Interview presentation · November 2026

Measurement &
Audiences Strategy

Samba Measurement & Audiences: Product Assessment Framework and Vision

1. Assessment  ·  2. Measurement  ·  3. Audiences  ·  4. Roadmap & P&L  ·  5. Team

Sr. Director, Measurement & Audiences PortfolioUse ← → or the tabs to move between chapters

Vision statement

Build the measurement and AI systems advertisers use to decide where ad dollars go.Those dollars fund the free press, art, and entertainment we all love, and TV is where much of that money is still decided.Samba is the independent layer that makes those decisions accurate, privacy-durable, and actionable.

Product principles

01

Trust over short-term wins

Trust is earned over time. For a measurement company, trust in the methodology is the product.

02

Simplicity

Depth when needed. The common case stays effortless, and the complexity is available behind ellipses.

03

Ship to learn

Small, testable releases over big launches. Every release should teach us something.

Assumptions

  • Everything here comes from public sources / assumptions. Treat it as a set of hypotheses to test by Day 30.

If only three things

What I prioritize

01

Systems thinking

Start from first principles, zoom out by default, go deep when it matters.

Know every team's incentives.

02

Platform mindset

Build for the self-serve, automated future.

Design every product with awareness of the fully automated future.

03

Process focus

Run repeatable playbooks, adapt when reality changes.

Run empowered teams. Use templates and guardrails so others move fast without breaking quality.

Why now

TV measurement is fragmenting and closing at the same time

US CTV ad spend, 2026

$38B

Up ~14% year over year

CTV upfronts, 2026

$17.7B

First year above primetime linear ($17.0B)

Agentic buying

~2/3

of US ad buyers plan to focus more on agentic buying in 2026

Cost of duplication

-43%

lower cost per new household reached when Samsung Ads cut duplicate exposures 25% for Skip

Sources: eMarketer CTV forecast (via Adwave) · eMarketer upfronts · IAB 2026 Outlook Study (n=205 buyers) · eMarketer, Samsung Ads x Skip (Oct 20, 2026)

Market forces

What's changing

01 · Spend

Spend is shifting to streaming

US CTV reaches ~$38B in 2026; CTV upfronts pass primetime linear for the first time.

So what

Buying stays TV-shaped: buyers need reach and frequency across every screen.

02 · Currency

The currency referee is weakening

The JIC recertified Comscore, iSpot and VideoAmp; VideoAmp and Nielsen One Ads left the MRC in 2026.

So what

Buyers run several currencies and need a neutral layer that reconciles them.

03 · ACR

ACR is closing into walled gardens

Samsung, LG and Walmart-owned Vizio keep viewing data for their own ad businesses.

So what

Independent, cross-OEM data is worth more, and is at more risk.

04 · Consent

Consent goes opt-in

Texas made Samsung (Feb) and LG (May) get express ACR consent; more states may follow.

So what

Footprint durability is a product requirement, not a legal footnote.

05 · Agents

Activation is going agentic

AdCP and IAB Tech Lab specs define how buyer agents find and activate audiences.

So what

If agents can't call it, they buy around it.

TakeawayThe winners will be independent, privacy-safe and tied to activation. Samba has the pieces.

Sources: eMarketer · MediaPost (JIC) · MediaPost (MRC) · Texas AG and Samsung · Texas AG and LG · Samba newsroom (AdCP)

In the news

Recent market events

Oct 2026 · Netflix

The premium streamer missed its ad target

Netflix cut its 2026 ad outlook to ~$2.7B. Buyers told Variety it priced too high and tried to build its own ad stack first. It's now buyable through The Trade Desk, Amazon, DV360, Yahoo, Magnite and Microsoft.

For Samba · Closed tech slows adoption, and premium sellers need independent proof to defend price.

2026 · Streaming goes programmatic

More premium inventory is biddable

Warner Bros. Discovery is opening more live sports to programmatic buying, and the big streamers are now selling through the major DSPs rather than only direct.

For Samba · Buying runs through a few pipes, so measurement has to sit across all of them.

2026 · Apple

Apple is buying live sports, and sports is where it sells ads

Apple took exclusive US Formula 1 rights from 2026 (reported ~$140M a year for five years) and moved every MLS match into the base Apple TV subscription. Live sports are the only place Apple sells ads today. Analysts expect the bigger audiences to grow that ad business, even though Apple says no ad tier is planned.

For Samba · Another premium sports seller with no TV glass of its own will need independent reach and outcome proof to price its ads.

Oct 2025 → now · Agents

Agentic buying gets a protocol

The Ad Context Protocol (AdCP) launched with Samba as an early member, and the IAB Tech Lab published agentic specs.

For Samba · Measurement and audiences need an API surface agents can call.

Variety via Yahoo Finance (Oct 27, 2026) · Digiday · Engadget (F1) · PCWorld (MLS) · Benzinga via inkl (ads outlook) · Samba newsroom: AdCP

The data moat

Why Samba

There are two kinds of data moat. Samba has the first and can build the second.

Moat 1 · Data control

Sole control of a critical asset

Uniqueness, aggregation, records. For Samba: opted-in ACR viewing across 24 TV brands, web signal from Semasio, and an identity graph of 1.5B profiles.

Moat 2 · Data loops

Every use strengthens the core

Campaigns get measured and activated, outcomes flow back, and the models and segments get sharper. That's the flywheel this portfolio should own.

Framing after Pivotal, "Economics of data businesses"

How it fits together

Data in, value out

01 · Data in

Smart-TV ACR24 TV brands, opted-in viewers
Web (Semasio)2.5B pages analyzed a month
Partner dataSet-top box, retail, TiVo
Client first-partyOnboarded through Project Gravity
02 · Samba core Knowledge graph
+ Samba ID 1.5B profiles · one exposure spine

03 · Value out

Focus of this roleMeasurementReach & frequency, outcomes, lift
Focus of this roleAudiencesSegments, Gravity lookalikes
Integrations40+ platforms, DSPs, clean rooms
Samba AI & agentsBridge, APIs, AdCP

The LLM-era point

LLMs have made front-end insight cheap. Companies that only connect to other people's data will struggle; owned signal is what stays scarce.

Product and partner facts from samba.com and AdExchanger (Semasio).

From ad to decision

How the data turns into decisions

Samba sees the ad on the TV, collects exposure data, classifies it and connects it in one graph. Buyers and sellers each get something different out of it.

  1. Auto brand0:30 spot · Linear
    Streaming promo0:15 spot · CTV app
    Retailer0:30 spot · Live sports
    Ad detected

    01 · See

    Every ad on the TV

    ACR on 24 TV brands detects each ad that airs, linear and streaming.

  2. imp · CTV app · hh_8f2 · 21:04

    imp · web video · hh_1c9 · 21:04

    imp · mobile · hh_8f2 · 21:05

    imp · CTV app · hh_77a · 21:05

    imp · web · hh_3d0 · 21:06

    imp · CTV app · hh_8f2 · 21:04

    imp · web video · hh_1c9 · 21:04

    imp · mobile · hh_8f2 · 21:05

    imp · CTV app · hh_77a · 21:05

    imp · web · hh_3d0 · 21:06

    02 · Collect

    Exposure data from clients

    First-party impression logs from advertisers and publishers: CTV apps, web and mobile.

  3. Genre · dramaTopic · travelBrand · autoLogo · on screenShow · live sports

    03 · Classify

    Tagged by Semasio and Samba AI

    Every show, page and ad gets tagged by genre, topic, brand and logos.

  4. 04 · Connect

    Connected in one graph

    Samba ID ties every exposure to a household, across screens.

Metrics & insights

Buyside · Agencies & brands

Main audience buyer

Spend smarter while the campaign runs

They see

  • Unique reach and frequency
  • Incremental reach by screen
  • Outcomes and lift

They act

  • Cap homes that see the ad too often
  • Reach in-market homes the buy missed
  • Target viewers of a competitor's ads
  • Move budget mid-flight

Sellside · Networks & publishers

Show what the inventory is worth

They see

  • Who their audience really is
  • Reach they add beyond linear
  • Outcome lift vs. competitors

They act

  • Back a higher CPM with independent proof
  • Benchmark against competing networks
  • Package curated audience deals
  • Defend price in upfront and programmatic deals
The loop Better ROAS→More budget into TV→More ads on screen→More data for Samba

ACR footprint, products and partners from samba.com; Semasio per AdExchanger.

1

Portfolio assessment
& priorities

How would you assess the current state in your first 30 to 60 days? What would you look for to understand portfolio health? What are likely to be your top 3 to 5 priorities?

Acquisitions and launches

Samba timeline

Oldest → newestAcquisitionLaunch or partnership

2008

Founded

Smart-TV ACR viewing data

The data asset everything else is built on

Dec 2018

Screen6

Cross-device identity graph

Ties TV exposure to homes and devices

Aug 2019

Axwave

Fingerprint ACR for ads and content

Automatic ad detection on the glass

Nov 2022

Disruptel

On-screen AI: logos, products

Powers sponsorship and placement measurement

Oct 2024

Semasio

Contextual web understanding

Privacy-durable audiences beyond TV

Jun 2025

Acxiom

Audiences in IPG's Interact, 40+ markets

Global distribution through a holdco

Oct 2025

AdCP

Early member of the Ad Context Protocol

A seat at the agentic standards table

2026

Samba TV → Samba

Rebrand to media intelligence

Positioned beyond TV

May 2026

Project Gravity

Client data onboarding and lookalikes

Opens the graph to first-party data

Jun 2026

Bestever AI

GenAI creative and agentic planning

AI product leadership joins

Sep 2026

X partnership

Tune-in measurement

X saw a +55% tune-in lift

Sources: AdExchanger (Axwave) · AdExchanger (Semasio) · Samba (Disruptel) · Samba (Bestever AI) · MarTech Series (X) · samba.com

Portfolio today · My read from public sources

How the products fit together today

Client surfaces

Analytics DashboardR&F, conversions, tune-in
Real-time viewershipLive, by market and demo
BridgeOne client portal, in build
Activation40+ platforms, DSPs, clean rooms

Products

Measurement FOCUS OF THIS ROLEKnow who you reached, how often, on every screen

Reach & frequency · Entertainment outcomes · Digital outcomes & lift · Sponsorship & placement · Industry data inputs

Audiences FOCUS OF THIS ROLETurn TV and web behavior into reach

Behavioral & contextual · Competitive conquesting · Incremental reach · Tentpole retargeting · Project Gravity (new)

Platform

Knowledge graph1.5B profiles
Samba IDTies exposure to homes and people
Samba AIShows, ads, logos on screen

Data

Smart-TV ACR24 TV brands, opted-in viewers
Web (Semasio)2.5B pages analyzed a month
Partner dataSet-top box, retail, TiVo

To validate in the first 30 days: which features clients actually use, what each earns, and where the acquired stacks still run separately. Sources: samba.com; Project Gravity (May 2026).

Pragmatic Framework

Public data assessment

18 of the 37 boxes can be filled from public sources (highlighted). Key reads below; all 18 are in the appendix.

Framework: Pragmatic Institute. Faded boxes need internal data.

Scored at Day 30

Portfolio health

Reach & frequencyOutcomes & liftAudiencesProject GravityIndustry dataSamba AI
LensWhat I'd look atHealthy looks likeScore 1–5
CommercialGrowth, NRR, client concentrationNRR above 110%; recurring revenue rising; no client dominatesDay 30
Customer valueAdoption, usage depth, renewal driversClients act on our data in-flight; renewals without discountingDay 30
DifferentiationWin rate vs. Nielsen, iSpot, VideoAmp, TV makersWe win on independence and action, not on priceDay 30
Data foundationOpted-in footprint and OEM mixFootprint stable or growing by OEM; no single-OEM dependencyDay 30
MethodologyBias correction, validation, error boundsPublished method, validated against panel and currencyDay 30
DeliverySelf-serve share, time to first read, automationMostly automated; little analyst work per readDay 30
ActivationLive destinations, match rates, segment usageSegments get bought, not just listed; usage tied to renewalDay 30

Scores come from finance data and interviews by Day 30. Revenue, growth and margin are not public.

30–60 day plan

Days 1–30

Learn the business and teams

People1:1s with every PM, data science, engineering, top-10 account owners, BD, privacy and finance
Clients10–15 conversations across holdcos, brands, networks and platforms: what decision do you make with our data?
Business resultsP&L by product and client, renewals, usage, win/loss, delivery SLAs
TeamsEmpowered or feature teams? Trace the last three cross-platform projects end to end

Days 31–60

Assess and set direction

Product healthScore each product line on the seven lenses with real data
Team topologyWhich team owns which customer problem? Rate each acquired system on strategic fit
Draft strategy12–18 month plan and investment ask
Early proof2–3 credibility wins that ship inside 90 days
Day 60 outputProduct healthPortfolio mapTeam topologyDraft strategy

To confirm by Day 60

Assessment priorities

PriorityWhat it meansHow we'd measure
01Know what clients pay forFind the problems clients pay the most to solve, and the revenue each product line brings inRevenue by product and client problem; client concentration
02Understand why we loseLearn the common reasons we lose to competitors or a pilot goes badly, and rank the top threeTop loss reasons ranked from win/loss reviews; baseline win rate and pilot-to-contract rate
03Find methodology weak spots and investment needsUnderstand where the methodology is weakest (identity, exposure definitions, numbers that don't tie out across the acquired tech) and where investment is neededGaps ranked by client impact; investment ask sized by Day 60
04Reduce the cost to serveUnderstand which service reports clients actually need, and which could be self-serve.Share of revenue from the platform; cost to serve per campaign
05Agree where to focus, where to betDraft focus areas and explicit non-goals. Treat bets as discovery experiments before scaling with clear success criteria.Investment split agreed with leadership; reviewed each quarter

2

Measurement vision

What does best-in-class measurement look like for advertisers today? Deep dive on cross-platform reach and frequency: methodology, data requirements, accuracy and scale tradeoffs, differentiation.

Measurement vision

Know who you reached, how often, on every screen

One deduplicated household spine behind reach and frequency, projected to the US population and joinable to outcomes data.

01 · See

True reach and frequency

Unique households and exposures across linear, CTV and digital, including which publishers add incremental HH reach.

02 · Understand

Where the frequency curve bends

The point where more exposures stop adding outcomes, by audience, publisher and creative.

03 · Optimize

Move budget mid-flight

Shift spend to lines with headroom, within ceilings and the inventory publishers actually have.

Current products from samba.com; the vision is my proposal.

Measurement products

FeatureClient problemHow Samba solves itValue to the client
Reach & frequencyAgencies & brands"Each platform caps frequency on its own, so the same homes get hit again and again."Deduplicated reach and frequency across linear, CTV and digital, read daily while the campaign runsLess spend wasted on over-exposed homes; budget moves to homes the buy missed
Reach & frequencyNetworks & streamers"We can't prove streaming adds reach beyond linear."Incremental-reach reports by network and platformEvidence to defend price in upfront and programmatic deals
Entertainment outcomesStudios, networks, streamers"We can't tell which media drove people to watch."Tune-in and box-office measurement tied to ad exposureSpend on what fills seats and drives viewers. X saw a +55% tune-in lift.
Digital outcomes & liftBrands, platforms"My lift numbers don't match my reach numbers."Site visits, sales and brand lift (with Cint) on the same exposure data as reachOne set of numbers that tie out, and a clear ROAS read
Sponsorship & placementBrands, rights holders"We pay for logos in content but can't see what we got."Samba AI detects logos and products on screenSponsorships valued on real on-screen exposure
Industry data inputsCurrencies, industry bodies"Our cross-media measurement is missing streaming."Samba streaming data feeds ANA's Aquila and Australia's OzTAMMore complete measurement for the whole market

Features from samba.com; client problems and value are my framing.

Deep dive

Cross-platform reach & frequency

Four views a buyer would use mid-campaign. All data is illustrative.

Cross-platform frequency: how to build it

Methodology, data requirements, accuracy vs. scale, and differentiation.

Data requirements (assumed)

  • Samba ACR exposure for linear and CTV, with timestamps and screen-on state
  • Ad impression logs with IDs and identity-join IDs (raw or via clean room)
  • Samba identity graph plus Semasio web context
  • Client first-party outcomes (e.g., site visit logs) with IPv4/IPv6, user agent and timestamps for joining, plus rich event metadata

Methodology

Layer 1

Capture exposure

  • ACR on the TV screen sees linear and CTV ads across TV brands
  • Publisher and DSP impression logs add what was served, and the screens ACR can’t see (mobile/desktop/etc.)
  • Walled gardens: aggregate reach curves or secure sketches

Layer 2

Resolve identity

  • One household graph across linear, CTV and streaming, refreshed weekly
  • Deterministic joins: TVs, CTV devices and phones on the same home Wi-Fi (IPv4 and IPv6), IDs (Ramp/UID2)
  • Probabilistic links only above a confidence bar

Layer 3 · Credibility

Clean, scale & calibrate

  • Drop ads played to an off or idle screen (a CTV left on overnight) and cap extreme counts
  • Weight Samba's footprint to US
  • Compare homes with similar viewing levels; reconcile totals to the deal's currency

Layer 4

Act and improve

  • One deduplicated exposure count per household-week: reach, frequency
  • Outcomes by frequency decile vs. matched unexposed homes
  • Push unreached and over-frequency homes to DSPs; callable by API/AdCP

Accuracy vs. scale

  • Latency vs. accuracy: Optimize on fast daily reads; report on calibrated weekly reads.
  • Deterministic vs. modeled: Wi-Fi joins are precise but only cover the panel; weight to the population and show the modeled share.
  • Noise vs. heavy viewers: Trimming outliers cuts noise but can undercount real heavy viewers.
  • Observed vs. causal: Heavily exposed homes differ; compare similar viewers and confirm with cap tests.

Differentiation

  • Best cross-platform OEM panel: know what homes watched.
  • ACR confirms the screen was on, so overnight and TV-off impressions are removed, not counted.
  • Outcomes by decile show where the next exposure stops paying.

My proposal, built from public information. To validate with Samba data science in the first 60 days.

1 · True cross-platform reach

Which screens and publishers add households nobody else reached.

Holiday launch · Linear + CTV + digital · Day 12 of 30Illustrative data
Linear TV
33.8M
+ CTV apps
+5.1M incremental
+ Digital video
+2.3M incremental
Unique households
41.2M
Unreached, in-market
9.1M
PublisherHouseholds reachedIncremental HH reachCross-publisher freq.Same-publisher freq.What it means
NBCLinear33.8M6.2M · 18%4.4x3.2xOver-served by one publisher
PeacockCTV14.6M2.3M · 16%5.0x1.5xSpread across publishers
HuluCTV11.0M0.7M · 6%6.3x1.4xMostly homes others already hit
Pluto TVCTV6.8M1.8M · 26%2.4x1.2xBest source of incremental HH reach
YouTubeDigital9.5M2.3M · 24%3.0x1.2xIncremental reach at healthy frequency

Incremental HH reach = homes this publisher reached that no other publisher on the plan reached.
Cross-publisher freq. = how many times, on average, this publisher's households saw the ad across every publisher on the plan.
Same-publisher freq. = how many times they saw it on this publisher alone. The gap between the two is frequency added by other publishers. Same total frequency can mean different problems: NBC drives most of its own frequency; Hulu adds to homes others already saturate. How it ties to the bars: the +5.1M from CTV is 4.8M seen on only one CTV app (2.3 + 0.7 + 1.8) plus 0.3M seen on two or more.

2 · Frequency dashboard

How exposure is spread, and when reach stops growing.

Unique households

41.2M

across every screen

Avg cross-platform freq.

4.0x

across every screen

3+ reach

24.3M

59% of reached homes

Homes at 7+ exposures

19%

past where outcomes flatten

Reach & frequency curve

41.2M households · avg frequency 4.0 · 59% reached 3+

Saw it exactly n times (bars add to 100%)Saw it n or more times

Under-exposed

24% saw the ad once. Most outcomes start at 3+, so these homes are worth another exposure.

On target

59% reached 3+. The share most buyers plan against.

Over-exposed

19% saw it 7+ times, past the point where the response curve flattens.

From measurement to decision: model where the next dollar still adds reach.

Reach curves by line

Unique households vs. impressions since flight start. Dots are observed; dashed lines are fitted curves, extended 50% past today. The shaded band on each line is its most efficient range: impressions still add incremental HH reach quickly, up to 70% of the line's ceiling. Hover a line to isolate it.

3 · Reallocation recommendation

Move budget from saturated lines to lines with room to grow. Try the controls to see how each limit changes the plan.

Same budget, $2.0M

+0% tune-ins with Samba

KPIWithout SambaWith Samba
Tune-ins
Cost per tune-in
Avg frequency4.0x

Where the budget moves

LineShare of total budgetRecommended spend

Today's shareRecommended shareMost this line can take (ceiling or inventory)

Ideal: apply it programmatically, not by hand

ConnectSamba links to the DSP or SSP by API (The Trade Desk, DV360, Amazon, Magnite).
Adjust bidsBid up on under-reached homes, down on over-frequency ones; caps and suppression update automatically.
Show the gainUsers see KPIs with vs. without Samba. They approve the guardrails, not every bid.

Future bet · Autopilot reallocation

An agent rebalances daily within your guardrails through DSP APIs or AdCP, keeps a holdout to prove the lift, and explains every move.

Assumptions for this projection

  • Past behaves like future. It assumes ads measured so far keep performing the same way: same creative, season and audience.
  • Inventory has to exist. Each publisher, ad length and format can only absorb what it has available, so every line has an inventory cap.
  • Ceilings prevent "put it all in one place." Without them the model moves everything to the best line, which then saturates and frequency climbs there instead.
  • Floors protect commitments. Upfront deals and partner minimums mean some spend can't move, even when the model would move it.
  • Adjusted for who sees more ads. Heavy viewers respond differently, so frequency effects are compared within similar viewing levels, not read from raw buckets.
  • More budget earns less per dollar. Extra spend on a line saturates it, so outcomes grow slower than budget.
  • Only measured cells count. Gains are weighted over the daypart and publisher cells Samba can measure.

4 · Audience overlap by property

The overlap report clients already use. Each cell is the share of the row property's audience that the column property also reached.

Darker = more overlap: budget paying twice for the same homes.

Overlap under today's plan. Switch to see it after a reallocation recommendation, which shifts budget toward properties that add incremental HH reach.

Avg overlap between properties

24%

Unique households

41.2M

Budget

$2.0M

Views 1–4 use illustrative data to show the product, not real campaign results.

Competitive comparison

Everyone else helps buyers decide where to stop spending. Samba can show what the next exposure is worth.

CapabilityCurrenciesNielsen, Comscore, iSpot, VideoAmpAd serversInnovidWalled gardensGoogle, AmazonDSPsTrade Desk, DV360Sambawith this plan
Frequency across every publisher●●◐◐●
Incremental HH reach by publisher◐●◐○●
Same-publisher vs. cross-publisher○●○●●
Frequency tied to an outcome○○○◐●
Independent of buyers and sellers●●○○●
Signal back to DSPs and agents○●◐●●
Certified for buying and selling●○◐○○

● offered · ◐ partial, or only inside one ecosystem · ○ not offered. Based on public product documentation, 2026. DSPs only see the media bought through them. Walled gardens are only accurate inside their own walls. The big clean rooms are now owned by agency holdcos (InfoSum at WPP; LiveRamp agreed to Publicis in May 2026).

The gap

No one puts an outcome on the frequency curve. Currencies have no outcomes; DSPs only see their own media.

Samba's edge

ACR exposure across TV brands, plus outcomes (tune-in, site visits, lift) on the same household graph.

Honest gap

Not a currency. Pursue accreditation for narrow metrics, and reconcile to whatever currency the deal uses.

3

Audiences vision

What are the must-haves for Samba in audiences? What are clients looking for from us? How do audiences tie into activation, and how do we make that real?

Audiences vision

Turn TV and web behavior into targeted reach

FeatureClient problemHow Samba solves itValue to the client
Behavioral & contextualAgencies & brands"I need to target without relying on IDs."Viewer, reader and in-market audiences; contextual from Semasio works without IDsScale that holds up under privacy rules, here and abroad
Competitive conquestingBrands"A rival is outspending us on TV."Homes exposed to a competitor's ads, ready to targetReach the rival's audience on any screen
Incremental reachAgencies"My TV buy keeps missing light-TV homes."Homes the TV buy missed, pushed to CTV and digitalMore unique reach for the same budget, with less over-frequency
Tentpole retargetingBrands"Our big live-event spot is one and done."Live-event viewers, reached on other screens in real timeThe moment keeps working after the broadcast
Project Gravity NewBrands & agencies"Onboarding our first-party data takes weeks."Onboard client data, model lookalikes, push to 40+ platformsUsable segments from client data, without a long setup
Curated dealsPublishers & SSPs"We need audience packages buyers trust."Samba audiences packaged into curated deals through SSPs and publisher directHigher-value inventory with independent backing

Features from samba.com; client problems and value are my framing.

What clients want from us

Agencies & brands

Scale with accuracyReach that holds up when measured
TransparencyHow each segment is built
ProofPerformance measured on the same data
SimplicityOne place to plan, activate and measure

Publishers & platforms

Independent validationThird-party proof to sell premium inventory
Curated packagesReady-made audience deals
PartnershipData that strengthens their own offer, never replaces it

Everyone

Privacy confidenceData use they can explain to legal

Must-haves

What Samba needs in audiences, and where it stands today.

Must-haveWhat it meansSamba today
What we sell
Viewership audiencesShow, genre, light-TV and competitor-exposed homes. Only ACR can see these.● Strong (ACR)
ContextualTargeting that works without IDs, here and abroad● Strong (Semasio)
Outcome audiencesLikely-to-convert models, client lookalikes, and lookalikes of homes that already bought◐ New (Project Gravity)
What makes it usable
Linked to measurementUnexposed and over-frequency homes straight from the R/F product◐ Possible, not yet packaged
Everywhere the buy happensDSPs, SSP curation, publisher direct, clean rooms● 40+ platforms
Transparent by defaultSegment makeup, freshness and match rates○ Not visible publicly

● strong · ◐ partial or new · ○ gap or not visible. Samba today is my read from samba.com and AdExchanger (Semasio); treat it as a hypothesis.

Deep dive

Audiences: how we'd build them

Same foundation as frequency: one household graph and one data set serve both.

Data requirements (assumed)

  • ACR viewing data per TV
  • IPv4 and IPv6 history plus device IDs for the graph
  • Semasio contextual signals for ID-free targeting
  • Client first-party data and purchase-data partners (receipt or card panels) through clean rooms, as seeds and outcomes
  • Destination IDs (UID2, RampID, CTV device IDs) and match-rate feedback

Methodology

01

Viewing data

ACR on opted-in TVs records the shows, networks and ads watched.

02

Shared household graph

The same graph as measurement: the TV linked to phones, laptops and CTV devices on the home Wi-Fi (IPv4 and IPv6), refreshed weekly.

03

Build the segments

Viewership segments, Semasio contextual, client lookalikes (homes that watch like recent buyers).

04

Quality and privacy checks

Minimum segment sizes, freshness windows, sensitive-category exclusions, and IVT removed.

05

Activate and close the loop

Push to DSPs, SSPs, MCPs, curated deals and clean rooms.

Accuracy vs. scale

  • Precision vs. scale: Strict viewing rules (e.g., 3+ episodes) are precise but small; looser rules scale. Offer both.
  • Observed vs. modeled: Lookalikes reach past the TV footprint. Label the observed vs. modeled share.
  • Match rate vs. reach: Match rates drop at every hop. Report reach after matching, not before.
  • Granularity vs. privacy: Minimum segment sizes; small segments get modeled estimates.

Differentiation

  • Measurement-derived segments (unexposed and over-frequency homes): our unique value.
  • Most audience data is modeled from purchases or surveys. Samba's starts from what was actually watched.
  • Same data and graph as measurement, so the audience you buy is what you measure.
  • Semasio contextual keeps working where IDs don't.

Future bet · Responsive audiences

Combine ACR viewing, Semasio interests and measured lift to find homes the ad actually changes, not just homes that would have tuned in or bought anyway. Each segment ships with its own frequency cap and a holdout to prove the lift. Guardrails: no sensitive categories, and buyers can see why a home was added.

My proposal, built from public information. To validate with Samba data science in the first 60 days.

4

Portfolio strategy
& roadmap

How would you prioritize investment across these products? What is the 12 to 18 month roadmap and its key bets? How do you think about P&L and business outcomes?

12–18 months · one budget, three lanes

Roadmap

Months 0–6

Months 6–12

Months 12–18

Core

50%

Protect

  • Automate the top 3 reports
  • Reduce time to insight
  • Improve core methodologies: identity, calibration and outlier rules
  • Standard reads fully self-serve

Grow

35%

Scale

  • Turn measurement into audiences: unexposed and over-frequency homes
  • Act on it programmatically: bid changes and converter lookalikes
  • Bundle measure + activate by vertical

Bet

15%

Agentic advertising

  • AI chat over R/F and audiences, internal first
  • Connected to bidding
  • Further automate intake
  • Open the chat to clients; fully automated reports
  • Automated sales outreach: self-serve leads based on new ad changes
  • Customer request → full feature deployment

Starting split, revisited quarterly.

P&L and business model

Start with a scorecard for every client, then price the platform in tiers.

Client scorecard · who is profitable, who is at risk, who should move up a tier

Usage

  • Reports viewed and logins
  • API calls and feature use (Mixpanel-style events)
  • Active seats per client

Revenue

  • ARR and expansion
  • Tier and add-on reports bought
  • Measure + activate attach

Cost to serve

  • Analyst hours per client
  • Data and compute cost
  • Share of delivery that is automated

Health

  • Renewal date and risk
  • Usage trend vs. last quarter
  • Ready to move up a tier?

Business model · one platform, tiered access

The base is data access to the core platform. Dimensions, metrics and reports unlock by tier. Anyone can still buy a report, at the window price.

Pay at the windowCore SCore MCore L
Annual TV + CTV ad spendAnyUnder $25M$25–100M$100M+
Platform access: R/F dashboard and API—●●●
Dimensions: demo, geo, daypart—Totals only●●
Metrics: outcomes and liftPer report—●●
Incrementality and audiencesPer report——●
Reports included per year041224
Extra reportsFull rateMember rateMember rateMember rate

Illustrative tiers with no prices. Core M is highlighted as the likely default for mid-size brands. Tier set by annual TV + CTV ad spend.

5

Team & stakeholders

How would you structure the team and its ownership? How would you partner with Data Science, Engineering, Sales and BD? How do you make the strategic initiatives function effective?

Teams are grouped by client problem.

Cross-platform R/F

Mission

Help every buyer see who they reached, how often, and where the next dollar does the most good.

R/F product, currency reconciliation, reach curves for planning

Owns: % of campaigns on in-flight reads

Outcomes & incrementality

Mission

Prove which TV ads changed what people did, in terms a CFO trusts.

Attribution, lift, incrementality, outcome-data partners

Owns: % of campaigns with a lift or ROAS readout

Audiences & activation

Mission

Turn what people watch into audiences buyers can use anywhere, and show they worked.

Segment catalog, audience builder, destination performance

Owns: measure-to-activate attach

Integrations & APIs

Mission

Put Samba data wherever the buy happens, for people and agents, with no manual steps.

DSP, SSP, clean-room and TV-maker connectors; agent interfaces

Owns: destinations live; agent call volume

Operating cadence · based on Cagan's Empowered

Vision & principles

Set once, revisited yearly

Annual

1–2 year plan

Quarterly

OKRs mapped to strategy; plan 3–6 months out

Monthly

Customer-facing roadmap update

Weekly

Agentic program updates, near-zero program management

Working with partner teams

PartnerHow we work togetherRitual
Data ScienceMethodology is a product: DS leads sit in each pod with shared OKRs and co-own error budgets and validation gatesMethodology council with an external advisor
EngineeringPlatform first: shared exposure, identity and activation services, with product pods on top; reliability owned jointlyJoint quarterly planning
SalesProduct builds enablement (narratives, case studies, battlecards) and runs a design-partner program with top clientsMonthly commercial council: win/loss, pipeline, pricing
Business DevelopmentPartner tiers follow the strategy: data supply, destinations, outcome data. Product sets requirements; BD closes termsJoint business cases, one P&L owner
Legal, Finance, MarketingPrivacy review in every PRD; finance co-owns the P&L model; marketing turns validation into the market storyPrivacy gate in the PRD template

Strategic initiatives

A stage-gated process for bets.

1 · Explore

Frame the bet

And its success metric

2 · Validate

Prove it

With design partners

3 · Scale

Fund it

Against the Core / Grow / Bet mix

4 · Graduate

Hand it to a pod

With an owner and a budget

EVERY GATE HAS A SUCCESS METRIC AND A KILL CRITERION

Charter

Owns cross-portfolio bets no single pod can deliver.

Cadence

Monthly portfolio review of the scorecard and bets; quarterly reallocation across Core, Grow and Bet.

Integration playbook

Acquired capabilities such as Semasio and Bestever AI join the loop within two quarters.

Thank you

Questions &
discussion

Currencies tell you what happened. Samba helps you change it.

Appendix

Measurement data requirements +
DataPurposeSourceRisk
Opted-in ACR viewingLinear and CTV exposure spineOwned footprint plus TV-maker partnersConsent rules; TV-maker walled gardens
Ad fingerprint libraryDetect ads on the glassOwned, plus agency and client creative feedsCreative versions at scale
Impression logsCoverage beyond the TVTags, publisher and SSP log deals, clean roomsPartner willingness; latency
Identity graphHousehold and cross-device linksOwned graph plus UID2, RampID, partner IDsIP address loss
Person-level panelCo-viewing, demos, truth setLicensed panel or survey partnerCost; small-demo sample size
Universe estimatesProjectionCensus and ACS, third-party demosLow
Currency feedsReconciliationClient-provided or licensedCommercial terms
Public data assessment: all 18 boxes +
Market detail +

US CTV ad spend, 2026

$37.95B

~+14% (eMarketer)

IAB's narrower CTV figure

$29.3B

+11% (IAB Outlook 2026)

CTV upfronts

$17.73B

vs. $16.98B primetime linear

Retail media on CTV

$9.03B

by 2029, from $4.46B in 2025

eMarketer and IAB define CTV differently, so their totals differ by ~$8B. Sources: eMarketer (via Adwave) · eMarketer upfronts · IAB Outlook

Portfolio inventory +
ProductWho buysJob it doesPublic milestones
Reach & frequencyBrands, agencies, networksDeduplicated reach across linear, CTV and digitalR&F module in Analytics Dashboard, Dec 2020
Outcomes & liftBrands, platformsProve tune-in, site visits, sales and brand liftBrand lift with Cint, Jul 2026; X study, Sep 2026
AudiencesAgencies, DSP buyersBehavioral, contextual and conquesting segmentsSemasio, Oct 2024; Acxiom in 40+ markets, Jun 2025
Project GravityBrands, agenciesOnboard client data, model lookalikes, push to 40+ platformsLaunched May 2026
Industry dataCurrencies, holdcosStreaming data as an input to others' measurementAquila, Oct 2025; OzTAM; Omnicom India, Dec 2025
Samba AIEvery product aboveRecognize shows, ads, logos and products on screenDisruptel, Nov 2022; Bestever AI, 2026

Revenue, growth and margin are not public. They're the first ask of finance, and nothing here estimates them.

Interdependencies +

Single point of failure

If OEM terms or consent rules shrink the ACR footprint, every product loses scale at once.

Single point of failure

If identity match rates drop, reach dedup and activation degrade together.

Bundle

Measurement findings (missed and overexposed homes) can ship as audiences: one loop, one contract.

Portfolio map +

Fix

Targeted turnaround plan. Low health, high importance.

Invest

Double down and accelerate. High health, high importance.

Sunset

Sunset or reposition. Low health, low importance.

Harvest

Maintain, minimize spend. High health, low importance.

Placement comes from the scorecard plus finance data, so nothing is plotted yet. Each move sets where people and dollars shift, and what stops.

Technology review +
AuditWhat we checkUnified looks like
IdentityIs there one graph? Does the same household ID carry from ACR to web to activation?One Samba ID behind every product
ScaleHow does each piece scale: footprint, markets, data latency, cost per query?Known coverage and latency, per market
ExposureIs an exposure defined the same way across ACR, impression logs and page views?One exposure spec: linear, CTV, digital
ReachIs reach deduplicated on the same spine, with the same projection and weights?One reach number, any cut
OutcomesAre lift and outcomes measured with the same controls and method?One published incrementality method
Sources +