its own merits: the user, the app, the format, the market, the moment.

User acquisition led the industry’s shift to deep learning. Retargeting is where it now pays off most, because retargeting lives on precision: the right existing user, at the right price, at the right time.

What Bigabid Alta delivers

Sharper pricing
Every impression is priced with granular precision. In August and September, more than 70% of our retargeting accounts beat their Day-7 ROI target. Better pricing ensures campaigns win high-value users efficiently without overpaying.
Every genre, one engine
The same models price a casual puzzle title and a complex mid-core game. Nothing has to be rebuilt per genre, so a new vertical starts from a model that already understands how users behave, not from a blank page. That opens verticals we could not serve well before.
Every market, its own calibration
US and Japanese audiences are priced on their own parameters, not on one set of assumptions stretched across every market. Each geography gets the pricing its users actually justify, which is what makes expansion beyond your core markets workable.
Faster learning
Models now retrain in days instead of monthly. When something changes in your game, in your audience or in the competitive market around you, the models adjust in days rather than weeks, so you spend less time paying for yesterday’s conditions.
Scale without compromise
Our platform processes over 2 million auctions every second at low latency, and the models keep their precision at the full speed of live bidding. Accuracy does not degrade as volume grows, which is what lets us bid across a far broader share of your target audience.
Every format, not just full-screen
Our older models could only price the most expensive placements with confidence. Bigabid Alta prices banners, native and other formats properly too, which is most of the inventory in the market. Your campaign can now reach the same users in far more places.
Far more reach
By accurately pricing a wider spectrum of inventory, our retargeting impression volume roughly tripled between May-June and August-September, leveraging supply already available in the market.

Reaching the users others pass on

Retargeting depends on reaching your existing users, and not all of them are equally easy to reach or to price. Some have been away for a long time. Some sit in spending segments that are hard to read. Some are only visible on inventory most models will not touch.

Most platforms play it safe. They bid on the easy slice of your audience, the users everyone else is also bidding on, and pay a premium for it.

Bigabid Alta prices the rest of your audience accurately too. That turns the harder-to-reach part of your audience into reachable, profitable inventory, and it is why retargeting gains more from this shift than any other channel.

The results

Comparing May-June with August-September 2026, across the same retargeting accounts:

Day-7 ROAS up 50%+. ROAS is our north star, and it rose sharply. In some accounts, performance now beats client KPI targets by 20% to 30%.
CPM down by about half, in close to 90% of accounts, as campaigns reach users across far more of the inventory available.
CPI down about 30% overall, because we win the same users for less instead of overpaying for the few we could always identify.
CPA down in most accounts, as more of the impressions we buy reach users who go on to pay.
Room to scale. Accounts can absorb 2x to 3x their previous budget, because the models keep pricing well as spend grows instead of running out of users worth buying.

What it means for your account

Android:
a High-Growth Retargeting Channel. On Android, Day-7 ROAS rose up to 60%, while CPI fell by more than 40% and CPM dropped by roughly 60%. As a result, more Android accounts consistently exceed their ROI targets. With acquisition costs reduced significantly, Android has evolved into a major growth engine for retargeting.
iOS:
Powerful Lift, Targets Beaten. Performance on iOS mirrors this success, with Day-7 ROAS also climbing up to 60%. Close to 90% of iOS accounts now beat their ROI targets, up from just over half.
More budget headroom:
Advertisers that used to hit a ceiling can now grow, and apps that were too small to run efficiently can run.
More consistent returns:
Weekly retraining and per-market calibration mean performance holds as conditions shift, rather than drifting until the next monthly reset.
Wider coverage:
New geographies, new genres and new formats come within reach without starting from scratch.

Built around your goals

ROAS is our primary optimization metric. We work hand in hand with your team to tune campaigns directly to the targets you are measured on, backed by our dedicated white-glove service.

Bigabid Alta optimizes precisely toward the targets you set. The more exact your KPIs (thresholds, attribution windows, revenue goals), the harder it works for you. Send them to your account team, and we’ll build the retargeting strategy that scales your ROAS.

Want to master retargeting methodology?

Explore Bigabid’s RT 101 course for deep dives on holdout control groups, optimal timing, and performance benchmarks.

Introducing Bigabid Alta: Scale Retargeting and Exceed ROAS Targets with Neural Networks

How Bigabid Alta, our platform based on neural-network models, lifts performance across core ROAS metrics, and what it delivers for our advertisers.

What changed

What Bigabid Alta delivers

Reaching the users others pass on

The results

What it means for your account

Built around your goals

Want to master retargeting methodology?

Results based advertising starts here.

Introducing Bigabid Alta:
Scale Retargeting and Exceed ROAS Targets with Neural Networks