A two-week proof of concept for brand search monitoring (with a 7-day version)

Key takeaways

• Agree four or five acceptance tasks before day one and score each as met, not met or unresolved.

• Keep the scope small: one brand, about five keywords, two or three US cities, desktop and mobile.

• Check that a completed check with no ads looks different from a failed or missing check.

• Test a handoff to a colleague who did not set up the trial.

• Record promises about future releases as unresolved, never as met.

A trial cannot prove long-term return on investment, and it cannot show that a service sees every ad in every auction. Nobody can, because search results change with location, device, time and many signals you do not control. What a trial can show is whether the observations are clear, whether the gaps are visible, and whether the people who will use the tool can finish real tasks with it.

Before day one: write the acceptance sheet

Give one person ownership of the configuration and a second person the job of checking whether findings make sense without the first person's explanation. Then agree the tasks that decide the purchase. A practical set for a brand search team:

1. Retrieve a specific observation. Find the ad a given advertiser showed for a given keyword, location and device on a given day.

2. Separate your own ads from external advertisers. Your own domains should never be counted as competitors.

3. Identify an unfamiliar advertiser. Get from a new domain to a reasoned note about who it is and what the ad said.

4. Explain a gap. Show the difference between a check that ran and found no ads, and a check that failed or did not run.

5. Hand a finding to a colleague. Someone who did not set up the trial should understand one finding from an export or shared view.

Add a task for any requirement that would block adoption, such as affiliate identifier detection or Bing coverage. If the vendor does not claim a capability, do not spend the trial hoping to discover it.

The configuration

Keep it small enough to review by hand. For a fictional brand, Northstar, selling outdoor gear at northstar.example, a reasonable trial scope is:

• Keywords: "northstar", "northstar boots", "northstar jacket", "northstar outlet", "northstar coupon". Five terms that answer different questions.

• Engines: Google, plus Bing if your audience uses it.

• Locations: two or three US cities where you sell, for example Denver, Chicago and Atlanta.

• Devices: desktop and mobile.

• Frequency: at least daily, more often if your plan allows and you will actually review it.

Record the configuration and why each part is there. If you change something mid-trial, note the date so that you do not compare different coverage later.

Days 1 and 2: set up and confirm what is being checked

Create the project, add your own domains (main site, regional storefronts, any separate domain used in ads) and enter the keywords. Ask the vendor, or read the plan page, to confirm how many checks the configuration uses per day and per month. Run a first check on demand so you have results to look at on day one.

Confirm the basics: each observation should show the keyword, engine, location, device and time, and the time zone should be clear. If a location cannot be selected or a keyword is misspelled, fix it now and write down the change.

Days 3 and 4: check collection context

Open results from every location and both devices. Compare planned combinations with completed ones. You are looking for three different states, and the tool should make them distinguishable:

• A completed check with ads, showing who appeared.

• A completed check with no ads, which is a real result.

• A check that failed, is pending or did not run, which is a gap and not a result.

If "no ads" and "failed" look the same, every later count becomes unreliable. Note how the product shows each state.

Remember that a location is part of the observation context, not proof of how an advertiser targeted its campaign. Google explains that its location targeting uses signals like IP address, device location and searches, and that it cannot guarantee accuracy in every situation (Google location targeting).

Day 5: review destinations with an analyst

Pick several observations that lead to different destinations and ask an analyst to say what the tool captured and what it inferred. Is the destination URL the one shown in the ad, a final URL after redirects, or a stored copy of the page? Google documents that with parallel tracking, people go directly from the ad to the final URL while click measurement happens in the background (Google tracking documentation). A visible URL may not show every tracking step, so ask exactly which parts of the path each vendor stores.

Days 6 and 7: run your normal weekend

Let the schedule run while the team follows its usual weekend or reduced-staff routine. Do not invent an emergency. On Monday, check what arrived, what needed attention and who was responsible for it. If notifications are part of the offer, confirm which events actually trigger them today and who receives them.

Days 8 and 9: classify advertisers and correct a mistake

Take a known retailer and an unfamiliar domain. Label the retailer's relationship using your partner records. For the unfamiliar domain, gather what you can: the ad copy, the destination, and public information such as the advertiser's entry in Google's Ads Transparency Center, which is searchable by advertiser or website name (Google Ads Transparency Center). If the evidence is incomplete, leave it unclassified and say what is missing.

Then deliberately correct a label. The tool should let you change your interpretation without rewriting the original observation.

Days 10 and 11: hand a finding to someone else

Give a colleague who did not set up the trial one finding. They should be able to name the keyword, location, device, time, advertiser and destination, and say what action is being proposed. Write down every question they had to ask. If the product exports tables, use the export. If it has no sharing workflow, test the manual process you would really use.

Time this handoff and the investigation behind it. These are your own measurements, useful for the decision, and they should not be turned into a general productivity claim.

Days 12 to 14: check the contract and decide

Review data retention, cancellation, support, and the price of the coverage you actually tested. Ask for a sample export if you may leave later. Then score each acceptance task.

Mark a task unresolved when the only evidence is a promise about a future release. Finish with a named decision owner and a written list of conditions for purchase.

The acceptance sheet above is ready to copy. It is also collected, with the vendor scorecard, the observation brief and the weekly report, on our templates page.

The compressed 7-day version

Many self-serve trials last a week. You can still test the same things if you set up on day one:

• Day 1: configure the project, own domains and keywords, and run an on-demand check.

• Day 2: check context and the three result states across locations and devices.

• Day 3: review destinations and classify a retailer and an unfamiliar domain.

• Days 4 and 5: let the schedule run, then check what arrived and who reviewed it.

• Day 6: hand one finding to a colleague using an export.

• Day 7: score the acceptance sheet and review price and cancellation terms.

Running this proof of concept in Ads Lighthouse

Ads Lighthouse is built for this kind of evaluation. Create a free account, then start a 7-day free trial from Plans & usage. Plans start at $49 per month and are listed on the pricing page. The free trial help article explains which allowances apply during the trial, and create your first project walks through the setup for day one.

In a project you add one brand, its own domains and the keywords to monitor, then choose Google, Bing or both, the cities or regions, desktop and mobile, and a schedule from daily up to the maximum your plan allows. You can also run an on-demand check at any time. Each observation keeps its keyword, engine, location, device and time. The paid results explorer lists advertisers with filters for device, operating system, country and domain, and segments for affiliate sites, coupon sites, lookalike domains and ads that mention your brand. The advertiser and keyword reports let you drill into saved ads, and you can label each advertiser's relationship, for example competitor, authorized retailer, affiliate or unclassified. Tables export to CSV or Excel for the handoff test, and you can invite teammates to the account.

Be clear about limits during the trial as well. An observation is what appeared for one search at one time; it does not prove what an advertiser bid on, how much it spent or that it broke a rule. Alert rules cover four events: a new advertiser, a returning advertiser, a brand mention in ad copy and a detected affiliate ID. The first check is the baseline, so advertisers seen there do not trigger a new advertiser alert: review those first results yourself. When one check finds more than 10 new events for a rule, you get one digest instead of a message for each.

The PPC teams use case shows the day-to-day workflow after a trial, the Google Ads and Microsoft Advertising (Bing) pages describe coverage per engine, and the glossary defines terms such as check and observation. If you are running trials with several vendors, start from the comparison pages. To prepare the scope, see how to choose brand keywords, how to compare pricing by coverage and the vendor scorecard.