The conversation was recorded over the course of an afternoon at a shared meeting room in central Manchester, in late May. The head of programmatic — she asked to be called Ivy, and to have the bank remain anonymous, and both requests have been honoured — has been in role for four years and had spent the previous six months running what she described, unprompted, as "the most demanding piece of work of my career". The audit had, by the time we spoke, produced a set of decisions the bank had implemented but had not, at least publicly, discussed. The conversation is her account of the process and the outcome.
The audit
Kaspar Ihre: Take me back to the beginning. What made you decide to audit every SSP?
Ivy: A specific piece of finance-team analysis, in the middle of last year. My CFO, who I get on well with, sent me a rather understated note asking whether I could explain a specific pattern in our programmatic delivery data. He had noticed that, on a small set of publishers we deliver against, we were buying through five or six different SSPs simultaneously. Some of the impressions were clearing at £0.80 CPM; some, on what appeared to be broadly similar inventory, at £3.40 CPM. He wanted to understand why we were paying so much more for what looked like the same impression.
The honest answer, which took me two weeks to work out and another month to confirm, was that we had never systematically deduplicated our supply paths. We were, in effect, bidding against ourselves through different SSPs on the same inventory, and the price we were paying was the price of the winning path, not the lowest path we could have won at.
Once I saw the pattern on that specific publisher, I ran the same analysis across our whole delivery footprint. The pattern was the same everywhere. On our best estimates, we were paying somewhere between 14% and 22% more than we needed to, in aggregate, on our open-exchange delivery, entirely because of unmanaged path duplication.
The process
Kaspar: How did you actually run the audit?
Ivy: Six months. Two people on my team full-time. A DSP partner who was, to their credit, more helpful than I had expected. A specialist SPO consultant we brought in for four weeks to help with the technical analysis.
The mechanics were, at one level, straightforward. We pulled six months of delivery logs at the impression level, joined the impressions across paths using the publisher-side deal IDs and slot IDs, and identified where the same impression had been available to us through multiple SSPs. For each duplicate path, we calculated the price we paid, the price we would have paid on the cheapest available path, and the aggregate cost of the difference.
The aggregate cost, on our six-month sample, was £1.4m — annualised, roughly £2.8m. Against a total open-exchange spend of about £14m annually, this was an 20% waste rate. Not universal — some paths were legitimately more expensive because they carried better-quality inventory on a like-for-like basis — but on the specific paths where we could identify no quality differential, the difference was pure operational waste.
The remediation work was where it got politically difficult. We had, in that six-month window, spent meaningful money with fourteen different SSPs. Some of those SSPs, on our post-audit analysis, were producing efficient paths against unique publisher supply. Some were producing efficient paths against publisher supply that other SSPs also carried. Some were producing consistently inefficient paths where we would have been better off buying elsewhere. And a handful were producing consistent evidence of either MFA-adjacent inventory or duplicative billing patterns that we could not, in good conscience, continue to spend against.
"The technical work was straightforward. Deciding what to say to the SSPs whose paths were consistently worse was, on any honest read, the harder part. Some of those relationships had been in place for four or five years. Some of the SSP account managers were people I would have described as friends. The audit forced conversations I would rather not have had."
What she cut
Kaspar: Can you describe the shape of the outcome, without naming names?
Ivy: We reduced our active SSP count from fourteen to seven. Of the seven cut, three were dropped because their paths were consistently duplicative of other SSPs and consistently more expensive on the duplicate paths. Two were dropped because their paths included a substantial share of inventory that our subsequent quality review flagged as MFA-adjacent. Two were dropped for a mix of reasons — reporting opacity, misalignment between the SSP's stated inventory pool and what actually delivered, unusual patterns in the billing that our finance team wanted separated from the wider account.
Of the seven we retained, three were kept without meaningful change. Two were kept with revised commercial terms — lower CPM ceilings, or explicit path-preference agreements that stopped them from bidding on inventory where a different SSP had a cheaper legitimate path to the same impression. Two were kept on trial, with a six-month review scheduled to determine whether they had improved on the metrics we had flagged as problematic.
The aggregate spend saved, at annualised rates, was approximately £2.1m against a £14m base — a 15% reduction in cost for approximately the same delivery outcomes. The MMM-attributed downstream performance was, if anything, marginally better than before, because the retained paths were, on aggregate, more efficient.
The political cost
Kaspar: What was harder than you expected?
Ivy: The SSP relationships. The five SSPs we dropped included two whose account teams I had known for years. The conversations with them were, in some cases, unpleasant. One of the two was, in his rendering, personally hurt by the decision — he had, over four years, been a good partner, and my decision to drop his SSP was, from his perspective, a betrayal of that relationship. I understood his position. I still made the decision.
The other unexpected thing was the internal blowback. Our head of marketing, when I described the number of SSPs we were dropping, was concerned that the reduced supply-side diversity would hurt reach. She was, in the strict sense, correct — we did lose some reach, particularly on specific mid-tier publishers where the dropped SSPs had been our only cost-effective path. On aggregate the reach reduction was small (perhaps 3-4% on total impressions), and it was more than offset by improved efficiency on the retained paths. But defending the reach reduction, in the abstract, was harder than defending the cost saving.
Kaspar: Last question. Would you recommend other buyers run the same audit?
Ivy: Yes, but with a warning. The audit will, on my strong prior, uncover substantial waste that you did not know was there. It will also, in almost every case, require you to have politically difficult conversations with people you have worked with for years. If you are not, at the moment you consider the audit, willing to have those conversations, you should not start the audit. Starting the audit and then declining to act on the findings is worse than not doing the audit at all — you will have identified specific waste that you are then knowingly tolerating, which is a substantially worse governance position than not knowing.
My strong recommendation is to start the audit only when you have, in advance, established the political air cover to act on whatever it reveals. Get your CFO's explicit endorsement. Get your CMO's explicit acceptance that some SSP relationships may end. Then start. Otherwise, don't.