The useful thing about a cannabis receipt is how ordinary it looks. A gummy, a vape cartridge, an eighth of flower, a discount. The troublesome thing is everything behind it. One store may label a product “Blue Dream Joint,” another “Blue Dream Pre-roll.” Markets have different rules. Brands launch and vanish quickly. Prices shift, promotions blur demand, and the same supplier can be staring at twelve incompatible reports from twelve retailers. Headset, a Seattle company founded in 2015, made a business out of turning that clutter into a shared language.
Its three founders - Cy Scott, Scott Vickers and Brian Wansolich - already knew the industry’s peculiar nouns. They had created Leafly, the consumer cannabis guide, in 2010 and sold it to Privateer Holdings. Their second company went backstage. Instead of helping a shopper find a strain, Headset would help a retailer know what to reorder, a brand see where it was losing shelf space and an analyst understand whether a category was gaining market share.
That distinction matters. Headset does not sell cannabis, and it is not a cash-register system. It connects to participating retailers’ point-of-sale software, receives receipt-level sales and inventory information, removes store-level identity from the market sample, cleans the product catalog and produces dashboards and forecasts. The company says more than 3,500 retail partners in the United States and Canada supply its core data. By November 2023, it said it had processed more than $50 billion in transactions.
First, teach the receipts to agree
The first failure is usually visibility. A product sells out, but the supplier learns after the missing sale. A slow item sits on the shelf while cash remains trapped in inventory. A sales representative downloads a spreadsheet, renames columns and discovers that every store describes pack sizes differently. None of this is a cinematic software problem. It is also exactly the kind of recurring annoyance on which durable enterprise companies are built.
Headset’s normalization layer is the important part. Machine-learning systems and human data specialists assign inconsistent retail records to a standardized product catalog. Its analysts then select representative store samples, check projections against government-reported totals where available, reduce outlier effects and produce daily product-level market estimates. A colorful chart is the last inch of a fairly grubby mile.
That work supports a family of products. Retailer gives dispensaries a view of their own sales, customers, inventory and staff performance. Insights gives brands, producers, investors and strategists a market-wide view of categories, prices, demographics and competitors. Bridge lets a retailer give a supplier permission to see that supplier’s sell-through and inventory, creating the conditions for vendor-managed inventory. Vault pipes normalized Retailer, Bridge and Insights data into Snowflake or familiar business-intelligence tools. Newer products add e-commerce menu coverage, conversational queries through Ask Headset and AI-assisted ordering through Bridge Nexus.
“We chose to focus on the retail side because we consider it to be the source of the richest set of data on how cannabis is purchased.”Cy Scott, co-founder and CEOThe change in posture
A chart is nice. An avoided stockout pays the bill.
Headset’s evolution shows what changed its customers’ minds about data. The industry became more competitive, margins tightened and a report that merely explained last month started to feel like homework. Operators needed a shorter distance between signal and action. Bridge turned a brand’s retail blind spot into a replenishment list. Vault stopped asking data teams to abandon the tools they already liked. Nexus and the 2026 Distru partnership connect demand signals to ordering and fulfillment.
Consider Kiva Sales & Service, a distributor serving hundreds of California retailers. Its problem was not a lack of software. It already had an ERP. The failure was the translation work between retail partners: product names did not match and inventory feeds required manual attention. Kiva accepted Headset’s normalized data into its existing system rather than adopting another daily interface. The implementation, according to the company’s case study, took one hour-long meeting and a follow-up. Kiva estimated that its team saved roughly 80 hours each week, partly because account work that once consumed hours became a simpler conversation.
The arithmetic is more persuasive than a slogan. Using the California hourly-pay figure in Headset’s case study, 80 hours represented about $2,320 in weekly labor. Kiva also reported a reduction of more than $3 million in accounts receivable that year, although many factors can influence collections. The clean lesson is narrower: data became valuable when it appeared inside an ordering workflow and gave salespeople a reason to contact customers at the right moment.
What did it cost?
Headset uses custom, sales-led pricing and does not publish a standard rate card. One competing vendor claims typical contracts fall around $2,000 to $5,000 per month with annual commitments, but Headset has not confirmed that range. The honest buying test is therefore operational: compare a current quote with the value of avoided stockouts, reduced inventory, saved analyst time and better promotion decisions.
The database gets smarter when the network gets wider
Headset sits between retail software and classic consumer packaged-goods measurement. Its direct alternatives include BDSA, Hoodie Analytics, Pistil Data and New Frontier Data. A small operator may use a POS dashboard or spreadsheets instead. A large brand may build an internal warehouse. Headset’s claim to difference is a combination of direct, retailer-approved POS depth and broad e-commerce visibility, with the normalization and statistical machinery to make the feeds comparable.
Direct POS data is the gold standard for actual sales and inventory, but it depends on integration and consent. Online menus cover more stores but are weaker evidence of what truly sold. Headset’s current strategy is to join the two. Its e-commerce product says it watches thousands of menu endpoints across legal markets, while thousands of approved POS integrations provide a more exact core. Depth checks breadth; breadth reveals gaps outside the direct network.
This is a familiar data-network loop. Retailers contribute feeds and receive free or useful store analytics. Brands pay to understand the market and, with permission, their accounts. More connections improve coverage. Better coverage attracts more buyers. The difficult-to-copy asset is not an AI chat box. It is years of integrations, retailer relationships, taxonomy decisions and corrections when a new gummy arrives with an imaginative name.
Start with the messy recurring input, not the executive dashboard. Whoever standardizes the weird nouns controls the useful output.
Give the data contributor a reason to participate. A network grows faster when the supplier of the raw material also gets a product.
Meet customers inside their workflow. Vault became useful to Kiva because it fed the ERP instead of demanding another destination.
Move from diagnosis toward action carefully: identify a stockout, suggest the call, generate the order, then measure what changed.
Where the map goes fuzzy
No sample is the market itself. Headset’s method works best when a legal market has enough participating stores, reliable government totals, compatible POS systems and a sample that resembles the whole population. A thin network, a sudden regulatory change or a retailer mix skewed toward one format can weaken a projection. Direct store data also requires trust: a retailer must approve a Bridge connection before a vendor can see its information.
Do not copy this blindly
The model is a poor fit when the underlying commerce is mostly informal, customers will not grant data access, records cannot be normalized economically, or decisions happen too rarely to justify a subscription. A five-store brand that only needs a quarterly market snapshot may be better served by public reports. A multinational operator with a large engineering team may prefer to own its pipeline. AI cannot rescue an unrepresentative sample, and automated ordering can magnify a bad assumption faster than a spreadsheet can.
There is also a practical cultural requirement. Vendor-managed inventory only works when retailers and suppliers agree about incentives. Claybourne reportedly accepted responsibility for inventory that did not move while using Bridge with Kind Delivery. That is more than software. It is an operating promise backed by shared information. Remove the trust or shift all downside to the retailer, and the connection becomes surveillance rather than collaboration.
What comes nextA decade of cleanup earns the right to automate
The recent AI layer is less interesting as a trend than as a sequence. Ask Headset lets a manager type a plain-language question instead of assembling a report. Bridge’s read-only conversational tools can chain a product lookup, a sales trend and an inventory check. The Distru connection is meant to pass that finding into execution. Headset can now travel from “Which stores are low?” toward “Prepare the right order.” It spent the previous decade making sure “the right product” meant the same thing in every system.
That is the most portable idea in the company. In a messy vertical market, intelligence begins with translation. The flashy feature comes last. First make the receipts agree, prove the view is useful, put it where work already happens and measure the avoided waste. Headset’s product is a market map, but its craft is cartography: collecting imperfect observations, correcting the labels and knowing where the blank spaces still are.