--- title: "How to measure B2B awareness advertising | AdBuyMCP" description: "See what your weekly report counts on each channel, what to watch in your own numbers, and when a geo-lift test can show whether the ads caused a change." url: "https://adbuymcp.com/measurement" --- [Home](https://adbuymcp.com/) / Measurement Measurement # See what each channel delivered against your plan, every week. Your weekly report counts each channel in its own unit, from plays on a digital billboard to podcast ad deliveries, and sets each count beside your plan. Your own numbers show whether more people are looking for you, and when they're big enough, a geo-lift test can show whether the ads caused it. [Get my plan and price](https://adbuymcp.com/start) [Book a 30-minute call with the founder](https://adbuymcp.com/contact#book) Free, with no card. We read your homepage, draft who buys from you and show your monthly price before you join the waitlist. Nothing is bought until you've approved the plan and the ads. The weekly report ## Each channel is measured in its own way, and your report shows what each one counted. Some counts are logged, like a screen's plays, and some are modelled, like Route's audience, so the report says which is which and who produced each one. What the weekly report counts for each of the four channels, who counts it, and what the count cannot show. | Channel | What it counts | Who counts it | What the count can’t show | | --- | --- | --- | --- | | [Out-of-home](https://adbuymcp.com/channels/dooh) | Proof of play, each digital screen's log of when your ad ran, and the reach, impacts and frequency Route models for every site and time you bought. | The screen's owner logs the plays. Route, Great Britain's out-of-home audience measure, models who passed. | Whether any one person saw it. Route models adults aged 15 and over and has no B2B decision-maker audience. | | [Streaming TV](https://adbuymcp.com/channels/ctv-streaming) | Impressions and completed views, with reach and frequency by household within each seller. | Each streaming service or ad platform, for its own inventory only. | Whether a person in your market watched. Counts stop at the household or device, and no seller sees another seller's viewers. | | [Podcasts](https://adbuymcp.com/channels/podcast-audio) | Downloads and ad deliveries, counted from server logs to the IAB Tech Lab standard. | The podcast host or network that serves the ad. | Whether the ad was played or heard, or by whom. A download is not a confirmed play. | | [ChatGPT ads](https://adbuymcp.com/channels/chatgpt-ads) | Impressions, clicks, spend, click-through rate, cost per click, CPM and conversions. | OpenAI, in the ad account you own. | How many different people saw it, or how often. OpenAI reports no reach or frequency. | Sources, from desk research checked 20 October 2026: [Route, Audience measurement](https://route.org.uk/audience.php); [IAB Tech Lab, Podcast Measurement Technical Guidelines](https://iabtechlab.com/standards/podcast-measurement-guidelines/); [OpenAI Ads developer documentation, Reporting](https://developers.openai.com/ads/reporting). Frequency ## Each week, your plan aims a set number of ads at your market. Frequency is planned: each week we buy enough ads aimed at your market for it to see you that many times on average. Some of every channel's audience sits outside your market, and the weekly report shows what each channel delivered against the plan. For example, a market of 12,000 people at 5 times a week means the plan buys 60,000 impressions a week aimed at that market, split across the four channels. The report sets each channel's count beside that channel's part of the plan, so you can see which channels delivered what was planned. ### No UK service can tell you how often one person saw you across all four channels. Each channel's count covers that channel alone, and nothing joins the four counts up, so someone who passes your billboard and hears your podcast ad is counted once by each. Adding the channels' frequencies together would overstate how often anyone saw you, so the report keeps each channel's count apart. Your own numbers ## Watch your own numbers for signs your market knows your name. These numbers live in your own tools and move for many reasons, from a launch to a press mention. Read them beside your weekly report as signs to watch rather than proof that the ads worked. ### Check how often people search for your name. Google Search Console shows the searches your site appears for, including your company and product names. Compare the campaign weeks with the weeks before them. ### Check how many visits arrive direct. Visits from people who type your address or use a bookmark show up as direct traffic in your analytics. It's a rough sign of how many people remember where to find you. ### Ask every new lead where they first heard of you. Add the question to your demo form and your first sales call, and keep a tally. Mentions of a billboard, a streaming TV ad, a podcast or ChatGPT are the ones to count, week by week. Geo-lift tests ## A geo-lift test can show whether the ads caused a change. It runs the campaign in some regions, holds comparable regions back and compares a number you already track between the two. AdBuyMCP's geo-lift engine checks whether the design can detect a change before the test runs. A test is possible when three things are true. Your plan can run in some regions and leave comparable regions without it. You track a daily number by region, such as sign-ups or visits to your site. And that number is large and steady enough for the change you're looking for to show above its normal ups and downs. Out-of-home is bought by place, which suits a regional test, while a plan that has to run nationally leaves nothing to hold back. On a managed first test, the power check runs before you pay, and the test is part of it when your numbers can support one. [See managed services](https://adbuymcp.com/managed-services). 1. 01 ### Match the regions Candidate UK regions are paired on how your number behaved before the campaign, so the regions with ads and the regions held back start out alike. 2. 02 ### Check the power first A calculation on daily observations, at a 5% significance level and 80% power, works out whether the design can detect the change you're looking for. 3. 03 ### Report what the design supports A design that passes the power check compares how much your number moved where the ads ran with how much it moved in the regions held back. It reports the lift with its confidence interval. ### The engine stops when the design is too small to answer. It returns no lift figure. You get the number of days of data the design would need and the smallest lift it could detect, so you know what would make the test worth running. ### The engine stops when a held-back region received the ads. It names the regions that received the campaign, so you can keep the ads out of them for the test or hold back other regions instead. Bring your daily numbers to a call, and we'll work out with you whether a test is possible at your plan's size. [Book a 30-minute call with the founder](https://adbuymcp.com/contact#book) ## Could your test detect a useful change? Start with observed daily outcomes and their variability. The initial values are illustrative inputs. This independent planning screen does not run the product’s geo-lift engine or measure lift. Baseline daily outcomes per arm Daily standard deviation per arm Minimum useful relative change (%) Usable post-period days per arm Observed baseline days Lag-one daily correlation (0 to 0.94) I have separate treatment and control markets, with a plan to check spillover and pre-period comparability. I estimated variability and daily correlation from the observed baseline. The initial example values are not evidence. ### A control design is still needed The approximation needs **212 usable post-period days per arm** for the chosen change. With 28 days, its minimum detectable relative change is about **27.5%**. Review the control markets and use variability and correlation estimated from your own baseline before commissioning a test. **Read the statistical assumptions** Equal-sized treatment and control samples of daily arm means, equal known daily variance, a two-sided 5% significance level and 80% target power. Required days per arm = ceiling of 2 × (1.960 + 0.842)² × standard deviation² ÷ absolute change², multiplied by (1 + correlation) ÷ (1 − correlation). Absolute change = baseline mean × relative change ÷ 100. The serial-correlation adjustment is our chosen AR(1), long-series planning approximation. It does not replace a fitted geo experiment, capture all seasonality or correct measurement bias. The 28-day baseline rule is a readiness check, not a guarantee. The normal approximation assumes known population variance; an estimated standard deviation and short samples need t-based review. Counts must be large enough for the normal approximation; rare outcomes and few independent markets require a tailored analysis. The actual product uses its own geo-lift power analysis and evidence gates. Observed geo outcomes, live-line delivery mapping and verified control integrity are additional requirements. Spend alone supplies none of them. [NIST’s sample-size reference](https://itl.nist.gov/div898/handbook/prc/section2/prc222.htm) explains the normal approximation and its known-variance limit. The factor of two here comes from two equal-sized independent arms. See also NIST’s [autocorrelation guidance](https://www.itl.nist.gov/div898/handbook/eda/section3/eda35c.htm). [Read the geo-test design guide](https://adbuymcp.com/guides/designing-a-geo-lift-test). Our trust page sets out how each channel is bought, what your weekly report shows and how we handle your data. [Read how we buy each channel before you start](https://adbuymcp.com/trust) ### Look up the measurement terms in the glossary - [Reach](https://adbuymcp.com/glossary#reach) - [Deduplication](https://adbuymcp.com/glossary#deduplication) - [Proof of play](https://adbuymcp.com/glossary#proof-of-play) - [Geo-lift test](https://adbuymcp.com/glossary/geo-lift) - [Holdout](https://adbuymcp.com/glossary/holdout) - [Matched market](https://adbuymcp.com/glossary#matched-market) - [Statistical power](https://adbuymcp.com/glossary/statistical-power) - [Minimum detectable effect (MDE)](https://adbuymcp.com/glossary/minimum-detectable-effect) - [Brand lift](https://adbuymcp.com/glossary/brand-lift) - [Run a geo-lift test](https://adbuymcp.com/how-to/run-a-geo-lift-test) ## Questions founders ask about measuring awareness ### Can the report show which people in my market saw my ads? It shows what each channel delivered at the level that channel counts: places and times for billboards and screens, households for streaming TV, downloads for podcasts, and impressions and clicks for ChatGPT ads. None of them tells you who the person was, so the report can't name who saw you. ### Can I trace a sale back to a billboard or a podcast ad? In aggregate, rather than sale by sale. A billboard has no click and a podcast download carries no buyer's name, so the routes are your own numbers, what new leads tell you, and a geo-lift test when your numbers can support one. ChatGPT ads count clicks and conversions, for that channel only. ### What should I set up before my plan starts? Make sure you can see searches for your company name and direct visits in your own tools, and start asking every new lead where they first heard of you. A few weeks of those numbers before your start date give you something to compare the campaign weeks against. ### Has a geo-lift test run on a live AdBuyMCP campaign? The engine, its power check and both refusals are built, and today they run on labelled sample data. No media has been bought through AdBuyMCP yet, so the first live result will come from the first campaign whose numbers can support a test. ### Can you run a brand-lift survey? A brand-lift study asks a panel whether they recall you, comparing people who saw the ads with people who didn't. The product can scope and quote one through Cint, a research panel company, but no Cint contract is signed yet, so brand lift isn't part of a plan today. Your plan and price ## See what your market costs at 3, 5 or 12 times a week Give us your company website or work email. We draft who buys from you, you confirm your market and pick 3, 5 or 12 times a week, and you see one all-in monthly price across out-of-home, streaming TV, podcasts and ChatGPT ads. Talk your market and your plan through with James Rooney, who built AdBuyMCP and wrote its code. [Get my plan and price](https://adbuymcp.com/start) [Book a 30-minute call with the founder](https://adbuymcp.com/contact#book) Free to price, with no card. Media is bought only after your plan, ads and start date are confirmed with you. [See how the price is worked out](https://adbuymcp.com/pricing)