Why last-click does not apply here
In e-commerce, the conversion happens inside your own systems, so the measurement question is which touch to credit. In cannabis, the conversion happens inside someone else's system, and you often do not learn about it for weeks — if at all, and rarely at the individual level. Everything downstream of that is affected.
Without a lawful data connection to the retailer, you usually cannot identify purchasers, retarget them as purchasers, or calculate customer lifetime value from brand-owned data. Person-level holdouts and cost-per-acquisition comparisons are also incomplete when the acquisition is not observed. The wrong response is to fall back on what is easy to count — impressions, followers, engagement rate — and present it as performance.
These activity metrics tell you the work happened, but they do not tell you if your cannabis marketing measurement strategy actually worked.
Layer one: the demand you own
Start with signals that live entirely inside systems you control. These are comparatively direct signals when event definitions, consent, and tracking are reliable.
Branded search volume and direct traffic
When someone hears about your brand and goes looking for it, that intent surfaces as branded queries and direct visits. It is the closest thing to a demand thermometer you own, it moves within days of real awareness work, and it is still affected by tracking quality, brand-name ambiguity, seasonality, and other activity. Track it weekly, not quarterly.
Store-locator use
This is a high-intent signal that many brands under-instrument. A visitor searching your locator for a dispensary near them has declared purchase intent. Instrument the search, the result click, and the outbound handoff to the retailer as three distinct events. Most brands track none of them, or track a page view and call it a metric.
Product-page depth
Which SKUs get researched, which specifications get read, which strains hold attention. This is a preference signal your sales team can use with buyers immediately.
Owned audience growth and response
Newsletter subscribers, open and click behavior, SMS response. A permissioned list reduces dependence on a social feed and can usually be exported, while still remaining subject to consent, suppression, legal, deliverability, and provider requirements. These four give you a weekly read on whether demand is being created.
The instrumentation work behind them belongs to the same discipline as cannabis search visibility — if your pages are not crawlable and your events are not defined, none of this data exists.
Layer two: the movement you can get from retail
The purchase data exists. It just is not yours by default. Getting it is a commercial negotiation and an operational habit, not a technical problem. The metrics that matter are sell-through rate by door, reorder rate and interval, velocity per door per week, door count and net new doors, and out-of-stock frequency. Together they answer the question marketing is actually accountable for: is the product moving faster, in more places, more often?
Reorder rate is one useful signal in the set when transfers, promotions, stock, and account practices are also considered. A first order can be won with a good pitch or a discount. A second order is stronger evidence of replenishment than a first order, although inventory transfers, promotions, and account practices still affect the signal.
If reorder rate is flat while impressions climb, you have an awareness program that is not reaching the shelf. Out-of-stock frequency belongs in the marketing report even though it is a supply metric.
Demand generation for an unavailable product can send a shopper elsewhere, and it will show up as a marketing failure in a quarterly review unless somebody put it on the page. Getting this data is the constraint. Ask for it as part of the account relationship, use whatever retail analytics your distribution partners provide, and record what your own reps observe door by door.
Document the dataset’s limits and review it consistently rather than waiting for information the partners cannot provide. The full method sits in turning dispensary sales data into marketing decisions.
Layer three: the bridge — geography and timing
When person-level purchase data is unavailable, a market and time period can support a directional test with documented limitations.
Matched-market testing
Pick two regions with similar door count and baseline velocity. Run the campaign in one. Compare movement over the following weeks. It is not a randomized trial and it will not survive an academic review, but it produces a directional answer from real sales in a category where nothing else does.
Time-boxed lift
Establish a pre-period baseline, run a defined activity with a hard start and end, and measure the delta in both layer-one and layer-two metrics. The discipline is the definition — vague campaigns that trail off cannot be measured at all.
Geographic correlation
When locator searches spike in a zip code and velocity rises in the doors serving it two weeks later, that is a real pattern. One instance is weak evidence; repeated patterns justify further testing but still do not prove causation.
Ask the customer
A single question at newsletter signup or on an event intake form — how did you hear about us — is unfashionable and useful. It will not be statistically clean, but it will surface channels your analytics cannot see, including budtender recommendation, which may be important and is often absent from brand-owned analytics.
What to build first
- Instrument the store locator properly. Search performed, result viewed, retailer handoff clicked, plus market.
- Establish a branded-search baseline so you can tell demand creation from demand capture.
- Get sell-through and reorder data from your top ten doors, not all of them.
- Define one matched-market test for the next campaign.
- Add the how-did-you-hear question to every owned intake point.
Notice what is not on the list: a dashboard. Dashboards get built before the underlying events exist, and then get abandoned. Instrument first, report second.
What most brands measure instead, and why it fails them
Follower count, impressions, reach and engagement rate dominate cannabis marketing reports because they are free, immediate, and readily available. They are also the metrics least connected to whether anyone bought anything. A post can reach a hundred thousand people who are outside the product’s lawful market or intended audience, in states you do not distribute to, and the report will look excellent. Rank alone is the equivalent mistake in search.
A position with no clicks, or clicks with no locator use, is a vanity number wearing a technical costume. Judge search by whether it produced qualified behavior, not by where a keyword sat. The deeper failure is measuring channels in isolation.
In cannabis the channels are a chain: an event produces content, the content feeds creators, creator content drives branded search, search lands on a product page, the product page sends someone to a locator, the locator sends them to a door where a budtender decides the outcome. Scoring one link of that chain and ignoring the rest produces confident wrong conclusions — usually that the last visible touch deserves the budget.
The chain view is the core of the cannabis digital marketing system.
The limits, stated plainly
This system will not give you a defensible return-on-ad-spend figure, and any partner promising one in a market where the brand does not own the transaction is selling a model, not a measurement. Retail data arrives late and inconsistently. Matched markets are not perfectly equivalent. Seasonality, promotions, competitor launches and supply gaps all contaminate the read. Small brands with few doors will have samples too small for confidence, and should treat directional movement as directional.
Use it accordingly: to decide what to do more of and what to stop, at the level of channels and campaigns, over quarters — not to settle an argument about which post drove which jar.
What most operators miss
Measurement is a design decision, not a reporting task. If nobody defines the baseline, window, market, and metric before the campaign runs, later analysis will usually be incomplete and easier to overstate. The most common reason cannabis marketing looks unmeasurable is that measurement was never designed into it. The second: budtender recommendation may influence purchase while remaining absent from the brand’s dashboard.
Any measurement system that ignores it will systematically under-credit offline work that may contribute to product movement.
The next step
Pick one campaign in the next quarter and design its measurement before its creative. Name the baseline period, the test market, the control market, and the three numbers that will decide whether it worked. Then hold the report to those three numbers only. Talk to High Rise about building the measurement layer before the next campaign, rather than reconstructing it afterwards from whatever data survived.