Frequency Capping per Account for Long Industrial Cycles

A cookie-based cap of three impressions a day is not a cap of three impressions a day. In the log-level data I pull from campaigns in this sector, the same browser reappears under a fresh identifier roughly every ten days on Chrome and inside a single day on Safari, so a three-a-day setting delivers closer to nine or twelve a week to one person and reports that person as three or four. Every reach figure built on top of that number is inflated by the same factor. How frequency should be counted here: The unit that matters in industrial and energy buying is not a browser and not a person. It is an account. A pump specification decision involves a plant engineer, a maintenance manager, a procurement lead and someone in finance, and the campaign is working when all four have seen it enough times across a purchase cycle that runs nine to eighteen months. So I cap per company per week, resolve the company from the domain or from IP-to-organisation data rather than from a cookie, and set the ceiling in the region of a dozen impressions per account per week across a flight that is measured in quarters. What the plan contains: A named account list, sized in the hundreds rather than modelled into the millions, and a reach curve built against that list showing where additional spend stops adding accounts and starts adding repetition. That inflection is the number the plan turns on, and it is usually reached far earlier than the media plan assumed. Below it, budget is buying reach. Above it, budget is buying frequency, which is a defensible choice but should be a deliberate one. Cross-platform overlap: Frequency cannot be deduplicated across two demand-side platforms. Nobody can do it, including the vendors who sell it, because neither platform sees the other's impression log. My answer is structural rather than technical: one seat carries the display and video budget, and where Google-side inventory is involved I dedupe reach in Ads Data Hub rather than accepting two separate reach reports that add up to more people than exist in the target market. If you are committed to running two platforms in parallel, I will plan a frequency ceiling for each and tell you plainly that the combined figure is an estimate. Boundaries: I do not model incremental sales impact from frequency. Reach and frequency are inputs; anyone selling you a curve that converts them into revenue for a capital purchase with an eighteen-month cycle is modelling, not measuring, and you should treat it accordingly. Creative rotation and message sequencing are adjacent and separate. I will tell you how many distinct executions your frequency ceiling requires so that the twelfth impression is not the first one again, but I do not produce them. Wrong fit: Consumer campaigns and anything with a short purchase cycle. The whole argument here rests on a long, multi-person decision, and against a two-week consideration window the standard platform cap is close enough and costs you nothing. Also wrong if you have no account list and no intention of building one. Without it this becomes generic reach planning against a modelled audience, which is a fair thing to buy and not what I do. What you receive: A frequency and reach plan with the caps expressed at account level, the platform settings written out per line item so they can be implemented without interpretation, the reach curve with its saturation point marked, and a two-page note on how far the platform's own unique-reach reporting will differ from what is actually happening.