On screenA matched-market test compares a place where ads run with a place where they do not. The comparison is only as good as the match. If the two areas differ in ways that have nothing to do with your campaign, you will end up measuring those differences and crediting them to the media.
This guide covers method only: how to pick the pair, how to record the setup before the flight, and what to do when no pair is clean. It makes no claim about what any campaign will deliver. For the wider plan this test sits inside, see build a DOOH measurement plan before you launch.
What a control market does and does not do
The control area answers one question: what would have happened in the test area if you had not advertised? You never observe that directly, so you estimate it from a place that was behaving similarly before the campaign and received no ads from you during it.
The IAB's digital out-of-home measurement guide lists matched market testing, along with synthetic control groups, among the attribution approaches used to isolate campaign impact. You can read its summary at Cracking the Code of DOOH. Both approaches rest on the same idea: the comparison group stands in for the missing counterfactual.
A control area does not prove attribution on its own. It gives you a defensible estimate of incremental change, with a margin of doubt you should state to the client.
The four matching criteria
Match on the factors that move your outcome metric, not on whatever is easiest to pull. For most local advertisers that means these four.
| Criterion | What to compare | Common trap |
|---|---|---|
| Customer base | Who buys: household type, age mix, income band, existing customer count, visit or purchase frequency | Matching on total population when your buyers are a narrow segment |
| Baseline sales | Weekly outcome metric (sales, bookings, calls, new customers) over the pre-period, in level and in trend | Matching on level only; two areas with equal averages can be moving in opposite directions |
| Seasonality | Whether both areas rise and fall together across the year: weather, school calendars, local events, tourism | Pairing areas whose peak seasons fall in different months |
| Promotions and other media | Your own offers, price changes, openings, and other paid media in each area, plus known competitor activity | A coupon drop, direct mail wave or local sponsorship that touches only one side |
Customer base
Start with your own customer file if you have it. Compare the areas on customer counts, average order or ticket, and the share of repeat versus new buyers. If you lack first-party data, use census-style demographics, but treat them as a rough proxy for your actual buyers.
Baseline sales and trend
Pull the outcome metric by week for the pre-period. Look at the shape, not only the average. Two areas that track each other week to week are a better pair than two that happen to share a mean. If you can, plot both lines and check whether they move together before you commit.
Seasonality
If your pre-period is short, you cannot see seasonal shape, so lean on prior-year data. A control that peaks in a different month will drift away from the test area for reasons unrelated to media. This matters most for flights that cross a holiday or a weather change.
Promotions and other media
Ask the client for a calendar of every planned offer, price change, store opening, and other paid channel by area, then freeze it where you can. A regional radio buy or a franchise co-op program that reaches only one area can swamp the effect you are trying to read. Agree before launch that neither area will get new promotions that the other does not.
A step-by-step way to pick the pair
- 01
Define the outcome metric and the unit
Decide what you are measuring (store visits, bookings, calls, branded search, sales) and at what geography: zip code, neighborhood, metro or a cluster of venues. The unit of the test has to be a unit you can both target and measure.
- 02
List candidate areas
Start with areas where you can both deliver media and observe the metric. Then screen out any with a recent opening, closure, remodel or one-off event in the pre-period.
- 03
Score candidates on the four criteria
Compare each candidate to the proposed test area using pre-period data. A simple correlation of weekly outcomes, plus a side-by-side of customer profile, is enough for most local tests.
- 04
Check for spillover
Look at how people move. If residents of the control area commute through, shop in, or visit venues in the test area, they may be exposed to the campaign, and your control is contaminated. Distance, a river or highway barrier, or a separate metro can all help.
- 05
Rank, then keep a backup
Choose a primary control and at least one backup. If the primary is disrupted mid-flight by weather, a closure or a competitor promotion, the backup keeps the test readable.
- 06
Confirm delivery on the test side
Before you lock the plan, confirm with your media partner that the test area can be served the way you intend, and how delivery will be reported. This is a practical gate, not a formality.
Same metro or different metros?
Splitting within one metro keeps the economy, weather and media environment the same, which helps the match. The cost is spillover: people cross the line you drew. Splitting across metros reduces spillover but adds differences in seasonality, competition and local culture.
There is no universal answer. Fine-grained targeting by neighborhood or zip code makes a within-metro split possible in principle, and geo-targeting lets you hold the boundary, but spillover is about human movement, not ad targeting. If your customers travel widely, favor separate metros. If your business is truly neighborhood-bound, such as a salon or a dental office, a within-metro split can work with a buffer area left out of the analysis.
For campaigns that already span several cities, see multi-city out-of-home campaign planning. If a Phoenix and Salt Lake City plan is on the table, the Mountain West plan shows how those two live CETV Now markets are framed. Treat any pairing as a hypothesis to test against your own data, not as an assumption of similarity.
Document the setup before the flight
The most common failure in market tests is not a bad match; it is a good match nobody wrote down, so the rules shift once results arrive. Put the following in a single shared document and have the client sign off before launch.
- Hypothesis and metric: what you expect to change, which metric reads it, and its source and reporting lag.
- Areas: the test and control definitions (zip codes, venue lists or metro boundaries), the buffer areas excluded, and the backup control.
- Why they match: the pre-period data and the criteria you scored, including the known differences.
- Dates: pre-period, flight and post-period, with time zones and week definitions.
- Media in each area: your planned spend, schedule and creative in the test area, and confirmation that nothing from you runs in the control.
- Other activity: a calendar of promotions, other paid media, price changes and known local events in both areas.
- Delivery evidence: what you will collect to show the ads ran, such as venue-level delivery reporting; see how to read a DOOH delivery report.
- Decision rules: how large a difference you will treat as meaningful, what would invalidate the test, and what you will do if the control is disrupted.
- Compliance: if the advertiser is in a regulated category such as healthcare, financial services, alcohol or legal, a note that the advertiser has checked its own compliance rules for the creative and the measurement.
What to do when the areas are imperfect
Perfect pairs are rare. Imperfect ones are still usable if you adjust the analysis and the claim to fit.
- Compare changes, not levels. Measure each area against its own pre-period, then compare the two changes. This absorbs a constant gap in baseline sales between areas, though not a gap in trend.
- Use more than one control. Averaging two or three partly matched areas often behaves better than relying on one that is a close but imperfect match. A synthetic control, which weights several areas to mimic the test area's history, is the formal version of this idea.
- Exclude contaminated weeks or zones. If one control week had a one-off event, flag it and show the result with and without it.
- Adjust for known differences. If one area ran a local promotion, record its dates and model or exclude that period rather than ignoring it.
- Lengthen the pre-period. A longer history makes poor matches easier to detect and correct.
- Narrow the claim. Report a range, name the weaknesses, and avoid language implying precision the design cannot support.
Be careful with small areas. A handful of locations produces noisy weekly numbers, so a visible gap may be chance. If you only have a few units on each side, treat the test as directional and say so. Pairing the test with a second read, such as branded search lift, can show whether a sales movement is supported by another signal.
Running a market test with CETV Now
CETV Now is a self-install digital out-of-home (DOOH) network: a media-player device plugs into a screen a venue already owns and runs licensed programming with video ads over live TV. Targeting can be set by neighborhood, city, metro, zip code, sector or vertical, and individual venue, and layered with audience type and daypart. Delivery is reported with venue-level impression reporting.
Live markets today are Phoenix, Dallas, Atlanta, Nashville and Salt Lake City, with Miami coming soon and not yet live. For any other area, availability is unconfirmed; send an inquiry before building a test around it. CETV Now has not published details here on how impressions are counted or on market-level inventory, so ask the team what delivery evidence you can get for the test and control definitions you have in mind.
Ad insertion has no minimums and no long-term commitment, so advertisers set their own spend and schedule. Pricing is quote-based, so a test design should go to the team before budgets are fixed. CETV Now reported figures such as a 4.2x average campaign ROI describe its network overall; results vary by campaign and none of that substitutes for a clean test of your own.
If you plan for clients, the agencies page explains how planning across markets and venue categories works through one network partner, and how to report out-of-home results to clients covers presenting a test read-out honestly. When your pair, dates and metric are drafted, share them through the contact page or at info@cetvnow.com so the team can map the plan to real venues.
Figures reflect CETV Now network data and are provided for informational purposes. Individual campaign results vary; no specific outcome is guaranteed.
