The Digital Market Was There. Why Were Georgian Small Firms Still Offline?
This series examines a specific puzzle among small firms in Georgia: many businesses had access to the internet, yet very few used it as a dependable route to customers. The central question is not whether digital technology was available. It is why firms that were connected—and, in many cases, interested in going online—did not complete the work of becoming online sellers. Understanding that distinction gives readers a way to judge the evidence for training, incentives, and other attempts to widen market access without treating every failure to adopt as a skills deficit.
The gap was not access to the internet
The setting matters. Georgia had spent years expanding broadband infrastructure, and the national innovation program delivered through the Georgia Innovation and Technology Agency was meant to help firms use that infrastructure productively. But the study’s formative evidence from 2,180 small businesses outside Tbilisi shows that physical access and commercial use were different things.
In 2018, 91% of firms in that formative sample had internet access for home or business use, and 60% of respondents had their own email address. Those figures suggest that the first layer of the digital problem—being able to connect and communicate—was already present for most firms. The next layer was much thinner: only 4% had a business email, 3% had a website, and 2% used an e-commerce platform. A Facebook business page was more common, at 11%, but only 6% had ever received an order online.
These numbers are not interchangeable. A social-media page can make a business visible without providing a reliable way to place, pay for, and fulfill an order. A website can display a product without creating demand. An email address can support communication without changing where customers come from. The meaningful gap is therefore between a firm’s ability to access digital tools and its ability to turn those tools into a transaction.
The experimental firms were a narrower group than the 2,180 businesses in the formative survey. They were drawn from six regions outside Tbilisi and had indicated interest in participating in e-commerce training. That detail matters in both directions. The firms were not random representatives of every Georgian business: they had already shown some interest. Yet even among firms willing to sign up or express interest, a digital sales channel was far from routine. The case is consequently about a barrier that can survive connectivity and stated aspiration.
The difference between private connectivity and commercial readiness is easy to underestimate because the tools can look deceptively simple. An owner may use a phone to communicate every day and still lack a business profile that tells a stranger what is for sale. A profile may exist but omit the information needed for a purchase. Even when a customer can make contact, the firm may not have settled how to receive a noncash payment, confirm an order, or deliver outside its immediate neighborhood. These are not separate “digital skills” in the abstract. Together they form a market relationship, and a weakness at any link can make the whole channel unattractive to use.
Small firms also face a distinctive allocation problem. The person who decides whether to try a new sales channel may be the person who must photograph products, answer questions, arrange payment, and fulfill the order. The cost is not only the time spent learning a platform. It is the opportunity cost of interrupting existing work for a return that may be delayed or uncertain. That is why a low adoption rate cannot by itself tell us whether firms lack knowledge, money, customers, confidence, or spare managerial attention. The visible outcome is the same, but the decision behind it is not.
“Going online” contains several different problems
It is tempting to describe the gap with a single phrase—small firms were “not digital.” That phrase hides the decisions that adoption requires. A firm must first establish a business-facing presence. It then needs a workable path for a customer to make contact, place an order, and pay. It may need to deliver a product or provide a service. After the first transaction, someone must maintain the channel, answer inquiries, update information, and decide whether the returns justify continuing.
Each step can fail for a different reason. A technical implementation problem might involve not knowing how to configure a profile or list an offering. A financial problem might arise when setup costs arrive before any customer revenue. A demand problem appears when the owner cannot see who will buy through the new channel. A behavioral or organizational problem arises when the work is easy enough in principle but competes with daily operations, leaving no clear moment when it must be done.
Those explanations can look similar from the outside. In every case, the firm remains offline. But they imply different remedies. If a business cannot perform the task, training or hands-on support may lower the cost. If it can perform the task but expects too little return, a subsidy or a buyer may change the calculation. If the expected benefit is adequate but distant, uncertain, or easily postponed, a deadline may matter more than another lesson. And if the business lacks the time or managerial bandwidth to coordinate setup, a larger financial offer may still leave the bottleneck untouched.
Training is an intuitive first response because digital adoption appears to require knowledge. The Georgian program was designed around that intuition: teach firms how to use familiar platforms, understand customers, register on e-commerce sites, and build an online presence. If firms stayed offline because they did not know how, lowering the implementation cost should have produced at least some immediate movement.
But the logic of training should not be confused with the conclusion that training must work. Knowledge is only one input into a decision. A firm can know how to open a channel and still decide that the opportunity is not worth the time, that the customer is too uncertain, or that the owner cannot fit the task into the week. The important question is therefore comparative: when the same underlying problem is approached by lowering the presumed cost of entry versus making the benefit concrete, which part of the decision changes?
This is also why the word “barrier” needs care. A barrier is not necessarily a permanent wall or a missing resource that can be identified by asking whether a firm has internet access. It can be a threshold: below a certain combination of expected benefit, urgency, and available attention, the owner does not act; above it, the same owner may complete the task quickly. Such a threshold can make a training program look ineffective even if participants learn something, while a buyer’s deadline can make a small change in circumstances look unusually powerful. The comparison matters because it reveals which condition moved behavior, not merely which program was more popular.
The distinction has a practical implication for how success should be measured. Counting profiles or course completions captures preparation, but it does not show whether customers can actually buy. Counting a single order captures activation, but it does not show whether the firm can maintain the channel when the special opportunity disappears. A serious account of market access has to follow the sequence far enough to see whether the new route becomes part of ordinary business. Otherwise, policymakers may reward visible setup while missing the recurring work that determines whether adoption pays off.
From a digital presence to a market relationship
The study’s subject is not a website, a course, or a payment. It is the movement of Georgian small firms from potential access to actual participation in digital markets. That movement has at least three useful milestones.
The first is readiness: the firm has the basic presence and information needed for a customer to find it. The second is activation: a customer can complete an online purchase or otherwise engage the firm through the channel. The third is persistence: the firm continues to maintain and use the channel after the initial reason to set it up has passed. A program can succeed at the first milestone and fail at the second. It can succeed at the second and fail at the third. Treating all three as “adoption” would make a short-lived opening look like lasting market access.
This distinction also explains why the Georgian case has significance beyond a single technology program. Digital markets reduce search frictions by helping buyers and sellers find one another, but reducing a friction in the wider market does not guarantee that every small firm can act on the opportunity. A business still has to recognize the opportunity, organize a response, and believe that the result will repay the effort. The market may be technically open while remaining practically difficult to enter.
The evidence was built to examine those layers rather than merely describe them. One randomized experiment offered interested firms in-person training intended to address the cost of capability. A later experiment told eligible firms that a research partner would place an order if the firm could make the purchase happen online. The two approaches changed different sides of the decision: one tried to make the task easier, while the other made the immediate value of completing it more visible and time-bound. Follow-up surveys then made it possible to ask whether any initial change endured.
That design does not turn the firms into a universal model of small-business behavior. The training sample was selected from firms outside Tbilisi that expressed interest, and the later demand sample consisted of firms that were still operating and able to offer goods or services at the intervention time, along with additional eligible firms. Results from such groups can identify meaningful causal responses within the setting while still leaving open how firms in other regions, sectors, or market conditions would respond.
The design also requires careful language about what is observed. An increase in the ability to sell online during a defined intervention window is evidence that a particular opportunity can unlock action. It is not automatically evidence of higher productivity, a permanent customer base, or broad-based growth. Conversely, a null result from a training offer is evidence that the offer did not change measured outcomes for the assigned firms; it is not proof that technical knowledge never matters.
Time is part of the question, not just a follow-up detail. In the first moment, an owner may be deciding whether to make a one-time investment in setup. Later, the decision changes: should the firm keep information current, respond to customers, absorb payment and delivery work, and continue when the original incentive is gone? A policy that helps with the first decision can still fail to change the second. Equally, an early difference between treated and untreated firms can narrow if market conditions improve and firms that initially waited begin to adopt on their own. The meaning of an intervention therefore depends on whether the goal is experimentation, a lasting sales channel, or broader firm growth.
The Georgian study is valuable because it keeps these questions connected without assuming that they have one answer. It asks what happens when capability is made easier to acquire, what happens when a concrete buyer makes the payoff immediate, whether those two changes reinforce each other, and what remains visible years later. The answers can support a more precise policy judgment than either “small firms need training” or “incentives solve adoption.” They can show which part of the journey is responsive to a particular change in circumstances—and where the journey still stops.
The practical lesson at the opening is simple but consequential: “offline” is an outcome, not an explanation. The Georgian firms’ position could reflect missing skills, weak expected demand, financial costs, hesitation, or the difficulty of making a new channel fit inside a small operation. The evidence becomes useful only when those possibilities are separated.
That is why the next question is about the architecture of the comparison itself. To understand what the interventions reveal, readers must see how the study distinguishes an offer from actual participation, a capability intervention from a demand shock, and immediate activation from a result that survives after the original opportunity is gone.