Pardot · Implementation in 2026
A first Pardot implementation in 2026
is usually the wrong project.
If you do not already own Account Engagement licences, every edition sold today comes with a Marketing Cloud Next edition, so the real choice is how much of each you will use, not which one to buy. If you do own them, implementing properly is often the cheapest thing you can do, because the licence is paid for either way. Which of those two you are decides everything below, and it takes about a minute to establish.
Updated 25 September 2026. Facts below are limited to what Salesforce has actually stated.
Before anyone quotes you
What does a Pardot implementation mean in 2026?
The word means something different than it did three years ago, and most quotes still describe the older version of it. Four facts change the shape of the project: three stated by Salesforce, and one that is my reading and labelled so.
The product you are implementing is called something else
Pardot became Marketing Cloud Account Engagement. Nothing about the software changed at the rename, and most teams still say Pardot, which is why this page sits at this address. If a proposal talks about Pardot and a licence order talks about Account Engagement, those are the same thing.
New AI capability is built in Marketing Cloud Next
Account Engagement still receives features in each release, but Agentforce, SMS and cross object merge fields in Account Engagement require Data 360, formerly Data Cloud. That matters for an implementation specifically, because you are choosing an architecture you will live inside for years, not buying a tool for a quarter.
Marketing Cloud Next now comes with it
Marketing Cloud Next was announced at Connections in June 2025, built on the Salesforce Platform with Data 360 and Agentforce. Every Account Engagement edition on the price list today includes a Marketing Cloud Next edition, and Salesforce says existing customers get its features with no migration required. Implementing Account Engagement today means owning both, whether or not you use the second yet.
There is still no announced end of sale
No published date, for sale or for support. So implementing is not reckless and nobody can honestly tell you the platform dies on a date. Anyone quoting you a deadline is either reading a private communication you have not seen, or inventing urgency. Ask them which.
The answer depends on one thing
Which of these three are you?
Not on the roadmap, and not on the size of your team. On whether the licence is already bought, and on what is already sitting in the org.
You already own the licences and nothing is built
Implement, and do it properly. You are paying for the platform whether it runs or not, and a working Account Engagement is what makes any later move cheap. The export you take to the next platform is the engagement history, and history only accumulates if something is actually running.
You are evaluating and have bought nothing
Then the question is not how to implement Pardot, it is which platform to implement. Ask your account executive directly what Marketing Cloud Next means for a new customer signing this quarter, and get the answer in writing. That one email is worth more than any implementation quote.
Somebody implemented it already and it does not work
This is the most common of the three and the one most often mislabelled. What gets asked for is a fresh implementation. What is usually needed is a repair, because the licences, the connector and the prospect history are all still there and only the architecture is wrong. A rebuild throws away the one asset you cannot buy back.
Scope
What is actually in a Pardot implementation?
Six pieces. Quotes vary wildly because most of them silently leave out the last two, and the last two are where implementations fail six months later.
The Salesforce connector and field mapping
Which records sync, in which direction, and what wins when both sides changed. This is the decision that is hardest to reverse later, and it is routinely made in ten minutes during setup.
Capture: forms, form handlers and landing pages
Where prospects enter. Form handlers behave differently from forms in ways that lose leads without raising an error, which is a whole article on its own.
Scoring and grading, kept apart
Scoring is behaviour, grading is fit, and merging them is the single most common architectural mistake in Account Engagement. A rebuild of a model that already exists is 45 hours at €2,250, which is a reasonable yardstick for building one from nothing.
Engagement Studio programs
The nurture itself. Build few and finish them. Most orgs I open have between fifteen and forty programs of which a third have produced no measurable outcome in a year.
Attribution back from won deals
Connected Campaigns and campaign influence, so the pipeline number and the marketing number come from the same place. Left out of most quotes, and the reason sales and marketing argue about which leads were real.
A named owner and written documentation
Who runs a campaign on day 91, and what they read to do it. An implementation without this is a system that works until the person who built it stops answering, which is the state most of the repairs on this site begin from.
From practice
What this looks like when it has gone wrong.
One case, with the numbers, because a page about implementations should show what a bad one costs.
A Canadian marketing consultancy, several Pardot clients at once
Every client org had been implemented separately by hand, and the recurring work of moving reports and fields between them took over eight hours each time. Nothing was broken in a way that raised an error. It was simply built without the assumption that it would ever have to be repeated.
Re-architecting the reporting and field migration logic brought the same process to about forty minutes, which moved delivery capacity from one client per ten hours to ten or more in the same window, with the sync errors eliminated along the way. The lesson transfers directly to a single implementation: the expensive part is never the initial build, it is everything you have to redo because the build assumed it was final.
The engagement
What exists at the end, and what I do not do.
Most implementation services describe activities. This describes what you are left holding, because that is the thing you are buying, and then the parts that are deliberately outside it. One person does the work, and the same person answers the email afterwards.
A written scope before anything is built
It names what is included, what is not, and the fixed price against it. If I misjudge how long something takes, the figure does not change.
The connector and field map, written down
Which fields sync, in which direction, and which side wins a conflict. Documented as the work happens, because this is the decision that is hardest to reverse later.
Capture tested end to end, not configured and assumed
Every form and handler submitted once for real, and the prospect found in Salesforce afterwards. A form that reports success proves nothing on its own.
Scoring and grading built as two axes
Behaviour and fit kept apart, negative scoring in from the start, and the threshold set from closed deals rather than from a template. The eight ways that model breaks is the same list I build against.
Attribution wired back from won deals
Connected Campaigns and campaign influence, so the pipeline number and the marketing number come from the same place rather than from two reports that disagree.
A named owner, and sixty days after handover
Who runs a campaign on day ninety one, and what they read to do it. The same sixty day handover the other builds on this site carry, and after that support is a separate plan with an end date.
Outside the scope, on purpose
Five things this is not.
Naming them is cheaper for both of us than discovering them in week three.
Website design and page building
Landing pages and templates are assembled from what your brand already has. If the brand needs designing, that is somebody else and I will say so.
Writing the nurture content
The programs get built, the sequencing gets designed, and the words in the emails are yours. Marketing automation cannot rescue content nobody wanted to read.
Any other platform's build
This is Salesforce and Account Engagement work. I do not build HubSpot or Marketo, and when the honest answer is that one of them fits you better, you will hear it.
An open ended arrangement
Every engagement has a written scope, a fixed price and an end date. Nothing renews by itself and nothing runs indefinitely.
Data you do not have
An implementation cannot invent consent, history or a list. What it can do is make sure what you collect from here is clean, and say plainly what was missing at the start.
Cost
What does a Pardot implementation cost?
Every price on this site is published, so here are the three that bracket the question. None of them requires a call to obtain.
Measure what is already there
Before an implementation is scoped, four measures taken against your own exports. Free in the browser version, €500 for the reviewed one delivered in a working day.
Scoring and grading built or rebuilt
Forty five hours. Scoring and grading kept separate, negative scoring, thresholds back tested against your own closed deals rather than against a template.
Move to Marketing Cloud Next instead
One hundred and twenty hours at the base shape. If you are case B above and have bought nothing yet, price this against the implementation before deciding, because the numbers are closer than most people expect.
Describe it in three lines.
Whether the licences are bought, what is already built, and whether there is a date attached. That is enough for me to tell you which of the three cases you are, and whether this is an implementation at all.
Serhii Skrypnyk · Senior Salesforce Administrator and developer · 7 Salesforce certifications · on the platform since 2018. Reviewed 25 September 2026.
This creates a record in Salesforce and sends you a confirmation. Elsewhere, an estimate is recorded only when you act on it, as the privacy policy lists, and a readiness result only when you ask for its breakdown.
Questions
Asked often enough to answer here.
Do you offer Pardot implementation services, or only advice?
Both, and they are priced differently. The advice on this page costs nothing. The build is a scoped engagement: a written scope in five working days, a fixed price against it, the six pieces above delivered in your own org, and sixty days after handover. It is one person doing the work rather than an agency, so there is no account manager and no proposal deck.
Should we still implement Pardot in 2026?
If the licences are already bought, yes, because you are paying for the platform whether it runs or not and a working setup is what makes any later move cheap. If you have bought nothing yet, ask Salesforce in writing what Marketing Cloud Next means for a new customer before you sign anything.
How long does a Pardot implementation take?
The build is rarely the long part. Connector decisions, capture and a first nurture program can stand up quickly, and what extends the timeline is agreeing which records sync and who owns the system afterwards. Any quote that does not name the owner on day 91 has left the expensive part out.
What does a Pardot implementation cost?
The published anchors on this site are €500 for a reviewed readiness measurement, €2,250 for forty five hours of scoring and grading work, and €6,000 for a one hundred and twenty hour migration to Marketing Cloud Next. An implementation is scoped against those rather than against a rate card you cannot see.
Is it cheaper to implement Pardot or to start on Marketing Cloud Next?
Closer than most people expect, which is exactly why it is worth pricing both before choosing. The comparison that matters is not the first invoice, it is the second one, when whichever platform you chose has to be moved or extended.
Our Pardot was implemented badly. Do we start again?
Almost never. The licences, the connector and the prospect engagement history are still there, and the history is the one asset that cannot be bought back or recreated. Measure which part of the architecture is actually wrong, then repair that part.