Define the cohort before looking at results
A cohort is a group of sites picked by one stated trait, such as launch month. For each microsite, record when it became usable, when its contact path was tested and any outages. A registration date doesn't show when a customer could use the site. At Trojan Digital Marketing, we report our own microsite networks this way, and Jeff Clerisier, our founder, will go through your site list with you. Send it over: the first strategy call is free, and during business hours we aim to reply within five minutes.
We pick one rule for assigning sites to groups and stick to it. If a domain was registered in January but its pages and form went live in March, our report keeps both dates and says which one sets the cohort. Otherwise a later discussion may compare unlike periods without realizing it.
The portfolio decision guide uses cohorts to inform investment choices. Here's how the rule plays out in our own numbers. Our original 200 Indiana residential roofing microsites brought 11 calls and form inquiries from June 29 through August 31, 2026. We added 88 more sites later, and they sit outside that count.
Keep the inventory larger than the performance subset
We keep an inventory of every scoped site, including ones in development, offline, awaiting review or hit by a technical problem. A performance view may focus on usable destinations, but it should not make the other sites disappear from the owner's understanding of the engagement.
Give each site a current state and a dated history where meaningful. A launch delay, broken form or extended outage changes the observation period. Recording that context is more useful than quietly excluding the site to improve a portfolio average.
| Field | Question it answers |
|---|---|
| Cohort rule | Why are these sites grouped together? |
| Scoped and usable sites | What exists and what can customers currently use? |
| Launch dates and active time | How long has each resource been available? |
| Search observations | What discovery does the available data show? |
| Unique inquiries and qualification | Which customer requests reached the defined stages? |
| Known gaps | What cannot yet be compared confidently? |
| Next action | What does the owner intend to change or inspect? |
Show totals before calculating an average
We start with the site count and the unique inquiry total for the stated period. Then identify the denominator for any average. Inquiries per scoped site, inquiries per usable site, and inquiries per active site month answer different questions.
Suppose, purely as an illustration, one group contains ten sites available for six months each and records twelve unique inquiries. That represents sixty active site months under this simplified definition, producing an observed ratio of 0.2 inquiries per active site month for that group and period.
Multiplying that ratio by twelve produces an annualized arithmetic scenario, not a measured full year result. Seasonality, early visibility and changing conditions may make the relationship nonlinear. The first year scenario guide explains how to keep that distinction visible.
Include distribution so one site does not speak for all
A portfolio total can come from a small number of sites while others receive no inquiries. Show a distribution that makes that concentration visible, such as site counts with zero, one, or several observed inquiries during the same defined period. Use ranges suited to the actual dataset.
Where the sample is small, avoid presenting a precise looking average as stable. Keep the individual destinations available for review. A strong result on a single service resource may justify investigating what made it useful, but it does not establish the expected performance of every new site.
Compare related business contexts where possible. A repair focused resource and a planned remodeling resource may face different buying cycles and inquiry definitions. A broad national average can obscure the practical decision for one trade.
Keep search discovery separate from customer outcomes
Search impressions and clicks can reveal whether a resource is being discovered for relevant questions. They do not establish how many unique customers became qualified opportunities. Keep search measures, website interactions and business records in distinct parts of the report.
Search Console documents how its performance measures work. Note the property, date window and filters used. A chart assembled from different properties or unequal periods should explain those differences before inviting comparison.
Use the Generative AI performance report for Google AI impressions. Keep inquiries in the business record and preserve the report type with each cohort export.
Deduplicate at the business level
A customer may interact with more than one microsite or use a form and a call. The business needs a rule for linking those interactions to a unique project inquiry. Otherwise a portfolio with more destinations may appear to create more opportunities partly because the same customer is counted repeatedly.
Keep source contributions where they are supported, but separate them from the unique total. An assisted interaction can be useful without creating another lead. The measurement playbook explains the inquiry record needed for that reconciliation.
Pending qualification should remain visible. Do not count every unresolved request as qualified to make a cohort look stronger, or as invalid because the team has not yet completed its review.
Explain what a cohort comparison cannot prove
- A: current page
- Observe the starting version.
- B: one change
- Compare with the same measurement.
A before and after comparison may reflect seasonality, changes in the business, different trade coverage, a stronger offer or measurement changes. It does not by itself establish what would have happened without the microsites.
Record major changes alongside the timeline: new project evidence, form revisions, market scope, advertising activity, staffing and content updates. Those notes help identify plausible explanations and better questions for the next review without presenting an unsupported causal claim.
Use the report to select an action. A group with relevant discovery and repeated form failures needs operational repair. A group with little history may need launch verification and continued observation. A group answering duplicated customer tasks may need an architecture review before further expansion.
End with a decision and a review condition
Assign the next task, its owner and the condition that will trigger another review. Keep the reasoning connected to the report. Aim for a portfolio the business understands and can improve with the same evidence at each review, even in months when the chart dips.
Bring us your site list and how you label inquiry stages. A baseline shows what's missing and makes the next membership or expansion decision easier to judge.
Frequently asked questions
What date should define a microsite launch cohort?
Choose and document a consistent usable launch milestone. Preserve registration, publication and contact verification dates separately when they differ.
Should sites with no leads be removed from the report?
No. Keep them visible with their operating state and observation period. Excluding them can make the reported portfolio performance misleading.
Is an annualized ratio the same as a first year result?
No. Annualization extends an observed rate through arithmetic. It does not establish what a complete year would produce under seasonality or changing conditions.
Can cohort reports prove that every inquiry was incremental?
Not on their own. They describe observed activity and timing. Customer overlap, other channels and business changes require separate analysis before making a causal claim.
