Skip to content
Delivr.ai
Back to Resources
Technical7 min read

A Migration Path for RampID Dependents

Practical and unglamorous: what a brand built on RampID actually does next. How to inventory the dependency, re-anchor on an identifier you hold, run two resolutions in parallel, and cut over without losing destinations.

By Delivr.ai

RampIDmigrationidentity resolutionhashed emailactivationimplementation

If your addressability runs on RampID and your team has spent the last fortnight being asked whether that is still a good idea, this is the document you actually need. Not a position on the Publicis acquisition — that is elsewhere — but the engineering sequence for reducing a single-vendor identity dependency without breaking the activation that depends on it.

Two things to establish before the steps, because they change the shape of the work.

Leaving the graph and leaving the destinations are different decisions. RampID is doing two jobs in most stacks: it resolves identity, and it reaches a few hundred downstream destinations. Teams treat these as one commitment and conclude migration is impossible. They are separable, and separating them is most of the value here even if you never cut over.

The identifier you keep is the whole of the outcome. A migration that ends with you holding a different vendor’s proprietary ID has not reduced your dependency. It has moved it, and it will have to be done again.

The five steps

In order. Step three is the one teams skip and the one that makes the rest defensible.

Five steps off a single-vendor identity dependencyFive sequential steps. One, inventory every destination, segment and measurement join that resolves through the incumbent today. Two, re-anchor on the hashed email, an identifier you hold, resolved from your own pixel and CRM. Three, run the second resolution in parallel for a full cycle and compare before deciding. Four, rebuild the destinations carrying spend first and keep delivering into RampID where it is still the best path. Five, cut over, retiring the dependency rather than the identifier.01Inventory what the dependency actually carriesEvery destination, segment and measurement join that resolves through it today. Mostteams find fewer live paths than they expect, and one they had forgotten.02Re-anchor on an identifier you holdThe hashed email is the portable anchor. Resolve to it from your own pixel and your ownCRM, so the graph is not the thing holding your addressability.03Run the second resolution in parallelKeep the incumbent live. Resolve the same traffic twice for a full cycle and compare, sothe cutover is a measured decision rather than a bet.04Rebuild the destinations that earn itNot every destination matters to you. Rebuild the ones carrying spend first, and keepdelivering into RampID wherever it is still the best path.05Cut over, and keep the anchorRetire the dependency, not the identifier. If the owner of anything else in your stackchanges, you repeat step four and nothing else.
The order matters more than the speed. Step three is the one teams skip, and the one that makes the rest defensible.

Step 1: Inventory what the dependency actually carries

Before anything else, write down every live path that resolves through RampID today. Four categories, and most teams have never had them in one list:

  • Activation destinations. Which DSPs, social platforms, CTV partners and publisher direct deals you currently reach through onboarding, and what monthly spend flows through each.
  • Segments. Which audiences are defined in the platform rather than defined in your own systems and pushed to it. These are the expensive ones: the definition lives there, not with you.
  • Measurement joins. Any attribution, incrementality or reach-and-frequency work that joins on RampID. These break silently and late.
  • Internal systems. Anywhere RampID has leaked into your own warehouse as a join key, which is more places than the data team expects.

The output is a list with a spend number next to each row. Nearly every team doing this honestly finds two surprises: a long tail of destinations carrying almost nothing, and one integration nobody remembered that carries a lot.

Step 2: Re-anchor on an identifier you hold

A proprietary graph ID is a rented anchor. You cannot take it with you, you cannot resolve against it independently, and its value is contingent on the relationship continuing. That is fine until the question of who owns the vendor becomes a question.

The portable anchor is the hashed email. It is not a product anyone sells you — it is a hash of a string you already have, which is exactly why it travels. Any vendor can resolve against it, you can store it, and switching costs collapse to re-pointing a resolution call.

Practically, re-anchoring means two things:

  • Your own pixel on your own property, resolving anonymous traffic to a hashed email you keep. First-party, your domain, your storage.
  • Your CRM and marketing records keyed on the hashed email alongside whatever vendor IDs exist, so your warehouse is not dependent on a vendor join to reconcile a person with themselves.

Delivr.ai accepts plain-text email, SHA-256 and MD5 on the resolve endpoint, and hashes plain text in-pipeline, so you are not forced into a single hashing convention to start. Downstream of the hashed email you get the person profile, the company, the household and the activation identifiers, including UID2 generated through The Trade Desk framework. The identity bridge covers how those relate.

The governing rule for this step: at the end of it, the most valuable identifier in your stack should be one no vendor can withhold.

Step 3: Run the second resolution in parallel

Do not cut over. Run both.

Point a second resolution at the same traffic for a full cycle — a month at minimum, long enough to cover your actual buying rhythm — and compare on your own data rather than on anyone’s published figures:

  • Overlap. Of the people each resolves, how many does the other also resolve? Pure overlap is less interesting than the two exclusive sets.
  • The exclusive sets. Who does the new path find that the incumbent does not, and the reverse. This is where the decision actually lives, and it is specific to your traffic mix.
  • Agreement on identity. Where both resolve the same visitor, do they agree on who it is? Disagreement here is the finding that matters most and the one nobody looks for.
  • Destination fill. For the destinations you care about, what proportion of the audience actually lands on each path.
  • Latency and freshness. Real-time resolution at the pixel, and how quickly a new signal becomes an addressable one.

A month of parallel running costs you a month. A cutover decided on vendor-published match rates costs you a quarter of unexplained performance change and no way to attribute it. We will run a resolution audit against your own traffic for exactly this reason; the comparison is only useful if the data is yours.

One warning specific to right now. Resolution rates across the whole industry are moving because of browser-level restrictions, not only because of vendor differences. Apple’s latest Safari release restricts identity resolution domains. If your parallel test spans that change, you will see movement in both paths, and attributing all of it to the vendor you are evaluating will mislead you. Segment the comparison by browser.

Step 4: Rebuild the destinations that earn it

Go back to the spend-ranked inventory from step one. Rebuild in that order, not alphabetically and not all at once.

In practice most teams find that a small number of destinations carry the overwhelming majority of spend, and those are usually the ones with the most direct integration paths. The long tail is long and cheap to leave in place for now.

Keep delivering into RampID where it is still the best path. This is the part that surprises people, and it is the honest engineering answer. A hashed email can be onboarded, and a destination that genuinely only reaches you through LiveRamp is a destination you should keep reaching. What you are changing is which system is the source of truth for identity, not which pipes exist. Delivr.ai delivers into Delivr DSP, The Trade Desk, HubSpot, Salesforce, S3, email and Slack, and into onboarding paths beyond those.

Sequence each destination as: build in parallel, verify audience sizes against the old path, run both for one flight, then retire the old path for that destination only. One destination at a time means one thing to debug at a time.

Step 5: Cut over, and keep the anchor

The cutover itself should be anticlimactic if the first four steps were done. What matters is what you are left holding.

Retire the dependency, not the identifier. At the end you should own: the hashed email for every person you have resolved, your segment definitions in your own systems rather than in a vendor UI, your measurement joined on an identifier you control, and destination integrations that can be re-pointed rather than rebuilt.

Tested properly, that structure means the next time an identity vendor anywhere in your stack is acquired, you repeat step four and nothing else. That is the actual deliverable of this exercise, and it is worth more than the specific vendor decision that prompted it.

What this costs and how long it takes

An honest estimate, with the caveat that it varies more by how much of your identity logic lives in the vendor’s UI than by your size.

  • Steps 1 and 2 are a few weeks of analyst and data-engineering time. Most of it is the inventory, and the pixel deployment is hours.
  • Step 3 is one buying cycle running two resolutions, so a month of wall-clock and relatively little effort once it is running.
  • Step 4 is the bulk of the work and scales with the number of destinations carrying real spend, not with the total.
  • Step 5 is a flag flip and a contract conversation.

The thing that makes this expensive is doing it under deadline pressure after a renewal has already lapsed. The thing that makes it cheap is step two, which has standalone value whether or not you ever migrate.

If you want the parallel resolution in step three run against your own traffic before you commit to anything, book a call. The comparison is yours regardless of what you decide.

See Delivr.ai in action

Book a call and see how Delivr.ai turns your anonymous traffic into actionable pipeline.