
Operations
Part of The honest guide to marketing attribution
Marketing attribution consultancies compared: independence, methods, price
Compare named marketing attribution consultancies on independence, method, and price before you sign, and test every claim on your own data.
What to take away
- Independence decides who a marketing attribution consultancy can afford to disappoint. Media agencies and resellers rarely recommend that you spend less.
- Method fit beats brand. Last-touch, multi-touch, incrementality testing and marketing mix modeling answer different questions.
- Ask for the pricing model, not a numberretainer, project fee, hourly, or software markup.
- Run a paid pilot on one bounded decision before any annual contract.
- Get exit terms in writingdata export, model definitions, and deletion.
The firms buyers actually shortlist
Seven names come up repeatedly in shortlists. They differ most on who owns them, what they measure, and how they bill.
| Firm | Independence | Method | Price model |
|---|---|---|---|
| Analytic Partners | Independent; owns no media | Commercial mix modeling, optimization | Quoted per engagement |
| Ekimetrics | Independent; Paris, New York, London | Mix modeling plus attribution | Project fees and retainers |
| Ipsos MMA | Owned by Ipsos, a research group | Unified measurement, mix modeling | Annual program |
| Merkle | Owned by dentsu, which owns media agencies | Identity resolution, attribution | Retainer plus licensing |
| Kepler | Inside Kyu, a Hakuhodo company | Attribution and incrementality | Agency retainer plus media fees |
| Marketing Evolution | Vendor with a services arm | Person-level attribution | Subscription plus services |
| Avinash Kaushik | Independent analyst; ex-Google | Method review | Day rate or fixed scope |
Analytic Partners owns no media and sells no ad inventory. It builds commercial mix models for large advertisers, and publishes client case studies. It suits brands whose spending justifies aggregate modeling. Engagements are quoted; multi-market programs typically run into six figures a year.
Ekimetrics is a Paris-founded data science consultancy with offices in New York and London. It pairs mix modeling with attribution work. It bills project fees and monthly retainers, not a software license.
Ipsos MMA sits inside Ipsos, a market research group that buys no media for clients. Its programs combine mix modeling with survey and panel data. Ipsos quotes the work as an annual program.
Merkle, a dentsu company, handles identity resolution, attribution, and analytics. dentsu also owns the media agencies Carat and iProspect, so ask how the measurement team is walled off from media planning. Pricing is a retainer plus platform licensing.
Kepler is a digital agency that markets itself as independent of the big media groups, though it sits inside Kyu, a Hakuhodo company. Its measurement practice covers attribution and incrementality tests.
The same firm can plan the media it later measures, so push hardest on independence here. It suits brands that want media and measurement under one roof, and bills an agency retainer plus media fees.
Marketing Evolution sells person-level measurement and attribution software with a services team attached. It suits marketers who want a platform plus a team from one vendor. Pricing is a subscription plus services, quoted per scope.
Avinash Kaushik, author of Web Analytics 2.0, spent years as Google's digital marketing evangelist and now advises on measurement method. He sells no media and no platform, which suits buyers who want method review without a platform sale.
Independent measurement advisors in US markets typically quote $1,000 to $3,000 a day. Fees are usually day rates or fixed-scope projects. Ask what his Google history means for his recommendations.
An in-house analytics team is the option buyers skip most. A senior analyst plus a clean data pipeline beats most retainers for a single-channel business. US senior analysts typically earn $100,000 to $140,000 a year.
If your shortfall is software rather than method, compare marketing attribution tools before paying anyone to advise on them.
Discovery pool cautions
- Compare tools before paying advisors
- AMA listing is discovery, not evidence
- 4As tool finds contacts, not rankings
- Membership is not endorsement
- Always run buyer due diligence
Directories such as the AMA's Marketing Services Search and the 4As agency finder are discovery records, not endorsements. Membership is not due diligence.
Independence is the first question
Ask who owns the media, the technology, and the implementation. A firm that earns media commission has a reason to keep the attribution model generous to paid channels. Merkle and Kepler sit inside groups that sell media, while Analytic Partners, Ekimetrics and Ipsos MMA do not.
Then ask what the firm has recommended against its own interest. A consultancy that has told a client to cut spend is worth more than one that never has.
Method fit decides the rest
Attribution models allocate credit. Incrementality tests measure cause. Marketing mix modeling evaluates aggregate investment. These are not substitutes, and a firm that sells only one will frame every problem as that one.
Analytic Partners and Ipsos MMA lead with mix modeling. Marketing Evolution leads with person-level attribution. Kepler leans on incrementality tests. Match the method to the decision before you compare fees.
For the model choices and their assumptions, see marketing attribution.
Price the engagement in parts
Never accept one blended number. Separate discovery, build, ongoing support, data costs, and exit. Ask whether the fee is a retainer, a project fee, an hourly rate, or a markup on media or software.
As of 2026, in most US markets, independent measurement advisors typically quote $1,000 to $3,000 a day. Agency measurement retainers commonly start near $10,000 a month. Enterprise mix modeling programs typically land between $150,000 and $400,000 a year.
Where a figure matters and you do not have it, write the arithmetic in your own variables: annual fee, expected hours, and the value of the decision the work informs. If the fee exceeds the value of the decision, the engagement is priced wrong already.
Run a paid pilot first
Bound the pilot to one decision and a known period. Supply edge cases, not clean sample data.
Pilot test / Acceptance evidence
- Reconciliation
- Eligible source total and explained difference
- Method
- Versioned rules, model, window, identity, and code
- Quality
- Late, duplicate, missing, and corrected records handled
- Transfer
- Your staff can rerun and maintain the work
- Exit
- Data, definitions, logic, logs, and deletion confirmed
Read the result against published marketing attribution benchmarks, not against the pitch.
Check the published evidence
Separate provider claims, customer statements, and measured records. Reproduce one material claim with your own data before you sign. If the method is unavailable, treat the claim as a lead for testing.
The FTC advertising substantiation policy requires a reasonable basis for objective claims before they are published. The FTC guidance for marketers using reviews covers fake feedback, selective solicitation, and hidden relationships. Both are U.S. policy; for your own jurisdiction, ask qualified counsel.
The failure patterns these tests catch are set out in marketing attribution mistakes.
Common questions
Which attribution consultancy is best?
None is best in the abstract. The right firm passes your independence, method, data, and cost checks for one named decision, and proves it on a paid pilot.
Should a shortlist rely on awards?
No. Awards help discovery. They do not replace comparable references, work samples, a conflicts review, and a controlled pilot.
How should a firm be asked to prove independence?
Ask for its revenue sources, its ownership of any media or software it recommends, and one decision it argued against its own commercial interest. For Merkle and Kepler, ask who signs off when media spending and measurement findings disagree.
What belongs in the exit terms?
Data export in a usable format, model definitions and version history, run logs, and confirmed deletion. Agree these before the first invoice, not at renewal.







