← All playbooks·Buying and negotiation

Playbooks · Buying and negotiation

Trim the supply path to cut redundant fees

This job produces an authorized, trimmed SPO path list per publisher domain, keeping only the SSP paths that clear inventory at the lowest verified take rate. It feeds the decision on which SSPs stay in the deal ID map and which get pulled from the buy.

What you need first

  • Path-level spend, win rate and CPM by SSP for the trailing 30-60 days, from DSP delivery logs
  • SSP fee schedule or take rate card per path, from SSP rate cards or the TAC disclosure if the SSP supports it
  • ads.txt and sellers.json records for each publisher domain, to identify duplicate or reseller paths to the same inventory
  • Current deal ID / PMP map showing which SSPs carry which deals, from the trading desk's deal sheet
  • IVT/fraud rate by path, from the verification vendor, if available

The procedure

  1. Pull path-level spend, win rate and CPM from DSP logs for the trailing period, producing a ranked path list per publisher domain
  2. Cross-check ads.txt and sellers.json for each publisher domain to flag paths reselling the same seller ID through a different SSP, producing a duplicate-path list
  3. Calculate net-to-publisher rate per path from the SSP fee schedule (100% minus SSP take minus DSP fee), producing a TAC ranking per path
  4. Rank paths per publisher on net-to-publisher rate, win rate and IVT rate together, producing a keep/cut list
  5. Cut the lowest-ranked and duplicate paths, producing an updated SPO allowlist
  6. Rebuild deal IDs and PMPs against the retained paths only, producing an updated buy-side path map
  7. Hold the cut for two weeks and track win rate and CPM against the pre-cut baseline, producing a variance report that confirms fill held

Worked through with numbers

A German news publisher clears through 4 paths totaling €100,000/month spend: Path A (direct, 15% take, €40k), Path B (reseller of A's inventory via a second SSP, 25% take, €12k), Path C (18% take, €30k), Path D (12% take, €18k). Blended take rate = (40,000×0.15 + 12,000×0.25 + 30,000×0.18 + 18,000×0.12) / 100,000 = 16,560/100,000 = 16.56%. sellers.json shows Path B carries the same seller ID as Path A, so it is a reseller markup with no unique fill. Cutting Path B and rerouting its €12k to Path A gives new spend of A=€52k, C=€30k, D=€18k: (52,000×0.15 + 30,000×0.18 + 18,000×0.12) / 100,000 = 15,360/100,000 = 15.36%. That is a 1.2 point drop in blended take rate on €100k monthly spend, €1,200/month reclaimed as working media. Read it as: the saving only counts once win rate on Path A holds within the post-cut monitoring window, otherwise the volume just moved, not the cost.

Where it goes wrong

  • Don't cut a path on take rate alone without checking sellers.json for unique fill on that publisher, since some high-fee paths carry inventory no other path reaches; confirm duplicate coverage before cutting
  • Don't rank paths on SSP fee in isolation; rank on net-to-publisher rate combined with win rate and viewability, since a cheap path with low fill is not actually cheaper
  • Don't cut every resold path in one pass; stage the cuts and watch win rate and CPM variance for a week or two so a fill drop is traceable to a specific cut
  • Don't treat the fee schedule as fixed; refresh SSP take rates quarterly, since SSPs revise them without always flagging the change to trading desks

How to know it is right

Confirm the blended take rate on the post-cut spend report is lower than the pre-cut baseline while win rate stayed within about 2 points of it, before the new path map goes to the buying team.

Terms used

← All playbooks Calculated in Media buying software →