Skai vs Pacvue: The Decision Usually Sits in the Contract

Each product was built to answer a different question

Feature grids make these two look like near-substitutes. They are not, and the fastest way to tell them apart is to ask which recurring argument inside your business the software is meant to settle.

Skai was built for the team that has to defend a whole media budget. Retail media, paid search and paid social sit in one login so that a planner can ask which channel deserves the next increment. Its centre of gravity is planning, comparison and measurement across channels that report in different currencies of truth.

Pacvue was built for the team that has to operate many retailers at once. Its homepage today names Pacvue Prism, Pacvue Agent, Pacvue MCP and Pacvue DaaS alongside cross-retailer media planning, digital shelf performance and revenue recovery, and describes work spanning 100+ retailers and partners across 30+ countries. Its centre of gravity is execution: bids, budgets, share of shelf, bulk rules, and the commerce operations that sit next to media.

So the honest test is a sentence, not a spreadsheet. If the meeting you keep having is "how did retail media do against social this quarter", that is Skai's question. If it is "why did our Walmart bids drift again on Tuesday", that is Pacvue's. Almost every other difference is downstream of those two sentences.

One point of vocabulary before going further, because the search results mix them up: Amazon DSP here means Amazon's demand-side platform for programmatic display and video. It has nothing to do with an Amazon Delivery Service Partner, which is a logistics franchise. Both products in this comparison connect to the advertising one.

One of them publishes five tiers. The other publishes nothing

This is the cleanest factual split between the two, and it is worth stating precisely rather than rhetorically.

Skai publishes a ladder on its own pricing page. Read on 20 August 2026, in US dollars, on annual billing: Standard at $114k a year up to $4M of annual media spend, Advanced at $276k up to $10M, Enterprise at $504k up to $20M, Enterprise Premier at $756k up to $35M, and a fifth, custom-quoted band above $35M. The page also says the platform comes "with predictable annual pricing and the flexibility to review your commitment after the first three months", and lists incrementality testing as an Enterprise Premier inclusion rather than a base one.

Pacvue publishes no price. The URL people keep linking to, pacvue.com/pricing, returns a 404 rather than a page, there is no pricing link in the navigation, and every route on the site ends at a demo request. That is a normal enterprise motion and we are not going to dress it up as evasion. It is simply a cost you should count: with Skai you can rule yourself in or out in five minutes, and with Pacvue the cheapest route to a number is a call. On a shortlist of six vendors that difference is measured in weeks.

Third-party sites will happily fill the gap with estimates. We are not reprinting them. None come from Pacvue, several disagree with each other, and the ones that agree usually agree because they copied each other.

A rule that applies to every vendor including us: check which billing tab you are reading. Annual and monthly views of the same page produce different-looking numbers, and reading the wrong one is how most published comparisons of this category acquire their errors.

The arithmetic a spend-banded ladder hides

Skai's tiers are capped by annual media spend, which means the fee you pay has an effective percentage attached to it — and that percentage depends entirely on where you sit inside your band.

Run it at the ceiling of each band and the ladder looks gentle. $114k against $4M is roughly 2.9% of media. $756k against $35M is roughly 2.2%. Fees tapering as spend grows is exactly what you would want.

Run it at the floor of a band, which is where a lot of buyers actually live, and it looks different. A brand putting $1.5M a year through the Standard tier is paying about 7.6% of media for software. A brand that has just crossed into Advanced at $5M is paying about 5.5%. The band you land in matters more than the headline rate, and the most expensive place to be is one dollar over a threshold.

That produces the single most useful question to carry into either sales conversation. Not what does it cost, but what does my fee do at the spend I will realistically be at in eighteen months, and what happens the month I cross a line? Ask for the answer as a table, not a reassurance.

The same discipline applies to any variable rate a vendor attaches to your spend, ours included. At $100,000 a month in advertising, one undisclosed percentage point is $12,000 a year, and it grows precisely when you are trying to scale. Nobody puts that line in the comparison spreadsheet, because it does not have a headline to sit under.

Read the contract, not the feature grid

Here is the part that every ranking comparison of these two products leaves out, and it is the part that costs real money.

The decision-relevant document is not the feature matrix. It is the terms of service, and specifically four clauses: term length, auto-renewal, notice period and data portability. They rarely appear in a demo and they determine what happens in month thirteen, which is when most of these relationships are actually re-decided.

A worked example from this category, on a vendor's own published terms. Quartile's terms and conditions, read 20 August 2026, state that "the initial term of this Agreement shall be one year from the Effective Date", that it "shall automatically renew on each anniversary of the Effective Date for successive one-year periods", that either party may decline renewal "by providing written notice of non-renewal at least sixty (60) days prior to the expiration", that fees "shall be due in advance of Services rendered", and that the client "shall not be entitled to any refund in connection with any termination". A month-to-month arrangement exists only as a written exception the vendor may grant.

None of that is unusual and none of it is a criticism. It is a completely ordinary enterprise contract. The point is that it is published, decision-relevant, and cited by nobody — every listicle ranking for these keywords compares dashboards while the auto-renewal window quietly does the deciding. Miss a sixty-day notice date and you have bought another full year of a platform you had already decided to leave.

So before you compare Skai and Pacvue on modules, ask both for the master agreement and read these lines:

• How long is the initial term, and does it renew automatically?
• How many days' written notice, and dated from what — the anniversary, or the invoice?
• Are fees payable in advance, and is any part refundable if you leave mid-term?
• Is the rate fixed for the term, or can it be revised at renewal, and with how much warning?
• What is a change-of-control clause going to do to your price if the vendor is acquired?
• On exit, what do you keep — historical reporting, audience definitions, bid history, the raw data behind the dashboards?

That last one is the sleeper. A platform that has held two years of your optimisation history holds two years of your institutional memory, and "we can export a CSV" is not the same answer as "you keep the model".

Date the review before you weigh it

Both of these products have review corpora that are older than the products themselves, which makes star ratings a weak input and review dates a strong one.

Pacvue's homepage today leads with Prism, Agent, MCP and DaaS. A review written before those existed is describing software that has since been substantially rebuilt around an AI layer. Skai has carried its current name since June 2021, when Kenshoo rebranded — the footer still reads "Kenshoo, Ltd", which is why the old name keeps surfacing in search — so a chunk of the searchable commentary about "Kenshoo" is commentary about a product generation nobody is selling now.

Two practical rules follow, and they cost nothing:

• Sort by date, not by rating. Read the most recent twelve reviews of each and ignore the average entirely.
• Do not average a small sample in either direction. If a profile carries two reviews, state that it carries two reviews and decline to draw a conclusion from them. That is honest and it is also the only defensible reading.

The related diligence question is ownership. This category has moved a great deal: ChannelAdvisor is now Rithum, Kenshoo is now Skai, Perpetua has sat inside Omnicom via Flywheel since January 2024, Carbon6 was acquired by SPS Commerce, and Sunken Stone relaunched as Emplicit. None of that is a warning about any of those firms — several are stronger for it. It is simply why change-of-control, price-protection and data-portability clauses are a fair thing to ask any vendor for, including us, and why a five-year-old review is evidence about a company that may no longer exist in that form.

Three costs that sit outside both licences

Whichever way this comparison goes, the licence is not the bill. Three lines sit outside it in both cases.

• Media. Every advertising dollar is additional. Obvious, routinely forgotten when a platform fee is compared against an agency percentage that has media inside the conversation.
• Creative. Display and video units still have to be produced, sized to the placements, versioned and refreshed before they fatigue. Neither platform makes them. On a serious display programme this is not a rounding error.
• The operator. Both products multiply a competent trader. Neither supplies one. A licence with nobody accountable to it produces beautifully formatted evidence of a problem that nobody has time to fix.

The third line decides more of these projects than the first two combined, and it is the one that never appears in the comparison spreadsheet because it is a staffing decision wearing a software costume. Before you buy either, name the person, name the hours, and check that those hours are not already spoken for.

Worth adding the alternative honestly: Amazon's own DSP product page states that the managed-service option "typically requires a minimum investment of USD 50,000", with self-service available if you have a trader. So the real shortlist is usually four-way — software plus your own operator, Amazon's managed service, an agency holding seats, or doing nothing and calling it patience.

The measurement question that outlives the software choice

Whichever platform you license, you will eventually have to answer a harder question: did the display spend cause sales, or merely observe them?

Our position is that 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 somewhere in the path; it does not tell you which of those orders would have happened anyway. Those are different numbers and the gap between them is where display budgets get either defended or wasted.

It would be dishonest to present that argument as ours alone. Tinuiti runs a published measurement practice, Bliss Point, with an Incrementality Lab productised inside it — real-time experimentation to isolate causal lift, sitting alongside media mix modelling and holistic ROAS. That is substantially the argument we make, from a much larger agency, and a serious buyer will raise it. Good. It means the category is getting better.

Where we would compete is not on whether incrementality matters but on two narrower questions, both of which apply to us as forcefully as to anyone:

• Who designs and computes the test? A lift study run by the party being measured is a conflict of interest that a good process manages rather than pretends away. Ask who builds the holdout, who holds the query, and who can see the result before you do.
• Is it in the statement of work? Measurement described in a pitch and measurement written into a deliverable with a cadence are not the same product. Ask for the clause.

The third question is scope. Sponsored ads and DSP both claim credit for the same order constantly, and until they are reconciled in Amazon Marketing Cloud you are adding two numbers that overlap. Whatever you buy, ask how that reconciliation happens and who runs the query.

When a managed DSP seat is the better purchase than either licence

Some readers arrive here needing a tool. Others arrive because display is not working and a tool feels like the natural next purchase. Those are different problems and only one of them is solved by a login.

reMKTR runs Amazon DSP as a managed service on our own seats. We are part of 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.

The scoped evidence, from a direct Amazon DSP API pull covering 30 of those advertisers during July 2026 — a labelled slice, not the whole book:

• 6.04x return on ad spend across the 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

We publish the denominator because a return figure without one is a marketing asset rather than evidence. It is a reference point for how the book runs, not a forecast for your account.

We charge a percentage of media spend and we say so before you sign. That is how the category prices, and we are not going to pretend our model is uniquely virtuous — a revenue-banded flat fee genuinely wins when your spend is large relative to your revenue, and if that is your shape you should say so and make us argue for the cap.

Two honest redirects. If display is fine and the waste is really in sponsored ads — search terms taking money and returning nothing, bid structures nobody has touched in a quarter — Dr. PPC is the right door, and it is priced publicly at $300/month plus 3% of ad spend, capped, month-to-month, with the first 30 days free. If the constraint is broader than advertising, and listings, catalogue or stock are deciding whether ads can work at all, Full Circle is the parent and takes the whole scope.