Dividing the work between provider and business
The provider might be an agency, a freelance lifecycle specialist or a studio that writes and sends the newsletter. Whichever it is, the work splits into the roles that the lifecycle email programme playbook names: a marketer who owns the programme and its test plan, a developer who builds templates and journey logic, an analyst who owns segments and measurement, and a compliance reviewer who owns the consent and data-use questions. A single provider may fill some of those roles and not others, so the business lists the roles it needs before comparing proposals and says who will fill the ones left over.
The email marketing guide sets email out as five connected parts, and a proposal can be read part by part against what stays on the business's side of the table.
| Part | What a provider can do | What stays with the business |
|---|---|---|
| List | Design sign-up points, write the promise beside each, run the confirmation step and keep the register of sources | The decision about what the business relies on to mail each person, made with whoever owns data protection |
| Deliverability | Set up authenticationRecords and signatures, SPF, DKIM and DMARC, that let mailbox providers check that a message comes from the domain it claims., stage the sends and read the postmaster and DMARC reports | The domain itself and the access to its DNS records |
| Content | Write, design and check newsletters and flows | The claims, prices and offers, and the sign-off of the email and its audience |
| Automation | Build flows with their triggers, timing and exits and run the launch checks | A named owner for each flow after launch, and the means to pause every flow |
| Measurement | Report by layer, run tests and keep the log | The order data and the attribution rule, agreed with whoever owns analytics |
Two questions come first, and other pages answer them. Whether outside help is wanted at all is the subject of in-house against agency, and the contract, scorecard and reviews around any provider follow the cycle in the agency selection and review guide. The platform is the second. As general practice, a provider who arrives before the platform is chosen ends up choosing it, so the business decides first, from the programme's requirements as the guide to choosing an email platform sets them out.
Where the platform, the domain and the records sit
A business owns its email channel: the addresses were given to it and the sending identity is its own. A provider works through access to that channel. Withdrawing the access then leaves the list, the domain and the records where they are, as long as none of the underlying objects was built somewhere else. These are the objects to settle before the first invoice.
| Item | Reason it stays in the business's name | Access given to the provider |
|---|---|---|
| The email platform account and its administrators | The business needs the right to leave, which means an export that carries the consent fields, the suppression files, engagement history and flow definitions detailed enough to rebuild each flow | A named user with the roles the work needs, not a shared login |
| The sending domain and its DNS records | Authentication lives in DNS, which marketing cannot change alone, and each system that mails as the domain has to appear in SPF and sign with DKIM | A written request for each record, applied by whoever holds DNS |
| The consent recordsThe stored evidence of a marketing permission: the channel, the wording the person saw, the date and the source of each opt-in, kept with the contact record. and the sign-up wording | A record holds where and when a person agreed, on what wording, the choices made and any confirmation, which is what lets the list be defended | Read access, and a duty to add a complete record for every address it adds |
| The suppression listA list of addresses or contacts excluded from every send, such as unsubscribes, hard bounces and spam complaints. | It records who must not be mailed, and stays on file rather than being deleted, so that an import cannot bring a person back | A duty to load it before any contact in a move and to apply it to every send |
| Segment and flow definitions, templates and forms | Written definitions are what let another person rebuild a flow or diagnose a fault | Write access while engaged, with a dated note for every edit to a live flow |
| Order and customer data | Results are judged where the sale is recorded, under a window and model agreed beforehand | Read access to what the reports need |
| The test log and the reports | The log of tests, nulls included, is what the programme learns from | Write access while engaged |
The right-hand column is general practice: the least access that lets the work be done, granted to named people.
Put one question to every candidate before any other: in whose name would the account be opened? As general practice, an account opened under the provider's own business means a rebuild when the engagement ends; an account in the business's name carries on. The suppression file carries particular weight at a change of hands, because the email list hygiene SOP keeps opted-out contacts on file indefinitely and treats suppression as the opposite of deletion. When more than one tool acts on the same addresses, the provider's own tools are among them, so each belongs on the consent map across the marketing stack. Each grant of access goes on the agency onboarding checklist, and a dated Access ReviewA periodic check of who can use each marketing tool: every account matched to a current person, supplier or integration, with owners, deputies and narrow roles. checks the grants against the people who still work for the provider.
Splitting the opening engagement into quotable pieces
As general practice, ask for an audit, a domain and platform set-up (or a move), a flow build, campaign production, list growth and reporting to be quoted as separate items, so that the business can buy each one, judge it and stop it without taking on the rest, and begin ongoing sending and reporting on a retainerA standing fee paid regularly to an agency or supplier for continuous work with a stable scope, as distinct from project fees, time and materials or commission. only once the first pieces have been checked.
| Job | Settle before it starts | Proof when it ends |
|---|---|---|
| Programme audit | That the email programme audit scorecard is the yardstick | Fixes in ranked order with an owner for each, and consent and authentication scored first, as gates, ahead of content |
| Domain and platform set-up, or a move | Who holds DNS and who requests each record | Authentication verified with test messages read header by header, suppressions loaded before contacts, a register of what moved, and records sitting in DNS the business controls |
| Flow build | The trigger, timing, exits and owner of each flow | Test contacts taken through every branch of the flow, including one who buys during the sequence, as the flow launch checklist asks |
| Campaign production and sending | Who approves, and what the approval covers | A calendar, built emails and quality checks, with a sign-off on both the email and the audience, as the campaign send SOP sets out |
| List growth | The wording and opt-in method of each sign-up point | A register row for each point and a complete consent record for every address added |
| Reporting and testing | The attribution rule and which flows get a holdout | Layered reports and the test log, the rule printed on every outcome figure and holdouts running where they were agreed |
As general practice, where two suppliers split the job, say a studio that writes and a specialist who builds and sends, give each its own scope from the scope of work template, listing what that supplier delivers and what it leaves out, and name who approves the audience as well as the email.
The written request: questions that test consent and sending
As general practice, send every candidate the same written request, so that answers can be compared line by line.
| Put to each candidate | The answer should name |
|---|---|
| Which addresses will you add to our list, and from where? | Only people who asked for this mail, each with a source, a time, the wording they saw and any confirmation time, and no bought or rented files |
| Which opt-in method will each sign-up point use? | A method chosen by source: confirmation where forms are open to anyone or addresses are keyed in, single opt-in where the address is already tested, as at checkout |
| How will you set up and test authentication on our domain? | SPF and DKIM covering every system that mails as the domain, DMARC starting at monitoring, alignment with the From address people see, one-click unsubscribe headers and test messages whose headers are read |
| What will you watch after each send, and what makes you stop? | Bounces by type and source, complaints and unsubscribes, the postmaster view where the domain is verified, and sends staged so that they pause when the signals worsen |
| How will an opt-out that arrives by reply, call or message reach every tool? | One global suppression list, a log of each request and the checks the opt-out procedure sets |
| How will you decide who receives each send? | A short set of standard attributes combined at send time, engagement tiers built from clicks and activity, and inactive subscribers asked to confirm before they are suppressed |
| How will each flow end, and who owns it after launch? | Exits checked before every send, a rule for people who qualify for two flows, a named owner, a review date and a way to pause every flow |
| What will each report show, and from which data? | Layers kept apart, clicks and outcomes ahead of opens, the attribution window printed and flows read against a holdout where volume allows |
| How do you test, and what happens to a result that shows nothing? | One question per test, set before the send, judged on clicks or conversions and logged even when it found nothing |
| Who approves each send, and what do they approve? | An approver on the business's side who confirms the finished email and the list it will reach |
The site's own guides give a yardstick for each group of answers. On the list, the list building and consent guide regards a list as permission rather than a database, stages every import and records how each address arrived; it points to the regulators' own guidance, and the GDPR asks that consent be demonstrable. Choosing the method source by source is what the single versus double opt-in comparison is for, and the sign-up source register template gives every sign-up point a row that a provider can be asked to fill.
On the sending identity, the deliverability guide explains that Google's sender guidelines document requirements as a floor, not a promise of placement, and that DMARC moves from monitoring towards enforcement only as each legitimate source passes. As general practice, a provider's answer on authentication is a guess until it can list every system that sends as the domain. Each line of the set-up is on the sending domain checklist, and DeliverabilityWhether sent email reaches the inbox rather than spam or rejection, decided by sender reputation. itself is decided by mailbox providers from identity, reputation and behaviour, which is why the question after authentication is about the list and the sending pattern.
On opt-outs, a request made in a reply or on the phone has to reach every tool, and the marketing opt-out request SOP sets out the procedure that does that; as general practice, a provider's staff should be able to follow it without improvising. On flows, the automation section names exits as where flows are weakest and sets out the owner, review and kill switch that an unattended flow needs. On choosing who is mailed, the segmentation guide sets the tiers and the re-permission step that comes before suppression.
The questions in the first column can feed the questions section of the agency request for proposal template. Group the answers in a weighted scoring matrix under email-shaped criteria such as list and consent, sending and authentication, flows and segments, and reporting and testing. Ownership of the platform account and the sending domain sits outside the weights: the framework treats a must-have as a gate, so a proposal that fails it is out, however well it does elsewhere. As general practice, write the gates down with a date before any answer is opened, so that they cannot drift to suit a favourite.
Reporting that holds up against the business's own orders
Make each of these a line in the scope of work.
- Layers kept apart. The email metrics and testing guide separates delivery, engagement, outcome and health, because a report that merges them hides which one moved.
- One attribution rule, printed beside every outcome figure. Changing the window or the model changes the figure, and the change can pass for a swing in performance, so the rule is agreed with whoever owns analytics and kept stable.
- Results read against orders. If email-credited orders fall while clicks hold, start with the email clicks without sales diagnosis. When a provider's report runs ahead of the sales ledger, the agency reports against flat sales diagnosis puts the provider's number next to the business's own.
- A holdout beside the flows that carry the largest spend or incentive. As the metrics guide describes it, a holdoutA randomly chosen part of an audience kept from exposure to measure what exposure caused. is a randomly chosen share of the qualifying people who are held back and sent nothing; set against the people who received the flow, it shows what the flow caused and not merely what it touched. A split made at sign-up instead is worked through in the guide to subscriber value.
- Revenue per EmailEmail-attributed revenue divided by the number of emails delivered. with its caveat. It values a send under the attribution rule rather than a person, so a higher send frequency can lower it while total revenue still grows.
- List health beside results. Unsubscribes, complaints, bounces and authentication results show whether the list that the results depend on is being worn out.
- The test log. Each test with its question, its metric and its outcome, nulls included.
What a weak proposal gives away
- A growth plan built on bought or rented addresses. The list building guide says such lists almost never pass the test of whether the people agreed, and that the damage lands on the sending domain, so every legitimate subscriber feels it too.
- A promise of inbox placement. The documented requirements remove reasons for rejection and earn no inbox, and a figure quoted from other senders' lists describes those senders. A guaranteed rate is a promise about a decision that mailbox providers make.
- Sending as the provider instead of as you. A platform needs to sign with DKIM using the business's domain rather than only its shared one, so that DMARC passes, and a plan that leaves only the provider's domain on the mail does not meet that.
- Opens as the headline. An open is partly machine activity now. The open-rate and click-rate comparison sorts out which figure still earns trust for which job, and a target set on opens, or a flow that exits on an open, builds on the weaker signal.
- A whole-list send to reach a volume. Mailing the full list after months of segmented sends brings a volume jump and the inactive addresses together, which is the pattern the deliverability guide warns against.
- Moving platforms, importing a list and changing cadence all in one week. The deliverability guide advises keeping them apart and says a platform migration reads to mailbox providers as a change in behaviour.
- Consent shortcuts. Pre-ticked boxes, bundled consent, a download treated as a subscription and badge scans assumed to be consent all appear in the list building guide as things to avoid.
- A flow report with no holdout. The flow's own report credits every order it touched, so the better a flow targets, the more it flatters itself, and an incentive decision made from it overvalues the flow.
- No word on suppression. A proposal that does not say where unsubscribes, complaints and bounces go, or how the list survives an import, has not planned for the moments when lists are damaged.
How long to wait before judging the work
As general practice, hold off on a verdict until the changes behind the figures have settled and a full sending cycle has closed, seasonal sends included where they matter to the business. At the start a provider's work is tangled up with the list it inherited and with any change of platform, domain or cadence. A move to another sending platform reads to a mailbox provider as a change in behaviour, so the advice in the migration tutorial carries over: begin with the most engaged people, release the rest in stages and, as general practice, put the date of the change on every report.
Read each send beside earlier sends of the same type, not beside an industry figure. A small flow may not gather enough entrants for a holdout to be readable, and its report then carries that caveat. Some drops belong to neither side of the contract. The diagnoses of an open rate that dropped and of emails going to spam separate measurement, placement, audience and copy, and some of those causes sit with the list and the promises it was built on, which belong to the business. If the work still falls short after that, the agency turnaround tutorial gathers evidence against the scope and ends in a written plan.
Handing over and leaving: the file to keep from the start
As general practice, the email share of a handover is a written inventory, easier to keep up as the work goes than to rebuild at the end: the platform's administrators and a recent full export, the DNS records and which system each one authorises, the consent records, each suppression file as a separate type, the written definitions of segments and flows, the forms and their confirmation messages, the calendar, the test log and the source of each report. The register in the email platform migration tutorial is the model, and the tutorial also fixes the order of a move, which the handover list should follow.
The agency offboarding checklist covers the remaining steps of an ending. If results slip after a new provider starts, the diagnosis of results that dip after a change of provider lists what a handover can leave behind.
Taking over from a freelancer: an origami paper shop (illustrative)
Tibruvel Origami Supplies takes its email back before hiring
Tibruvel Origami Supplies, an invented online shop selling folding paper, kits and tools, has 6,200 subscribers. For two years a freelance friend of the owner has run its email from a platform account opened under the freelancer's own business, with mail signed by the platform's shared domain. The freelancer is moving on and the owner wants a provider to take over. Before inviting any proposal she lists the main objects this guide names and where each sits today.
| Object | Where it sits today | First move |
|---|---|---|
| Platform account | Under the freelancer's business, with one administrator | A new account in the shop's name; the freelancer exports into it and the incoming provider joins as a named user |
| Sending domain | Mail signed by the platform's shared domain | Records for a subdomain of the shop's own domain, requested in writing and published by whoever holds the shop's DNS |
| Consent records | 2,150 of the 6,200 addresses carry a source and wording; 4,050 came from an old prize-draw form with nothing on file | The prize-draw addresses wait in a staging segment while the owner and whoever owns data protection decide what the shop relies on for them |
| Suppression file | 310 unsubscribed addresses exist only inside the freelancer's account | Exported and loaded into the new account before any contact |
She sends one written request to a lifecycle specialist and a small agency and writes two gates down, dated, before reading either answer: the account and the domain's records are in the shop's name, and no address joins the list without a consent record. Both clear the gates, so she compares them on four weighted criteria, and the specialist's answers on consent and authentication are the more specific. The first paid job is the move, quoted as a separate item and run with the suppression file loaded first. The specialist then builds a welcome flow in the new account with a random holdout of 50 of the 500 new sign-ups in the period. After eight weeks 38 of the 450 who received the flow have placed a first order, against 3 of the 50 held out, a gap small enough that chance could produce it at that volume, so the report says the result is not yet readable. The holdout keeps running, and no lift is claimed or budgeted until it can be read.
The fundamentals section places this guide beside the platform, migration and consent guides it relies on.
Common mistakes
- Hiring before the platform and the domain are settled. A provider who picks the platform and holds the domain also holds the list's history; decide both first and grant access.
- Judging a proposal on creative alone. A strong newsletter sent to addresses nobody can account for, through a domain that is not authenticated, can underperform however well it is written.
- Letting the provider set the attribution rule. It shapes the figure, so it is agreed with whoever owns analytics and held between reports.
- Taking the flow's own report as the result. Until a holdout has been read, no one can say what the flow caused, the provider no more than the business.
- Treating suppression as clean-up. Deleting opted-out contacts to tidy a list lets the next import bring them back; the file is carried across, not discarded.
- Writing the handover list on the last day. Start the inventory at the outset, so that an ending begins from a written list instead of a hunt through the account.
