How do lookalike audiences work in India CTV advertising?

A lookalike audience is a modelled segment that expands a seed set of known high-value users — typically customers, converters, or app users — by finding other devices with similar behavioural and demographic characteristics. In CTV, the process works at device ID level: the advertiser uploads a seed list (CRM hashed emails or device IDs), the DSP or data platform matches the seed to its CTV device graph, profiles the common characteristics, and generates a larger audience of CTV devices with similar profiles. Lookalike audiences allow advertisers to extend beyond the seed group while maintaining intent quality — useful when the seed is too small for cost-efficient CTV targeting on its own.

CTV lookalike construction in India follows a three-step process: (1) Seed matching — the advertiser’s CRM data (hashed emails, phone numbers) is matched against the DSP’s identity graph to identify matched CTV device IDs. In India, match rates against CTV device IDs are typically 15–30% of the seed, due to fragmented device ID namespaces and lower CRM-to-CTV household overlap. (2) Profile modelling — the DSP profiles matched seed devices on signals like content genre preferences, app usage, geographic cluster, and viewing time. (3) Audience extension — the model scores all available CTV devices in the DSP’s graph and returns the top-N percentile as the lookalike audience. Minimum seed size for a reliable India CTV lookalike: 5,000–10,000 matched devices.

CTV lookalike audiences are generally less accurate than mobile lookalikes for two reasons: (1) Smaller device graph — India’s addressable CTV device universe (50–70 million devices) is 15–20x smaller than the mobile device universe, which limits the statistical depth of modelling. (2) Weaker signal diversity — mobile devices carry richer signals (app install data, location history, search behaviour) than CTV devices, which are primarily characterised by content viewing and sparse demographic data. For India CTV campaigns, lookalike audiences work best when: the seed is behaviorally distinct (e.g., premium OTT subscribers, not all TV viewers); the campaign objective is awareness or brand lift rather than direct response; and the lookalike is validated with a holdout test before full budget deployment.