Flywheel Digital Reviews: Reading the Signal Correctly

You are almost certainly reading the wrong kind of review

Type this query and the results fill with employment sites. Those are people who worked there writing about working there. They are not brands writing about outcomes, and the two things correlate far more weakly than the search results imply.

This is structural rather than particular to Flywheel. Enterprise commerce relationships are covered by mutual confidentiality, priced individually, and reviewed publicly almost never by the people who buy them. There is no G2 for holding-company agencies. So the client shelf stays empty, the employment shelf fills up, and searchers end up forming a view about service quality from data about compensation and workload.

Employee sentiment is not worthless. Sustained churn in an account team shows up on your side as re-briefing, lost context and a slower hand on the account, and that is a legitimate thing to ask about in a pitch. But it is an indirect signal about a different question, and it is a poor substitute for twenty minutes with two brands of your size who are in the account right now.

There is a second thing worth noticing about this search result page, and it applies to us. Almost every "best commerce agency" ranking that appears alongside these reviews is published by an agency that appears in it. Check the domain on any list before you read the order. Then apply the same test here: this page is published by reMKTR, which competes with Flywheel for Amazon DSP work. Use it to know what to ask, not to decide.

Who Flywheel is now, on their own published figures

The ownership history matters because it changed what the company is, and a lot of the commentary in circulation predates the change.

Founded in 2014 and acquired by Ascential in 2018, Flywheel was bought by Omnicom for a net cash purchase price of $835 million — announced in October 2023 and completed on 2 January 2024. It now operates as a practice area inside Omnicom rather than as an independent specialist, led by the former Ascential chief executive.

Scale, taken from Flywheel's own site rather than from a roundup: 4.5k clients currently served worldwide, 50+ of the top 100 publicly listed global CPG brands, partnerships spanning 400+ digital marketplaces and 100+ retailers, and operations across 30+ global markets. The service map runs Retail Media, Retail Operations, Market Intelligence, Managed Media Services, Content and Creative, Consulting and Omnichannel Campaigns, all wrapped around the Commerce Cloud platform. No pricing is published anywhere on the site; the route is a contact form.

Read that profile straight. It is a very large, global, CPG-weighted commerce practice with its own software, sitting inside one of the largest advertising groups in the world. That combination is genuinely rare, and it is a specific fit rather than a universal one.

One neutral structural observation, drawn with no conclusion attached: a holding company that owns both retail-media software and managed retail-media services will sometimes have its software serving agencies that compete with its own services arm. That is simply how the group is arranged. It is a reasonable thing to ask about, and an unreasonable thing to infer anything from without asking.

What employment reviews can and cannot tell you

If you are going to read them anyway — and you are — read them properly.

• Do treat consistent notes about account-team churn as a question for the pitch, not as a verdict. Ask who would be on your account and how long they have held their current roster.
• Do check whether reviews cluster before or after January 2024. A review written in 2019 describes an independent company; one written in 2022 describes an Ascential subsidiary; neither describes the practice you would be hiring.
• Do not infer campaign performance from employee satisfaction. Some excellent agencies are demanding places to work, and some pleasant ones are quietly coasting on an account.
• Do not average a small sample. Where a profile carries a handful of entries, note the count and decline the conclusion — that is the only defensible reading, and it applies to flattering samples as much as unflattering ones.

We deliberately have not repeated specific ratings or counts here. Review platforms return different figures to different readers and to search-result snippets, we have seen snippet and live page disagree repeatedly, and an unverified number that damages a company is the most expensive kind of mistake a page like this can make. Go and read the pages yourself, sorted by date.

Ownership change as diligence, not as a warning

Flywheel is one of many firms in this category that a buyer would now be hiring under different ownership from the one their reputation was built under. The useful response is a clause, not a suspicion.

The documented moves worth knowing, purely as context: ChannelAdvisor became Rithum after CommerceHub took it private; Kenshoo became Skai in 2021; Perpetua has been an Omnicom asset via Flywheel since January 2024, though its own site does not say so; Carbon6 was acquired by SPS Commerce; Sunken Stone relaunched as Emplicit; Teikametrics renamed its platform ARI at the end of 2025. Several of those firms are stronger for the change. None of this is a criticism of any of them.

What it justifies is asking every vendor — including reMKTR, including Full Circle — for three things in writing:

• Change-of-control language. What happens to your rate, your team and your term if the firm is acquired mid-contract.
• Price protection. Whether the fee is fixed for the term or revisable, and how much notice a revision requires. A written notice period for fee changes is a small ask and a large protection.
• Data portability. Who owns the advertiser seat, the ad account permissions, the audience definitions, the Amazon Marketing Cloud instance, and the historical reporting — and what you walk away with on the day the relationship ends.

That third point is the one agencies find easiest to answer well and buyers forget to ask. Agree offboarding at the start, when everybody is friendly and it takes five minutes.

The reference call that would actually tell you something

Since the public client record is thin, the reference call is not a formality — it is the primary evidence. Ask for two brands of similar size in a similar category, and take twenty minutes with each. Eight questions do nearly all of the work.

• Who is on your account today, and how many of them were there twelve months ago?
• What was the last significant change made to your display campaigns, and what evidence prompted it?
• Has display ever been tested against a holdout, and what did the test show?
• How are Amazon DSP and sponsored ads reconciled so that they are not both claiming the same orders?
• What is inside the fee, and what has been invoiced on top of it in the last year?
• What happened the last time performance dropped for a full month?
• How long is your term, and what notice do you have to give?
• If you left tomorrow, what would you keep?

Question three separates agencies more sharply than anything on a credentials deck. Any partner can produce an attributed number. Far fewer can tell you what would have happened without the spend, and fewer still have written that obligation into a statement of work.

The measurement argument, including the part that cuts against us

Our whole position rests on one sentence: last-click attribution cannot prove incrementality and never could. Holdouts and matched controls can. A platform-attributed report tells you which orders had an ad touch in the path, not which orders the ads caused, and on display the difference between those two numbers is frequently the entire business case.

It would be dishonest to sell that as a position only we hold. Tinuiti publishes a measurement practice called Bliss Point with a productised Incrementality Lab inside it — real-time experimentation to isolate causal lift, alongside media mix modelling. That is substantially our argument, made by a much larger agency, and any competent buyer will raise it. They should.

Where the honest competition sits is narrower, and it applies to us with full force:

• Who designs and computes the test? A lift study run by the party being measured needs a process that manages the conflict rather than ignoring it. Ask who builds the holdout, who holds the query, and who sees the result first.
• Is it in the SOW? Measurement in a pitch deck and measurement as a contracted deliverable with a cadence are different products.
• Does it cover both channels? Sponsored ads and DSP double-count each other constantly. Reconciling them in Amazon Marketing Cloud is the only way to stop adding two overlapping numbers together.

Where a specialist sits differently

reMKTR does one thing. We run Amazon DSP as a managed service on our own seats, inside Full Circle — a full-service Amazon management company with $500M+ in managed spend across 100+ brands — and we hold 109 live Amazon DSP advertiser seats.

Because the client-review shelf is empty for everyone in this part of the market, the substitute we offer is a scoped number you can interrogate. From a direct Amazon DSP API pull covering 30 of those advertisers during July 2026:

• 6.04x return on ad spend across the whole set
• 78.4 million impressions at a $4.00 CPM
• a blended $1.42 cost per click
• $5.49 cost per acquisition on 57,137 attributed purchases
• 20.1% of those purchases from shoppers new to the brand

The scope is stated deliberately: thirty advertisers, one month. A return figure without a denominator is a marketing asset, and this industry has a great deal of that already. It describes how the book runs; it is not a forecast for your account and we will not present it as one.

For completeness on the alternatives: Amazon's own DSP product page states that its managed-service option "typically requires a minimum investment of USD 50,000", with self-service available if you employ a trader. An agency seat is the middle path, usually with more flexible minimums and someone accountable weekly.

Two redirects worth making. If what appealed about Commerce Cloud was the analytics layer rather than the media service, Orbit is our software and it comes with the engagement rather than as a separate subscription. And if the real margin leak is stockouts, fees and reimbursements rather than media efficiency, no advertising partner fixes that — Dr. Stock addresses it directly.

Who Flywheel genuinely suits

If you are a global CPG brand selling across dozens of marketplaces in many countries, and you need retail media, retail operations, market intelligence and content coordinated under one roof, Flywheel is built for exactly that. Very few organisations on earth can do that job, and a specialist DSP shop is emphatically not one of them.

They also suit brands already inside the Omnicom ecosystem, where commerce work can be joined to existing media planning rather than bolted alongside it, and brands whose problem is coordination across markets rather than depth in one.

They suit you less if you are a mid-market, Amazon-led brand whose central question is whether display is producing incremental orders. At that size you are a small account inside a very large practice, and the thing you most need — a senior operator with unhurried time for your account — is the thing that scale makes hardest to guarantee. That is arithmetic rather than criticism, and it applies to every large agency including the ones we admire.