2027 Retail Leaders’ Series · The martech myths
Five martech myths you are still paying for.
Retailers were sold a software-era bargain: pay by volume, clean your data first, accept one platform built for every industry, and trust that the vendor’s AI is built for you. None of it holds in the AI era. This report takes each myth apart and shows where the money and the value actually go.
- 33%
of licensed martech capability actually used
Gartner, 2023
$0
per message for mobile push via Google’s Firebase Cloud Messaging
Firebase pricing
- $0.10
per 1,000 emails through a cloud email API, lowest public tier
AWS SES pricing
- 60–80%
of martech cost we expect to disappear in the agentic model
Replenit view
About this report
Myth one · Pricing
The myth
More contacts and more messages cost more to deliver, so the bill should grow with them.
The truth
Delivery is a commodity. Some channels cost nothing to send. You are paying a volume tax on a software-era interface.
Profile tiers, message bands, monthly active users, events processed. Every pricing lever grows with your customer base, whether results grow or not. Meanwhile the underlying delivery has collapsed in cost: Google lists Firebase Cloud Messaging for mobile push among its no-cost products, and a cloud email API sends 1,000 emails for $0.10 to $0.23. The premium is not for delivery. It is for the canvas you build journeys on.
Exhibit 1
What you are billed for vs what delivery actually costs
Channel
How martech typically prices it
What delivery itself costs
Per profile tier and per send volume band
$0.10–0.23
per 1,000 via cloud email API
Mobile push
Per monthly active user or per message
$0
Firebase Cloud Messaging is a no-cost Google product
Web push
Per subscriber
Open browser standard; no per-message fee from the browser
In-app and onsite
Per MAU, per impression or per module
Rendered by your own app and site from an API response
Data
Per event, data point or profile stored
Already stored in your warehouse; billed again by the tool
Sources: AWS SES public pricing ($0.10 to $0.23 per 1,000 outbound emails by plan); Firebase pricing page, Cloud Messaging listed under no-cost products. Martech pricing models vary by vendor and contract.
Myth two · The interface
The myth
The journey builder, the segment editor and the dashboards are what you are paying for, and they are worth it.
The truth
The interface was the value in the software era. In the AI era, software is a commodity and the value is in the decision.
Martech platforms were built so people could operate them. That made sense when people did the deciding. Now an agent can reason about every customer individually, and a canvas of nodes and triggers is no longer an asset; it is a workaround. Yet the price has not moved. Retailers pay 2015 prices for a 2015 way of working, and only 33% of what they license is used.
Exhibit 2
What you paid for in the software era vs what creates value now
Question
Software era (what you still pay for)
AI era (where value is created)
The asset
A UI to build segments and journeys
Skills that decide for every customer
Who does the work
Your team, inside the tool
An AI CRM Manager, inside your guardrails
What scales cost
Contacts, messages, seats
Outcomes
What a new idea needs
A new build in the canvas
A new skill, written in plain language
What the AI does
Tunes send time and subject lines
Decides product, timing, channel, tone, incentive
Utilisation: Gartner Marketing Technology Survey 2023, n=405 (58% in 2020, 42% in 2022, 33% in 2023).
Myth three · Data
The myth
Clean, complete, unified data comes first. Then, maybe in a year or two, AI.
The truth
Tools that can only read clean columns need perfect data. An engine that enriches data does not.
Retailers have been told to fund long data projects before any AI can run, because that is what a system of record needs to work. Gartner predicts organisations will abandon 60% of AI projects unsupported by AI-ready data, and 63% are not sure they have the right practices. The answer is not a longer data project. It is a layer that turns imperfect order and product data into a living understanding of each customer and product.
Exhibit 3
System of record vs living memory
Question
System of record (CDP, MAP)
Living memory
What it holds
The fields you gave it
What the data implies: household, routine, fit, cycle, price behaviour
Missing data
A blocker: fix it before you start
Inferred from purchases and enriched
Products
A catalogue feed
Derived relationships: what pairs, substitutes and replenishes
Over time
Stays as accurate as the last manual fix
Learns from every outcome
Data needed to start
A unified, clean profile
Order history and a product feed
Gartner press release, 26 February 2025, “Lack of AI-ready data puts AI projects at risk”.
Myth four · One size
The myth
A leading engagement platform works for any business, so it works for retail.
The truth
It was built to be sold to everyone. Its generic limits are then sold back to retailers as premium modules and services.
The same canvas is sold to a bank where no commerce happens, to a pure-play online retailer, and to an omnichannel retailer with stores and an app. None of them needs the same thing. Retail runs on replenishment cycles, size and fit, returns, stock, markdowns and households. The platform does not understand any of these; it gives you custom attributes and a services team to build around the gap.
Exhibit 4
Three very different businesses, one identical product
The same segment editor, the same journey canvas, the same AI add-ons
Retail-specific needs become custom attributes, professional services and premium tiers.
Exhibit 5
What retail needs vs what a generic platform offers
Retail need
Generic platform
Retail skill
Replenishment
A fixed “X days after purchase” trigger
Each customer’s own consumption cycle per product
Size and fit
A custom attribute, if someone maintains it
Inferred from what a customer keeps and returns
Margin
Blanket discount to a segment
Discount only where it changes behaviour
Product churn
Invisible while the customer stays active
Caught when one recurring product is dropped
Household
One profile, one email
Each purchase attributed to the member it is for
Myth five · Whose AI
The myth
Each new AI feature from your vendor is an investment in your business.
The truth
It is an investment in their retention. The smarter their tool gets on your data, the harder it is to leave.
Vendor AI lives inside the vendor’s tool, learns inside it and only works on journeys built inside it. Every model trained on your customers makes the switching cost higher and the renewal safer for them. You pay for the AI, you supply the data, and they keep the asset. 45% of martech leaders using vendor-offered AI agents say those agents miss the business performance they were promised.
Exhibit 6
Whose intellectual property is it?
Question
Vendor AI inside the tool
Your skills in a decision layer
Where the logic lives
In their journeys and models
In your skill library
Who owns the learning
The vendor’s product
You; never used to serve anyone else
What it optimises
The campaign you already chose to send
Your commercial objective, per customer
Cost of leaving
Rebuild everything in the next tool
Swap the delivery endpoint; skills keep running
Who it serves
Their renewal
Your P&L
Gartner, October 2025, n=413 martech leaders: 45% say vendor-offered AI agents fail to meet expectations of promised business performance.
The value shift
Strip the myths away and the martech bill splits in two. The part that delivers messages becomes a commodity priced at cost. The part that decides moves to an intelligence layer you own and pay for on outcomes. Everything in between, the premium tiers, the volume tax, the canvas and the hours spent operating it, stops being necessary.
Replenit view
Our expectation: retailers will cut 60 to 80% of today’s martech cost as the agentic operating model matures. No rip and replace on day one; the cuts land at renewal, line by line, as decisions leave the channel tools.
Share of today’s martech cost we expect to disappear
Exhibit 7
Where each line of the martech bill goes
Cost line today
Why you pay it
In the agentic model
Premium AI modules
To tune campaigns inside the tool
Removed: decisions are made once, for every channel
Journey and segment tiers
To build plays by hand
Removed or down to a core tier
Volume-based sending
Profiles, messages, MAUs
API-driven delivery at cost
Data and event fees
Re-billing data you already store
Processed once in the decision layer
Services and workarounds
To fit a generic tool to retail
Replaced by retail and industry skills
Team time operating the tool
Building, QA, reporting
Moved to objectives, skills and results
Intelligence (new line)
Did not exist
Paid on outcomes, measured against holdout
Before your next renewal
What do we pay per message, per profile and per event, and what does delivery actually cost you?
The gap is the volume tax.
Which features do we pay for and not use?
The market average is two thirds idle.
What would it take to start AI without a data project first?
If the answer is “nothing until the data is clean”, the limit is the tool.
What in your product is built for retail, and what is a custom attribute?
Workarounds are the generic tax.
If we leave, what do we take with us?
If the answer is “nothing”, the AI was built for their renewal, not your business.
Don’t buy another tool
Maestro is an AI workforce for retail CRM. You hire it like a team member: your team defines the decision logic and the guardrails, and Maestro runs the work for every customer, every day, on the stack you already have. It is accountable for the commercial outcome, not for usage.
The next step
A sizing session to map your objectives to Maestro’s skills and size the case on your own numbers.
You were sold software. What you need is decision logic you own.
Keep delivery cheap and interchangeable. Put the intelligence, and the intellectual property, on your side.