inPowered AI

What is inPowered AI?

inPowered AI is a seller-side AI decision platform designed for brands, advertising agencies, and media providers. It deploys customized machine learning models on the ad exchange side to predict the likelihood of achieving specific business outcomes each time an ad is put up for bidding.

The platform organizes high-potential display opportunities into private market transactions, and provides a PMP Deal ID that buyers can use to activate them in existing DSPs. It is neither an ad copy generator for individual users nor merely an old-style tool for promoting content.

A one-sentence summary

inPowered AI uses real-time models on the exchange side to filter CTV, video, and display ad inventories; it then packages business outcomes such as brand enhancement, search performance, in-store visits, customer value, and attention into AI PMPs that can be activated directly in DSPs.

How does the seller’s AI decision-making system work?

Traditional DSPs primarily enable the buyer to control the audience, bids, frequency, and budget, whereas inPowered AI filters the supply before it enters the buyer’s bidding process. It has access to information related to the context in which ads are displayed, as well as data on auctions, supply, and attention levels.

Main functions

Display-grade real-time scoring

The model analyzes context, auctions, devices, geography, attention, and outcome signals at the ad display level, in order to determine whether a given opportunity is more likely to help achieve the desired goals. Decision-making takes place on the exchange side, making it suitable for large-scale real-time bidding.

Custom AI models

Each PMP can be trained using the specific KPIs and data of individual clients, rather than relying on a generic model to handle all advertisers. The quality of the training data, label latency, and business definitions have a direct impact on the results.

PMP Deal ID activation

The inventory identified by the model is delivered to the buyer via a unique PMP Deal ID, allowing the advertiser to add it directly to their preferred DSPs. Most solutions do not require any separate technical integration on the part of the buyer.

Continuous model learning

Brand enhancement plans are typically updated on a weekly basis using new research data, while strategies related to searches, visits to the store, customer attention, and interaction can be refined at a continuous or daily pace. The actual frequency of updates depends on how quickly the relevant data becomes available.

Cross-format media

The platform is designed for CTV, online videos, and display ads; it allows inventory to be filtered based on results in various media formats. The specific exchanges, regions, devices, and availability options need to be confirmed depending on the campaign.

First-party data training

Customer value and application solutions can be developed by using the advertiser’s own data on conversions, value, or revenue to train models. When transmitting data, it is necessary to avoid including any personal identification information in ads or tracking pixels.

Online reporting and activity analysis

Customers can view online reports that show various performance metrics such as views, clicks, interactions, time spent on the site, and conversions. The data in these reports is intended for internal use by the customers; sharing it with external analysis services is possible only under the terms of a contract.

Work in conjunction with measurement partners

Different outcome scenarios make use of professional measurement data, enabling the model to optimize more than just CPM or visibility rates. The partnership data made available by the platform includes those from Kantar, Cint, EDO, Innovid, PlaceIQ, and Adelaide.

Supported business outcomes

Brand enhancement

The brand enhancement model combines survey results, context, attention, and auction signals to predict which displays are more likely to enhance ad recall, awareness, consideration, and purchase intent. Measurements provided by firms such as Kantar and Cint can offer outcome labels for training and validation.

• It is suitable for brands that do not want to measure the upper stages of the funnel solely based on clicks and conversions;
• Brand research KPIs can be directly linked to media purchases;
• The model manages inventory in real time on the seller’s side and generates a Deal ID;
• The newly updated data can be used in the weekly retraining process;
• The study sample, significance level, and control group design still require independent review.

Search improvement

The search enhancement solution utilizes EDO search to analyze relevant data, in order to predict the likelihood that consumers will search for a particular brand, product, or category after seeing an advertisement. It is suitable for using proactive searches as indicators of brand interest and purchasing intent.

For searches of competing brands and for searches within specific categories, it is necessary to define a clear attribution window, in order to avoid attributing all natural trends or influences from other media sources to PMP. The feedback from the model should be analyzed in conjunction with the grouping of campaigns and the baseline period.

Visit the store in person

The in-store strategy combines location measurement data with contextual, auction-related, and attention signals, giving priority to those displays that are more likely to result in actual visits to the store. The in-store data can be fed into the learning cycle on a daily basis.

• Suitable for retail, food and beverage, automotive, consumer goods, and telecommunications brands with physical stores;
• It is possible to analyze customer flow data by store or event;
• Geographical accuracy, attribution window, and the baseline for natural in-store visits can affect the conclusions;
• Sensitive locations, small sample sizes, and cross-device matching require privacy controls;
• Visiting the store does not equate to a sale; it should be combined with sales efforts or testing strategies.

Application installation and subscription value

The solution in question does not focus solely on reducing installation costs; instead, it makes use of installation logs, in-app conversion data, and revenue information to identify users who are more likely to subscribe or make payments. The model helps to optimize in-app purchases, ROAS, and long-term value.

Games, finance, health, retail, and entertainment applications need to make use of mobile measurement partners, SKAdNetwork, or other attribution methods. Revenue reporting should be carried out using the permitted identifiers and aggregation techniques.

Long-term customer value

The customer value framework uses conversion logs, modeled value scores, and various indicators to identify opportunities with high LTV, repeat purchases, or a high ROAS. It is suitable for subscription services, as well as retail and DTC brands.

Companies should define in advance the criteria for value windows, returns, cancellations, gross profit, and repeat purchases. Short-term conversion rates may conflict with long-term value, so it is not sufficient to rely on only one indicator of immediate revenue for training purposes.

Attention and interaction

Attention is closer to actual engagement than mere visibility, but it still isn’t a direct substitute for indicators of brand enhancement or sales. The level of interaction is also influenced by creativity, placement, and differences among the audience members.

Activity implementation process

1. Identify the specific business outcomes that need to be optimized, rather than opting for CTR or the lowest CPM by default.
2. Select the measurement partner, data fields, attribution window, baseline, and success threshold.
3. Ensure compliant integration of first-party data, pixels, mobile attribution, or survey data.
4. Showcase models are trained using inPowered AI and deployed on supported exchanges.
5. Obtain a unique PMP Deal ID, and configure the budget, target audience, and frequency in the existing DSP.
6. Launch a small-scale test to monitor supply, delivery, pricing, and the feedback of outcome data.
7. After conducting comparisons and establishing a baseline, scaling up the deployment is carried out, while continuous monitoring of model drift and marginal returns is performed.

PMP activation process

1. Create or select the corresponding campaigns and ad groups in the DSP.
2. Add the PMP Deal ID provided by inPowered AI and confirm that it is visible on the exchange.
3. Retain the advertiser’s audience, creative elements, budget, geographic targeting, and frequency settings.
4. Check bid requests, conversion rate, CPM, supply quality, and budget schedule.
5. Verify whether the result data is returned as agreed and fed into the model for learning.
6. Avoid making significant changes to the creative, audience, and bid at the same time, in order to explain any changes in performance.
7. Decide whether to continue, adjust, or stop based on business results rather than just delivery volume.

Data collaboration and input signals

Relationship between exchanges and DSPs

The models of inPowered AI operate within exchanges or on the supply side, while the PMP Deal ID is entered by advertisers into the DSP of their choice. The ecosystems that are used for public display include Index Exchange, OpenX, Magnite, PubMatic, and Nexxen.

Which users is it suitable for?

• Brands and agencies that manage budgets for CTV, video, and display advertising;
• A team that focuses on optimizing the brand for improvement rather than merely focusing on clicks in the upper stages of the funnel;
• Retail and food service brands that have physical stores and place importance on verifying customer traffic;
• Marketing teams for mobile apps that focus on revenue, subscriptions, and LTV after installation;
• Companies that wish to use first-party conversion or value data to train media models;
• Media and exchange partners who wish to present high-quality offerings as result-oriented PMP services.

In what situations is it not very suitable?

• Only individual marketers who need to automatically write ad copy or generate images;
• Very small advertisers that lack programmed DSP, PMP, and measurement capabilities;
• Users who need to register on their own, have access to public pricing, and can make purchases using credit cards immediately;
• Teams that are unable to continuously send back result data or establish stable KPI metrics;
• It is hoped that this model will directly replace organizations responsible for brand safety, audience management, and creative content review.
• Platforms that require downloading models, viewing training code, or enabling private deployment must be purchased.

Product advantages

• The model makes real-time decision-making at the exchange level for display purposes, closely aligned with supply signals;
• Integrate with existing DSPs via Deal ID, thereby reducing the integration efforts required by the buyer;
• The optimization objectives cover brand, search, in-store visits, value, attention, and reach;
• It can be trained using professional measurement partners and first-party data from advertisers;
• CTV, video, and display ads can use the same result-oriented approach.
• Continuous feedback allows the model to adjust itself based on the performance in new activities;
• The seller’s curation can be combined with DSP parameters such as audience, creativity, frequency, and bidding amount.

Usage restrictions

• There is no fixed public price; the cost depends on the media, PMP, and measurement arrangements.
• The performance of the model depends on the quality of the result labels, the amount of data available, and the speed of feedback.
• Success stories of collaboration do not guarantee that the same results can be achieved in all brands, regions, and campaigns.
• The seller’s AI can only optimize the available supply; it cannot replace the entire media strategy.
• First-party data, location, and mobile attribution raise privacy and consent issues;
• Ongoing retraining may change the inventory structure, so it is necessary to monitor for drift and marginal returns;
• Brand security, fraud, creative quality, and incremental experiments still need to be managed separately.

Prices and billing

The price information was verified on August 23, 2026; the actual amounts, taxes, exchange rates, and discounts may vary, and the final figures will be those displayed on the settlement page.

inPowered AI does not have any publicly specified subscription fees or fixed PMP prices; access is obtained by requesting demonstrations and engaging in business partnerships. The costs are related to media purchases, exchanges, data partners, measurement services, and the specific solutions provided.

The old content promotion service used a CPE model based on billing per 15 seconds of interaction, but the current core products rely on seller AI and PMP decision-making. The historical CPE figures cannot be used as the uniform pricing standard for the present.

Privacy and data usage

The privacy policy was updated on March 28, 2025, and covers website accounts, contact information, devices, IP addresses, page visits, Cookies, as well as marketing purposes. The platform also processes data related to views, clicks, interactions, time spent on pages, and conversions as part of its activity analysis.

Regarding third-party AI models

The privacy policy specifies that user data shall not be shared with third-party tools, including AI models; it also allows service providers, affiliated companies, partners, and retargeting suppliers to process personal information under specified circumstances. Enterprises should define the protocols for processing data related to advertising campaigns as well as the processes involved in handling models, through specific data processing agreements.

Restrictions on personal identity information

The terms of service prohibit customers from using ad tags, pixels, or content distribution to insert personal identification information, and they also forbid combining campaign data with PII in order to re-identify individuals. Targeting based on children’s-related inventory and persistent identifiers of users under 13 is also restricted.

Contracts and data rights

API, SDK, and open-source status

inPowered AI offers enterprise integration for ad exchanges, DSPs, and data partners, but it does not provide any public developer API references, SDK packages, or official GitHub repositories. The platform and the models are considered proprietary commercial technologies.

Suggestions for measuring effectiveness

• Set a unique primary outcome and a limited number of secondary indicators for each activity;
• Establish a control group, a baseline period, or incremental experiments to avoid focusing solely on the differences before and after;
• Fix the attribution window, time zone, as well as return and cancellation rules;
• At the same time, monitor CPM, conversion rate, reach, frequency, and supply composition;
• Break down by creativity, equipment, region, and inventory to identify confounding factors;
• Record the model update date and the Deal ID version to ensure reproducibility of the results;
• Judgments are made based on profits, long-term value, or brand enhancement, rather than solely on the scores predicted by models.

Basic information

Frequently Asked Questions

Is inPowered AI a tool for generating advertising ideas?

The focus at present is not on creating copy and images, but rather on displaying the prediction results and filtering the available options on the exchange side. Creativity remains under the control of the advertisers or agencies.

Is it necessary to replace the DSP?

It is usually not necessary. Once the client obtains the PMP Deal ID, they can activate it within the existing DSP, while continuing to have control over the audience, budget, bids, creatives, and frequency of exposures.

Which ad formats are supported?

It currently covers CTV, online videos, and display ads. Native displays can also be used for interactive objectives; the specific options available depend on the campaign and region.

How often is the model updated?

The data related to brand enhancement can be retrained on a weekly basis, while feedback regarding visits to the store, level of attention, interactions, and outreach efforts is typically provided on a daily basis or continuously. The actual frequency depends on how quickly the relevant data becomes available.

Can first-party data be used?

It can be used for models related to customer value, revenue from applications, and other customized outcomes, but only the conversion or value signals permitted by the contract should be utilized. Ad pixels and tags must not be used to obtain personal identification information.

What is the price?

There is no fixed, public pricing; the costs for media, data, measurement, and modeling services are determined through business agreements. The prices associated with historical content marketing CPEs cannot be used as a reference for the current seller’s AI PMP services.

Does inPowered AI provide APIs?

The platform offers exchange and data integration capabilities, but it does not provide public self-service API or SDK documentation. The scope of technical integration needs to be determined as part of the cooperation agreement.

Is the platform open source?

No. Models, exchange decisions, and activity platforms are proprietary commercial technologies; there are no licenses available for their core code.

Can case improvements guarantee reproducibility?

It’s not possible. The degree of improvement is influenced by factors such as the brand, creativity, target audience, supply chain, region, budget, as well as the methods used for measurement and setting benchmarks; it should be verified through controlled tests.

Summary

inPowered AI integrates AI-driven decision-making into the programmatic advertising supply chain; it uses actual business results to train the models used for content delivery, and it packages the selected inventory as PMP Deal IDs. It is suitable for brands and agencies that already possess DSP capabilities, measurement tools, and the capacity to manage large-scale media campaigns.

The value of such a platform lies in its ability to enhance the buyer’s experience by connecting brands, search functions, in-store options, as well as various signals related to value and attention. When selecting a platform, it is important to assess factors such as data rights, measurement methods, the scope of coverage provided by the platform, actual costs, and the incremental benefits it offers, rather than relying solely on individual success stories.