Marketing and automation

Someone withdraws consent. It is recorded, correctly, in whichever system they withdrew through. Three weeks later a campaign goes out, built from a list assembled before they withdrew, and they receive it. Consonas asks the consent question again for that individual at the moment of sending, which is the only place the answer is current.

They complain, and they are entirely right to. The organisation is baffled, because the withdrawal was recorded and everybody did their job. Nobody was careless. The failure is structural: the system checked the list instead of the person, and the list was a snapshot of a moment that had passed.

This is the shape of nearly every consent failure we have seen described, and it matters more than the individual complaint suggests. In a small organisation the reputational cost lands directly on the business, and in regulated or charitable sectors it can land somewhere considerably more expensive.

Two decisions that prevent it

The first is that an audience is a definition instead of a snapshot. Customers in one county with no activity for ninety days means whoever satisfies that description at the moment you look. It does not freeze when you save it, so it cannot go stale.

The second is that consent is checked for every individual at the moment of sending rather than at the moment the audience was built. That is the difference between checking the list and checking the person, and it closes the window in which the failure above happens.

Neither of those is clever. Both are simply where the check belongs, and putting it anywhere else creates a gap that will eventually be found by someone who did not deserve to find it.

And consent is not one switch

Agreeing to receive an invoice is not agreeing to a newsletter. Agreeing to service updates about a job is not agreeing to a promotional campaign. Organisations that hold one consent value per person produce two failures at once: people who wanted operational messages unsubscribe from everything to escape marketing, and then miss something they needed.

Consent here is per channel, held as an event with a date and a source, and withdrawal is recorded the same way. Both sit in the timeline, so the question of what someone had agreed to on a particular date has an answer, which a single current value field can never give you. Electronic marketing in the United Kingdom is governed by the Privacy and Electronic Communications Regulations, which do not treat an invoice and a newsletter as the same thing either.

Not a sophisticated marketing platform

No deep visual template builder, no split testing, no landing page designer, no lead scoring model, no advertising integrations, no attribution modelling across channels. If your marketing is sophisticated enough to need those, a dedicated platform is the right purchase.

What this has that most dedicated platforms do not is that the audiences and the consent are computed from the same records your sales and service run on, rather than from a copy that has to be synchronised. For a small organisation, that is usually worth more than the features it lacks.

Opens are not treated as a measure

Clicks are recorded. Opens are not treated as meaningful, because mail clients now fetch images on the recipient's behalf and an open has become a measurement of someone's mail provider rather than of a person. Reporting it as engagement would be reporting a number we know to be wrong.

Not for sending on behalf of someone else

An agency should not run client campaigns from here. The consent records would belong to your organisation rather than your client's, which is the wrong controller relationship, and sender authentication is per organisation. That is a data protection problem waiting to happen and the tooling will not save anybody from it.

It will not make the legal judgement for you

Whether a particular send is lawful depends on your jurisdiction, your relationship with the person and what you are sending. For organisations here the guidance is published by the Information Commissioner's Office rather than by a supplier. The product records what was agreed and when so you can decide with facts and evidence it afterwards. Any supplier claiming their software makes you compliant is describing something that does not exist.

Find out what your consent records actually say

Before building anything. If your list came from a spreadsheet with a column marked yes, you have a tick rather than a consent record, and a tick tells you nothing about when, for what, or how.

Consent here is a fact with a date and a source, per channel. Where you have those, import them. Where you do not, the honest position is that you have a list of people you may not be able to demonstrate a basis for, and the right response is to fix that deliberately rather than to send and find out.

Build the audience as a question, not a list

An audience defined by a rule keeps itself current as records change. A list exported on a Tuesday is wrong by Thursday, and the ways it is wrong are the expensive ones: somebody who objected last week, somebody who is no longer a customer, somebody who has become a complaint.

Send to fifty first

The first send from any new system should be small enough that a mistake is recoverable. Read what arrives, in a real mail client, on a phone. Almost every problem anybody has ever had with a first campaign is visible in the first fifty.

Understand what happens at the moment of sending

Consent is checked for every individual person as the message is being sent, not when the audience was built. Somebody who objects between building and sending is not sent to.

That distinction is the whole difference between a system that can be relied on and one that produces an apology once a quarter. It also means an audience count and a send count can legitimately differ, which is the system working instead of a fault.

Watch what came back, not what was opened

Objections and complaints are the numbers that matter and they are the ones nobody puts on a slide. A campaign with excellent engagement and four objections has told you something about the next campaign that the engagement figure never will.

The audience definition

An audience is a name, an optional description and a set of conditions. Each condition names a field, a comparison and a value: the relationship type is customer, the address has something in it, the record was created before a date. Every condition has to hold for someone to be in it.

The fields a condition may name are a fixed list: the name, the address, the kind of record, its status, its relationship type, its owner and the date it was created. A definition naming a field that did not exist would match no one whilst looking exactly like a working audience, which is the worst of the failures available. Twelve conditions is the ceiling, because no one writes a sensible audience with fifty clauses in it and every clause costs work on every record.

The definition is written back to you in words on the list of audiences, so what you are about to send to is legible without opening anything. Relationship type is customer, and email has anything in it, reads as a sentence somebody can check.

The template

A name, a subject and a body. Anything inside double braces is filled in for each person as the message goes out: your organisation's trading name, telephone and address, and the person's full name, first name and address. When you save a template it tells you which of those it found, because a template quietly missing one is how somebody receives an email addressed to no one.

A placeholder with nothing to fill it is left exactly as it was written rather than replaced with a space. That is deliberate. A gap in a sentence reads as a mistake in your writing, whilst braces still sitting in the middle of a paragraph read as a mistake in the template, which is what it actually is.

The campaign itself

A name, one audience, one template, a purpose and a state. The purpose is the important part: it is what the consent check is made against, so a campaign whose purpose is marketing reaches the people who agreed to marketing and nobody else. The state is draft or sent, and there is no third value.

What the preview tells you

Before sending you can ask what would happen, and nothing is written. You are told how many people are in the audience, how many would receive it, and how many would not, grouped by reason. There are four reasons and they are the words on the screen: no address, restricted from processing, never agreed to hear from you, and asked you to stop.

You also get a handful of names as a sample, and that sample is filtered by what you personally may see whilst the counts are not. The counts are a fact about the campaign and someone planning one needs the real number. The names are facts about people, and they answer to the permission on those people.

What is left behind afterwards

A row for every person in the audience, whether they were written to or not, carrying what happened, the reason where there was one, and the time. Counts of how many were reached and how many were left out. A timeline entry on each person who received it, naming the campaign. And on each of them, the campaign recorded as the last thing that touched them, and as the first thing if nothing was there already.

Those recipient rows are the reason a complaint can be answered at all. Somebody asking whether you wrote to them, and when, is asking a question that has a row as its answer rather than somebody's memory of a Tuesday.

Every change to a record is settled against every audience

Creating a contact, changing one, moving one to the bin, restoring one, converting an enquiry and taking a form submission all decide that one record against every audience definition as part of the same change. The cost is one pass over your handful of definitions, rather than every audience costing a pass over every record, which is why opening an audience is instant whatever you hold.

A workflow condition is an audience condition

The conditions you put on a rule are written in the same language as an audience and are settled by the same code. That is a deliberate refusal to have two condition languages: a second one would be a second thing to learn, a second thing to test, and two places for the same question to be answered differently.

The permission checked is the one on the person

Membership is a property of a record rather than of an audience, so someone who may not see a contact does not see them inside an audience either. An audience is not a way around who may see whom, and neither is the preview of a campaign.

The count is the sharper version of the same problem. The stored count is the true one, because that is what would be written to, but the count shown to a person is filtered to what they may see. An unfiltered total over a definition somebody writes themselves is a way to interrogate records they are refused: name starts with A, then AB, and a sensitive contact can be read out a character at a time by watching the number move.

Sending writes into the timeline and into attribution

Each person written to gets a timeline entry naming the campaign, which is what makes the send visible to whoever picks up the phone afterwards without them having to know a campaign happened. The campaign is also recorded on that person as the last thing that touched them, and as the first thing only where nothing was there before, which is what makes first touch actually first.

Opportunities carry both, so what a campaign produced is counted twice over: what it brought in and what made someone act. Those two answers are usually different, and an organisation choosing between them is choosing what to spend money on. Totals are grouped by currency, because adding pounds to dollars produces a number that is not an amount of anything.

Automation shares its vocabulary with webhooks

The events a rule can trigger on are the same closed list an outgoing webhook can subscribe to, published from the same place, so a rule and an external system hear about exactly the same things. One vocabulary for what can happen in the product, instead of one for automation and a different one for integrations.

Sensitivity travels with what automation creates

When a rule puts somebody on a sequence, the owner and the sensitivity of the contact are carried onto the enrolment. Without that, chasing a sensitive contact would be more visible than the contact is, and a list of enrolments would name people the reader is refused everywhere else in the product.

What the plan decides

Audiences, campaigns and automation sit on the plans that include marketing, which the pricing page states plainly, and all three are refused by the plan as well as by permission. Building an audience is included in that: a plan without marketing used to let you assemble a list you could never write to, which is a worse answer than a refusal because you find it out at the end. When the plan refuses, it names the plan that would include it rather than saying no.

The plan is checked again at the moment of sending rather than only when the campaign was written. A campaign drafted during a trial does not go out after the plan it needed has lapsed, and a rule stops running when automation stops being included. That second check exists because a marketing send has a direct cost and reaches people who are not your colleagues.

The allowance, and what actually counts against it

The allowance counts people written to over the previous thirty days, not the size of your audiences. Somebody left out because they never agreed, or because they asked you to stop, does not count towards it, which is the fair reading: you are charged for what went out rather than for what you considered.

A campaign that would cross the line is refused before anything at all is written, so it stays a draft rather than sending to some of the audience and stopping in the middle. Automation is counted the same way and refused the same way, run by run.

What is never behind a plan

Consent records, the audit trail, subject access tooling, two factor sign in and complete export are on every plan including the free one, permanently. That is deliberate and it is not a trial position. Recording that somebody withdrew, and being able to demonstrate it later, must never be something an organisation cannot afford, and a supplier who puts the obligations behind an upgrade has made the obligation into a sales tactic.

The consequence is worth stating in the direction that costs us something: an organisation on the free plan can hold consent properly, evidence it, answer a subject access request and take all of its data away, and can do none of the sending this page describes. What that free plan costs us to run is written up in the measurements behind it.

The trades where consent is the hard part

Consent records earn their keep where the same person hears from you in several different capacities. Charities hold supporters who gave for one purpose and not another, membership organisations have to separate a renewal notice from a campaign, education and training writes to learners, employers and funders about different things, and retail services has to keep an order confirmation apart from a promotion. Those pages describe how each of them models it.