How to build an ad audience sync

What matters here

• Never upload an exported list. It is stale the moment it is downloaded and it cannot be suppressed.
• Suppress customers, open opportunities, recent churns and anybody who opted out, before the audience leaves your systems.
• Hash identifiers locally. Raw email should never reach an ad platform, and hashing at the boundary is the cheapest privacy control there is.
• Refresh at least daily. An audience refreshed monthly is advertising to people whose situation changed three weeks ago.
• Track match rate per platform. A low match rate means you are paying for reach you are not getting.

Who this is for

You run marketing operations or demand generation. Audiences are exported to CSV, uploaded by hand, and nobody is quite sure whether current customers are in them.

How it works in practice

The sequence, from a segment being defined to an advert being served.

1. 1

   The audience is defined as CRM state, not a list

   A saved query evaluated at run time, so somebody who converted this morning leaves it today.

2. 2

   Suppression runs before anything leaves

   Existing customers, open opportunities, recent churns, opt-outs and competitor domains.

3. 3

   Identifiers are hashed locally

   Normalised then hashed before transmission, so raw email never reaches the platform.

4. 4

   The audience is synced as a delta

   Adds and removes, rather than replacing the whole list and resetting the platform learning.

5. 5

   Match rate is recorded per platform

   Because you are paying for reach and a poor match means you are not getting it.

6. 6

   Suppression is re-evaluated on every refresh

   A customer who signed yesterday leaves the acquisition audience today, not next month.

Paste these into Claude Code or Codex with the Tray Headless plugin installed. Each stage runs on its own. The systems named in them are the worked example rather than a requirement, and every prompt says so.

Once per project, run /tray-workflows:set-workspace to pick the workspace these build in. Point it at a sandbox first.

1. 1

   Start here: define suppression first

   Suppression is the reason to build this instead of exporting a CSV.

   Headless skills build-workflow

   Copy

   Use build-workflow. The systems in play are Salesforce and the ad
   platforms we run, or whatever we run in those seats.
   
   Before defining any audience, define the suppression list:
   
     Anybody at an account with an active subscription
     Anybody at an account with an open opportunity above a stage
     Anybody who churned in the last year, which is a win-back audience
     instead of an acquisition one
     Anybody who opted out of marketing
     Employees, competitors and partners by domain
   
   Tell me where each of those lives. Suppression is the reason to build
   this instead of exporting a spreadsheet, so it gets designed first.

2. 2

   Build audiences from live state, never a list

   An exported list is stale before it finishes downloading.

   Headless skills build-workflow

   Copy

   Use build-workflow. Define each audience as a saved query against the CRM
   and the warehouse, evaluated at run time.
   
   Never accept a static uploaded list. Somebody who converted this morning
   should leave the acquisition audience today, and a list cannot do that.
   
   Apply suppression after building the audience and before anything leaves
   our systems. Do not rely on the platform to exclude a list you also
   uploaded, because the exclusion rules differ per platform and the failure
   mode is invisible.
   
   Log the suppression counts per run. If suppression removes a large share
   of an audience, the audience definition is wrong rather than suppression
   working hard.

3. 3

   Hash before it leaves

   The cheapest privacy control available.

   Copy

   Normalise then hash every identifier before transmission: lowercase and
   trim the email, strip plus-addressing, normalise phone numbers to E.164,
   then hash with the algorithm each platform specifies.
   
   Raw email addresses should never reach an ad platform. Hashing at the
   boundary costs nothing and removes an entire category of problem from
   your data processing agreements.
   
   Send the minimum each platform needs for matching. Do not send company,
   job title or anything else because a field happened to be available.

4. 4

   Sync deltas and keep the audience alive

   A full replace resets platform learning every night.

   Headless skills tray-gotchas

   Copy

   Use tray-gotchas, then sync as a delta: adds and removes against what the
   platform already holds, rather than replacing the audience wholesale.
   
   A full replace resets the platform's optimisation and it shows in
   performance within days.
   
   Refresh at least daily. Handle these:
   
     An audience that falls below the platform minimum size, which silently
     stops serving rather than erroring
     A platform rejecting a batch, which must not leave the audience half
     updated
     Removals failing, which is the dangerous direction because it means a
     customer is still being targeted
     An audience that has not refreshed within its window, which should
     alert rather than quietly go stale

5. 5

   Measure match rate and suppression

   Because you are paying for reach you may not be getting.

   Copy

   Report per audience and per platform: size, match rate, suppression
   counts by reason, and how long since the last successful refresh.
   
   Match rate is the number that matters commercially. A 40% match means
   more than half the audience is not reachable and the spend is
   concentrating on whoever did match, which is rarely the segment you
   designed.
   
   Alert on: match rate dropping sharply, an audience below the minimum
   serving size, a refresh that has not completed, and any run where
   suppression removed nothing, because that usually means the suppression
   join silently broke.

6. 6

   Check it end to end, then hand the definitions over

   Because audiences change with every campaign.

   Copy

   Run the per-step schema checks and the whole-workflow audit before this
   touches production. Run once in preview, showing the audience and the
   suppression counts without sending, and have somebody read it.
   
   Then open the same workflow in Tray Build so marketing operations can
   define new audiences and amend suppression rules in the visual canvas.

What it connects to

The CRM decides who, the warehouse adds context, and the platforms receive hashes.

Named systems are the ones most teams run, not the only ones that work. Each is an authentication in your Tray workspace, referenced by name, so the workflow never holds a credential. Where we have a connector page, the name links to it.

Questions people ask

Further reading

Background on the same subject, for the case rather than the build.

• Drive down CPC and target the best leads automatically

Related guides