Brand and interest affinity enrichment

What it does

Microsoft's first-party affinity enrichments score each unified profile for how strongly its demographic cohort skews towards a set of brands or interests you choose, using aggregated Microsoft online search data. It produces affinity scores and a share-of-voice split, written to separate tables and shown as charts on the profile page.

Key facts

The documentation for brand and interest affinity carries a prerelease notice, so treat it as preview and check current status before writing it into a scope of work.
You pick up to five brands or five interests per enrichment. Selection is either from a curated industry list, which is region-specific, or by typing your own with system suggestions; there is a feedback route to request missing items.
Supported markets are Australia, Canada (English), France, Germany, the United Kingdom and the United States (English).
Profile field requirements: at least one of date of birth or gender, plus country/region together with either city and state/province or postal code.
Date of birth must be a DateTime type or an ISO 8601 string — yyyy-MM-dd or yyyy-MM-ddTHH:mm:ss. Anything else, including UK-style dd/MM/yyyy text, will not map.
Output lands in dedicated tables: BrandAffinityFromMicrosoft and BrandShareOfVoiceFromMicrosoft for brands, InterestAffinityFromMicrosoft and InterestShareOfVoiceFromMicrosoft for interests.
Affinity is banded as Very high (85–100), High (70–84), Medium (35–69) and Low (1–34), with the underlying score on a 100-point scale.
Share of voice is scored 0–100 and totals 100 across the brands or interests in that enrichment, so it is relative to your chosen set rather than absolute.
Profiles outside the selected country/region can still be enriched where data exists for their own region — selecting Germany does not exclude a US-based customer.
Microsoft enrichments are included with the Customer Insights - Data subscription; no separate vendor contract is needed, unlike the partner enrichments.

When to use / skip

This is a good fit when a client has decent demographic coverage and wants directional targeting signals for campaigns — which cohorts skew towards which competitor, what interest categories to build content around. It costs nothing extra, the setup is an afternoon, and the charts on the profile page demo well.

Skip it when demographic coverage is thin. If most profiles lack date of birth or gender and have only a country, you will enrich a small slice and the client will conclude the product does not work. Also skip it when the client's expectation is individual-level behavioural truth. This is cohort modelling from aggregated search data; it tells you what people like this person tend to search for, not what this person did. If the stakeholder cannot hold that distinction, the feature will create more argument than value.

Configuration decisions

Brands or interests, or both as separate enrichments — they are configured independently and each carries its own five-item limit.
Which five items. Competitor set, own-brand-plus-competitors, or a category read; you cannot have all three in one enrichment, so decide what question you are answering.
Country/region for the enrichment, which also determines the curated industry list you can pick from.
Whether to run across the whole Customer table or a segment. Restricting to a segment is the sensible move when demographic completeness varies by source system.
Which profile fields map to age and location, particularly where you have several address fields and need to choose between city/state and postal code.
Output table naming, and whether segment and measure builders are expected to work against the affinity table or the share-of-voice table — they answer different questions.

Gotchas

Share of voice is relative. Add or swap a brand and every existing SoV value changes, which makes historical comparison across enrichment versions meaningless. Say this before anyone builds a KPI on it.
The five-item cap is per enrichment and it is not generous. Clients with a long competitor list will want several enrichments, each producing its own output, and someone has to keep them straight.
Date-of-birth format failures are silent in the sense that the enrichment just yields low coverage. Check the type in the unified table, not in the source.
Enrichment output sits in separate tables joined to the profile by customer ID. Segments and measures need to traverse that relationship — the fields are not on the Customer table itself.
Coverage is a demographic function, not a data-volume function. A million profiles with no gender or DOB will enrich barely better than a thousand.
Preview status means the market list, the brand catalogue and the scoring bands can move. Do not hard-code score thresholds into client-facing logic without a plan to revisit them.

Consultant notes

Frame it to the client as a propensity signal for cohort targeting, in writing, at design stage. The single biggest delivery risk here is someone in marketing reading "affinity: very high" as a statement about an individual.
Run a coverage check on date of birth, gender and postal code before you promise anything. A quick count of populated values across the unified table sets realistic expectations in five minutes.
Demo the profile-page charts alongside the timeline — the two together are what sells the unified profile story, and this one is free with the licence.
Use affinity as a segment refinement layer rather than a primary selector. "High-value customers who also skew towards brand X" holds up far better than segmenting on affinity alone.
Before go-live, check who is allowed to change the five selected brands. An unannounced swap silently rewrites everyone's share of voice.

Worth another look when these enrichments move out of preview, or if the supported market list expands.