InMobi today unveiled Mobius, a single system that chooses which advertising options run across its exchange, its buying platform and the Glance consumer surface, and that keeps a written record of why each choice was made. According to InMobi, the system handles more than 25 trillion predictions a day and orchestrated a 12-day Grocery Outlet campaign that reached 8.6 million unique shoppers.

In Short

InMobi today unveiled Mobius, one system that picks which ads run across its exchange, its buying platform and the Glance app, then checks the results against its own forecast. It matters to brands, agencies and app publishers that buy or sell ads through InMobi, because the company says every choice is logged with a reason and the way success is measured is fixed before any ad runs. What changes is that InMobi's ad products now run on this single system, although the results shown so far come from InMobi itself and from one retailer.

What Mobius is, according to InMobi

The release, issued from San Mateo, California at 12:30 PM Eastern Daylight Time, describes Mobius as a decisioning layer that sits across InMobi's entire advertising stack, from the Exchange and the demand-side platform to the ads that appear on Glance. According to InMobi, the system weighs the options eligible for a given opportunity against an outcome the client has specified, picks what runs, and then feeds measured results back into later decisions.

The scale figures come first in the company's account. Mobius processes an average of more than 25 trillion predictions across signals each day, according to InMobi, in order to serve an average of more than 200 billion advertising opportunities daily. Both numbers are stated as floors, so any ratio is approximate, but together they imply something on the order of 125 predictions for each opportunity. The release does not define what counts as a prediction, how many models produce them, or what the latency budget for a single decision is.

Three tiers of signal

InMobi sorts the data feeding Mobius into three groups. Zero-party signal comes from the company's own consumer survey product, in which people state preferences and intent directly. First-party signal comes from two surfaces: the weather app 1Weather, which InMobi operates, and Glance, an AI-powered commerce platform that InMobi backs. Third-party signal, in the company's description, comes from more than 70,000 publisher apps that have integrated the InMobi SDK, from pixel data across the Exchange and the buying platform, from licensed data, and from direct ecosystem alliances.

One detail in the boilerplate bears on how that first tier is read. The release calls Glance an independently operated consumer platform, owned and run by Glance InMobi Pte. Ltd., which it describes as an associate of InMobi Technology Services Pte. Ltd. Yet the same release lists Glance as a source of first-party signal. It does not say how data moves between the two entities, or on what terms.

PPC Land has tracked that relationship before. When InMobi opened its premium connected TV and mobile in-app supply to buying agents on Scope3 Interchange on June 3, 2026, the company said, according to InMobi Advertising, that the connection made roughly 2 billion mobile users and Glance's 300 million active devices reachable through a single protocol connection. Scope3 has since been renamed Apostra.

The decision sequence

From the advertiser's side, each decision starts with the outcome the advertiser wants, according to InMobi. The system then assembles the audience, supply and creative components needed to deliver it and writes what the company calls a decision trace: a record of what ran and why. Permission, the release says, travels with the data, so the rights a brand holds over its data set the limits of what Mobius will consider.

Measurement follows the same order. For each campaign, the KPI, the measurement partner, the attribution rule, the holdout and the evidence window are fixed before media runs, according to InMobi. Mobius then compares measured outcomes with its own forecast and routes the evidence back into the decisions it can influence.

Abhay Singhal, co-founder of InMobi, framed the design in two statements. "Advertising systems are being asked to make increasingly complex decisions in milliseconds, with audience, media, creative and business outcomes all informing the choice," he said. "Mobius brings those decisions into one governed system. It forecasts what is likely to work, determines what should run and records the basis for that choice, so the next decision can start with evidence from the one before it."

On measurement he added: "Measurement and accountability need to be part of the decision system itself." He continued: "With Mobius, the outcome and the method used to evaluate it are established before the media runs. The system can then compare what happened with what it predicted and use that evidence to improve subsequent decisions."

Objectives covered

The release lists three objective families. For brand-building campaigns, Mobius can optimise toward ad recall and brand lift. For consideration, it can work against consideration lift and qualified visits. For performance campaigns, it can optimise toward return on ad spend and longer-term return on ad spend, using measurement sources suited to each objective. InMobi also says the system is designed to improve yield for publishers and traffic shaping for partners that run buying platforms of their own, and it argues that serving both sides of the market from one intelligence layer is what separates a full-funnel claim from a list of separate products.

The campaign figures

Two advertisers appear in the release. The first is LinkedIn, whose campaign promoting learning content used Mobius to find and reach the users most likely to engage and to move them into longer sessions. The release gives no numbers for it, stating only that the partnership delivered value for users and better outcomes for LinkedIn.

The second is Grocery Outlet, a supermarket chain, and here the release is specific. Its Founders Day campaign, orchestrated end to end through Mobius, reached 8.6 million unique shoppers across CTV and mobile over a 12-day flight. CTV video completion peaked at 98.7%, against a blended average of 82.7% across both channels. An interactive wipe-to-reveal unit produced a 6.6% wipe-through rate to a store-locator call to action. An AI companion that turned the end of each video into a two-way conversation about nearby stores and deals saw 35% engagement and a threefold lift in click-through rate.

Jim Porterfield, chief marketing officer of Grocery Outlet, said: "This campaign showed us what is possible when you pair the power of the Grocery Outlet brand with a truly intelligent platform." He added: "Beyond putting Founders Day in front of millions of shoppers, Mobius helped us build experiences that got people to engage, discover and act. That is exactly the kind of smarter, more captivating marketing we have been looking for, and it is changing how we think about turning a great deal into a great experience."

The numbers invite a few observations. The 16-point gap between 98.7% and 82.7% compares a peak, apparently for CTV alone, with an average taken across CTV and mobile together, so it is not a like-for-like measure of improvement. The threefold click-through lift has no stated baseline. And all of the figures are supplied by InMobi, with a client quote attached, rather than by an outside measurement partner.

What the announcement leaves out

The release is a product statement, not a technical paper, and several matters a buyer would ask about are absent. It names no pricing or fee structure for Mobius. It gives no count of advertisers or publishers using it, beyond the statement that the system is powering InMobi's advertising solutions today. It does not say which KPI, measurement partner, attribution rule or holdout design applied to the LinkedIn or Grocery Outlet campaigns, even though it presents those parameters as the foundation of every campaign. It does not say whether measurement partners are independent of InMobi. Nor does it explain how permission is attached to data in practice, which consent signals or standards are involved, or which regions are covered.

The same applies to the decision trace. The release says one exists; it does not say what fields it holds, who can read it, or whether an advertiser can export it. Audit records of this kind have begun to matter in practice. A group director at Rise, part of the Quad agency group, has been running media-buying agents for a supermarket client since June 2026 and uses the audit log PubMatic built into its platform to watch for drift.

How Mobius fits InMobi's 2026 sequence

The release does not mention earlier products, but the timing places Mobius within a run of InMobi announcements this year. On March 3, 2026, Kochava added InMobi to its Certified Partners Program alongside LG Ad Solutions, Digital Turbine, Jampp, Adikteev and Appnext. On May 28, an Amazon Publisher Services bridge running through the InMobi SDK gave app publishers access to additional bidders, and four unnamed gaming publishers in the beta reported revenue growth of as much as 14 times on that stream.

On July 24, InMobi released FRAME Attention, a certified attention product with a 3.5 times claim against equivalent YouTube video ads and validation from Lumen Research. And on August 13, 53 days before Mobius, InMobi Advertising placed a conversational agent inside Buyer Hub and opened that deal platform to agencies and managed-service buying platforms. According to InMobi Advertising at the time, that agent ran on what it called a unified intelligence layer, and traders kept the final decision on every recommendation. The Mobius release speaks of the same intelligence layer serving each side of the ecosystem, but it does not state whether Mobius is that layer. The overlap in terminology is plain; the relationship is not documented.

The supply footprint beneath all of this is considerable. In DataBeat's latest ads.txt sample, InMobi ranked fifth among supply-side platforms for lines added, with 4,088, and was one of only two companies in the top five of all three traffic bands. Older history explains how the Exchange got there: InMobi agreed to buy AerServ for $90 million in cash and stock in January 2018, and it later secured $100 million in debt financing from MARS Growth Capital in September 2024 to advance its AI capabilities. In between, a 5% workforce reduction in February 2024 came with a stated intention to keep expanding the buying platform after the Appsumer acquisition.

The release also sits beside other InMobi items listed on the same newsroom page: a free resource called "Ready, Set, Agent", which scores retailers on agentic commerce readiness, and a Glance experience that brings personal shopping help to iPhone and iPad users through Siri AI. Details of both are not in the Mobius release. Siri AI itself began reaching consumers on September 14, 2026 as an English-language beta, absent from iPhone, iPad and Apple Watch inside the European Union.

The wider evidence on automated buying

Mobius arrives in a market where claims about agentic AI outrun independent proof. Two-thirds of advertisers were prioritising agentic AI for campaign execution by January 2026, according to IAB research, while Gartner expects more than 40% of agentic projects to be cancelled by the end of 2027. Independent pricing data is thin. DataBeat's June 2026 report found conventional programmatic buyers clearing at $6.95 CPM against $6.13 for agentic buyers, a 13.4% gap, with agentic demand entering 86% fewer auctions.

Spend forecasts are modest. Michael Barrett, Magnite's chief executive, put the most optimistic 2027 forecasts for protocol-based agentic spend at $600 million to $700 million. Competitors have also moved toward integrated systems: PubMatic unveiled AgenticOS on January 5, 2026 with live campaigns already running. Results from such systems, like those in the Mobius release, are largely vendor-reported and drawn from individual campaigns.

Why this matters for the marketing community

Three points stand out for advertisers, agencies and publishers.

First, the sequencing of measurement. The release commits to defining the KPI, partner, attribution rule, holdout and evidence window before media runs. If that is applied as described, it moves argument about how success is judged from after the campaign to before it. Whether it is applied as described cannot be checked from the release, because the two named campaigns do not disclose their parameters.

Second, concentration of roles. InMobi runs an exchange, a buying platform and, through an associated entity, a large consumer surface, and Mobius makes decisions across all three while also forecasting and then measuring the outcome. A single system that predicts, selects and reports invites the question of who audits it. The release's answer is the decision trace and the measurement partner field; the details of both are what an outside reviewer would need.

Third, the shift in vocabulary from optimising bids to governing decisions. Describing a system as one that "records the basis for that choice" echoes a wider push, visible in audit-log products from other vendors, to make automated buying inspectable. For publishers, the promise of better yield is stated without a number. For agencies, the Grocery Outlet results show format and completion data but no incrementality figure, which is the measure that holdout designs exist to produce.

InMobi said it will be at Advertising Week New York, which runs from October 5 to 8, and that a video and further material on Mobius are available through its release.

Timeline

Summary

Who: InMobi, the Singapore-headquartered advertising and commerce technology group, with Abhay Singhal, its co-founder, as the named company spokesperson. Grocery Outlet, represented by its chief marketing officer Jim Porterfield, and LinkedIn appear as advertisers. Glance and 1Weather are the consumer surfaces named as signal sources.

What: Mobius, a unified decisioning system covering InMobi's Exchange, buying platform and Glance. According to InMobi, it evaluates eligible options against a client-defined outcome, selects what runs, writes a decision trace, fixes measurement parameters before media runs, and compares results with its own forecast. The company cites more than 25 trillion predictions and more than 200 billion advertising opportunities per day.

When: The release is dated today, with distribution at 12:30 PM Eastern Daylight Time. InMobi said it will attend Advertising Week New York from October 5 to 8.

Where: Issued from San Mateo, California, with the product operating across InMobi's global advertising solutions and the Glance and 1Weather surfaces. Advertising Week New York is the named event.

Why: According to InMobi, advertising decisions increasingly weigh audience, media, creative and business outcomes at once in milliseconds, and measurement needs to sit inside the decision system. The campaign results supplied are vendor-reported and lack disclosed measurement parameters, so the claims remain to be tested outside the company.