How to choose a direct mail platform: 9 questions before you buy

A direct mail automation platform has two jobs: get the right piece into the mail when a useful sales signal appears, and tell you whether sending it changed buyer behavior. That's it. The feature list can run for pages, but none of it rescues poor production or reporting you don't trust.

I start vendor reviews with nine questions. For every vendor, I want the same four things: a live workflow, physical samples, an itemized quote, and an actual campaign report. A polished demo isn't evidence.

1. Which mail formats does the platform support?

Start with what you'll really mail in the next 12 months. Not the campaign somebody might run someday. Postcards work well for plenty of event follow-up. A letter has room for a detailed offer; a handwritten note can make an executive touch feel more personal. Dimensional pieces and gifts can get noticed across a short list of valuable accounts, but now you're dealing with packaging, inventory, and shipping too.

Find out who produces each format. The vendor may do it, or the job may quietly move to a partner. Then get specific. For postcards, check sizes, paper weights, coatings, and full-bleed printing. For letters, check page counts, envelope styles, inserts, and first-class versus marketing mail. "Handwritten" can mean a person, a pen plotter, or a handwriting font, so ask. Gift programs raise a different set of questions around inventory, custom kits, perishables, and recipient choice. Domestic and international coverage belongs in this conversation as well.

The word "letters" on a feature grid tells you almost nothing. It might mean one page, one standard window envelope, and no variable insert. Get the templates. Get finished samples. If the medium itself is still undecided, compare postcards, letters, and dimensional mail.

Score the formats you'll use. Ignore the rest.

2. Does the direct mail automation platform connect to your CRM?

Importing a list of names doesn't make something a useful CRM integration. You need the events, fields, permissions, and write-back that keep the campaign out of spreadsheets.

Put one real workflow in front of the vendor. Here's a good test: an account moves to opportunity stage two, a letter goes to the economic buyer, the contact timeline gets the event, and the send stops if the opportunity closes before production. Make them run that sequence in the demo.

This part trips people up. "Salesforce integration" or "HubSpot integration" sounds complete until you ask which objects the platform can read and where delivery or engagement events get written. Check whether workflow actions are native or whether Zapier, webhooks, or CSV uploads are doing the work behind the curtain.

Break it on purpose. Feed it a rejected record, cause a temporary error, and try a duplicate. You want visible failures, sensible retries, duplicate prevention, and a record of rule changes. See how triggered direct mail works with CRM events.

3. How deep does personalization go?

Changing a name and company is mail merge. Useful personalization can also swap the message, image, sender, landing page, or even the format based on CRM data.

Ask the vendor to show every element that can vary inside one campaign. In practice, you might need the account owner's signature, an industry-specific proof point, and a unique page on the same piece. Look closely at conditional logic, defaults, image rules, and the preview someone approves before the job runs.

Missing data is where a handsome template turns ugly. What happens when there's no preferred name, industry, or owner photo? The record should be flagged or an approved fallback should appear. A blank field is bad. A printed {{first_name}} token is worse.

Approvals matter too, especially once sellers get access. Brand may lock most of the template and leave only a short note editable. The system should keep the template version and the exact content each recipient was sent.

4. Can you track each piece from production to response?

Campaign totals aren't enough for B2B work. Every piece needs a stable recipient record and a readable history: submitted, approved, produced, handed to the carrier, delivered when carrier data allows it, returned, QR scan, short-link visit, and reply.

Then ask how unique QR codes and URLs are made. One shared code gives you aggregate traffic, but it can't tie a visit to the intended recipient. Recipient-level codes are much better for sales follow-up. They also create obligations around access and data retention, which shouldn't be waved away.

Be skeptical of delivery language. USPS data may tell you a piece reached the final delivery operation; it doesn't tell you a named executive opened the envelope. I also want to know how the vendor filters security scanners, duplicate scans, internal tests, and bots.

Finally, export the raw events. Timestamps and recipient IDs included.

5. Does measurement show lift or only activity?

QR scans, page visits, and claimed gifts are useful operating signals. They're not proof of incremental pipeline. The mailed group may have been more likely to buy from the start because the team chose active opportunities or high-fit accounts.

I wouldn't buy the measurement story without seeing holdout testing. The platform should reserve a random share of eligible people or accounts, treat them the same in every other respect, and withhold the mail. You can then compare meetings booked, opportunities created, pipeline value, or closed revenue inside a defined window. If several contacts work at the same company, randomize at the account level. Otherwise the test can contaminate itself.

Ask who was eligible, how assignment worked, what got excluded, which attribution window was used, and whether the sample was large enough. "42 influenced opportunities" may look impressive on a slide. Without a control group or decision rule, it's a vanity number. Small campaigns often need leading outcomes and several pooled tests.

Sincerely is one example of a platform that combines recipient-level QR codes and short URLs with holdout-based campaign lift measurement. Whoever you're evaluating, ask for both the activity log and the holdout report.

Use this guide to measuring direct mail ROI before you write the vendor requirements.

6. What are the minimums and the full pricing model?

Get itemized pricing at three campaign sizes you might actually send, say 100, 1,000, and 10,000 pieces. Every charge belongs on the page: subscription, seats, implementation, creative services, printing, handwriting, envelopes, inserts, postage, pick and pack, storage, kitting, address checks, international handling, returned mail, and API usage.

Most pricing mixes a monthly or annual software fee with per-piece production and volume tiers. Gifts may bring storage and fulfillment fees. Postage moves with class, size, weight, presort eligibility, and destination. Small changes add up quickly.

Ask in writing:

• Is there a monthly send minimum or annual commitment?
• Do unused credits expire?
• Which services require prepayment, and can prices change during the contract?
• What happens when a campaign is canceled after submission?
• Are test pieces, replacements, and returns billed?

Compare the total at your expected mix, including internal labor. The lowest unit price isn't cheap if an operations manager loses six hours cleaning every upload.

7. How does the vendor control fulfillment quality?

The physical piece is the product. Inspect it before you sign anything.

I use a small test file containing long names, accented characters, apartment numbers, missing fields, several template branches, and one deliberately invalid record. What arrives matters, obviously, but so does what the system refuses to produce.

Ask where every format is made and how jobs are checked there. Barcode matching between letters and envelopes, camera inspection, color calibration, weight checks, seed pieces, and written reprint rules are all worth examining. With handwritten notes, look at repeated letters, spacing, and smudges. Does the result match the production method you were sold?

Production time belongs in the service levels. Get committed timing by format, holiday cutoffs, and the procedure for lost inventory. If jobs move across production sites, confirm that the materials and status data remain consistent.

And don't let the premium sample kit decide this. Have the vendor run your design through its normal workflow and mail the output to several office and residential addresses.

8. How does address handling prevent waste and risk?

Bad addresses waste money and make the sender look careless. Before charging for production, the platform should standardize addresses, spot missing secondary units and obvious duplicates, and run deliverability checks. Ask which countries are covered and exactly when validation happens: entry, upload, CRM sync, or final submission.

Next, try to suppress a record. Can you block a contact, account, domain, household, region, or individual address? Can a CRM opt-out stop a queued send before the production cutoff? You need an audit trail, but suppressed data shouldn't be visible to every user.

Returns need somewhere to go. The vendor may capture USPS return codes, scan the piece, update the CRM, or discard it. Find out which. At minimum, a returned piece should make the next campaign smarter.

Home addresses and gift preferences are personal data, not ordinary campaign fields. Review access, encryption, retention, deletion, subprocessors, and incident procedures. Who can download an address list? Is the export logged?

9. Does the API support automation and agent access safely?

You need an API when native integrations can't express the workflow, or when product, data, and AI agents will initiate sends. Get the documentation before buying. I look for endpoints covering recipients, validation, previews, submission, cancellation, templates, status, and tracking.

The less glamorous details matter most: idempotency keys, signed webhooks with retries, a sandbox, rate limits, stable recipient IDs, and errors a human can understand. Ask how old versions are retired.

An agent shouldn't inherit the power of a general API key. A sales agent might draft a note and request approval, but it shouldn't be able to send 20,000 gifts or export every home address. Look for scoped credentials, spend and recipient limits, template restrictions, approval gates, and complete logs of tool calls and human approvals.

Give an engineer a practical test. Create a sandbox send, repeat it, submit an invalid address, receive a webhook, and reconcile the final event back to the CRM.

Build a scorecard before the demos

Weight the categories for your program, not an imaginary average buyer. A gifting team will probably care more about fulfillment. A company sending triggered postcards may put API reliability and unit economics near the top.

Use a 1 to 5 score. One means the requirement is missing; five means the vendor proved it in your workflow. Roadmap promises don't get fives.

Multiply each score by its weight and put the open risks beside it. Marketing, sales operations, security, and finance should review the same sheet. The total narrows the field, but evidence and risk make the decision.

Make the final test physical

Pick two finalists and run the same small campaign through each. Keep the recipients, creative, postage, and timing comparable. Track setup time, rejected records, production speed, arrival quality, event accuracy, and the support response when something goes wrong.

Keep a finished piece from each vendor beside the contract and report. Then buy the workflow you verified, the quality your recipients will actually see, and the measurement finance will accept.