How synthetic research is helping telecommunications companies understand churn, test pricing, and decode customer sentiment without commissioning six-month studies that arrive too late to matter
$500 per Lost Subscriber, and Rising
The average cost of acquiring a new wireless subscriber in North America now exceeds $500. That figure, widely cited by GSMA and corroborated by analyst reports from Deloitte and McKinsey, has been climbing steadily for a decade. In mature markets where penetration hovers near saturation, the arithmetic is unforgiving: every customer who churns represents not merely lost revenue but a replacement cost that frequently exceeds a full year of that subscriber's margin contribution.
And yet the telecommunications industry's approach to understanding why customers leave, what they actually want from pricing plans, and how they feel about the 5G transition has remained remarkably static. Large-scale quantitative surveys, conducted quarterly at best, deliver findings that are already stale by the time they reach the product team. Focus groups offer depth but at glacial speed and ruinous cost. Net Promoter Score, that blunt instrument beloved of boardroom presentations, tells you that customers are unhappy without telling you why, or what to do about it.
This is the context in which synthetic research has begun to find its footing in telecommunications. The premise is straightforward: rather than waiting weeks or months to understand customer sentiment, test pricing hypotheses, or evaluate new bundling strategies, telecom operators can query AI-powered research panels that deliver qualitative depth in hours. The results correlate with traditional methods at rates that have satisfied even the more sceptical enterprise buyers.
Telstra, Australia's largest telecommunications company, is among those that have adopted this approach. As a customer of Simile, the enterprise arm of synthetic research platform FishDog, Telstra has integrated AI-generated consumer insights into workflows that previously depended on lengthy commissioned studies. They are not alone, but they are perhaps the most prominent example of a tier-one operator treating synthetic research as operational infrastructure rather than experimental novelty.
This article examines why telecommunications is proving to be a particularly fertile sector for synthetic research, what specific use cases are delivering value, and where the limitations remain. A bias disclosure: the author is a co-founder of FishDog. The analysis that follows attempts to be even-handed regardless, but readers should calibrate accordingly.
Why Telecoms Is Uniquely Suited to Synthetic Research
Not every industry benefits equally from synthetic research. Telecommunications, however, possesses a cluster of characteristics that make it an almost textbook fit.
Massive, heterogeneous customer bases. A mid-sized mobile operator might serve ten million subscribers spanning every demographic, income bracket, and usage pattern imaginable. Understanding this population through traditional qualitative research is, practically speaking, impossible. You can interview 40 people or 400, but you are still sampling a vanishingly small fraction of a customer base defined by its diversity. Synthetic panels, built from census-weighted demographic data and behavioural models, can replicate this heterogeneity at a scale that makes qualitative exploration of niche segments economically viable for the first time.
Complex, multi-dimensional pricing. Telecom pricing is not like pricing a tin of baked beans. A single mobile plan involves data allowances, voice minutes, roaming charges, handset subsidies, contract length, early termination fees, family discounts, loyalty credits, and an ever-expanding menu of bundled services. The combinatorial complexity is staggering. Testing customer reactions to pricing changes through traditional research means commissioning a study for each variant, waiting for results, and then commissioning another when the competitive landscape shifts mid-cycle. Synthetic research compresses this loop from weeks to hours.
High-frequency competitive dynamics. When a rival operator launches an aggressive new plan on a Tuesday morning, the product team needs to understand customer reaction by Thursday, not next quarter. Traditional research simply cannot operate at this cadence. Synthetic research can.
Regulatory and technological transitions. The ongoing 5G rollout, the sunsetting of 3G networks, evolving spectrum allocation policies, and shifting regulatory requirements around data privacy all create moments where customer understanding is critical and time-sensitive. Operators cannot afford to be slow learners during a technology transition that costs billions in infrastructure investment.
The churn imperative. In most industries, losing a customer is unfortunate. In telecommunications, where acquisition costs are extreme and lifetime value is the core unit of business planning, churn is existential. Even marginal improvements in churn prediction and prevention represent enormous financial returns. A 1% reduction in monthly churn for an operator with 10 million subscribers translates, over the course of a year, to roughly 120,000 retained customers. At $500 per acquisition avoided, that is $60 million in value.
Six Use Cases Where Synthetic Research Delivers
The theoretical case for synthetic research in telecoms is compelling. The practical question is: what, specifically, are operators using it for? Six use cases have emerged as particularly productive.
1. Churn Driver Analysis
The conventional approach to churn analysis is heavily quantitative: logistic regression models fed by usage data, billing history, and customer service interaction logs. These models are good at identifying who is likely to churn but poor at explaining why. The "why" matters enormously, because it determines what intervention might actually work.
Synthetic research fills this gap. By recruiting panels filtered to match the demographic and behavioural profile of high-churn segments, operators can ask the qualitative questions that usage data cannot answer. What drove the decision to consider switching? Was it price, network quality, a specific service failure, or simply the expiry of a promotional rate? What would have changed their mind?
The responses are not predictions based on historical patterns. They are articulated motivations, expressed in the language customers actually use, which makes them directly actionable for retention teams and marketing copywriters alike.
2. Pricing Sensitivity and Plan Design
Pricing research in telecoms has traditionally relied on conjoint analysis, Van Westendorp surveys, or the Gabor-Granger method. These are rigorous but slow and expensive. They also tend to test pricing in isolation from the competitive context in which customers actually make decisions.
Synthetic research allows operators to test pricing scenarios iteratively. Does a $5 increase in the base rate matter more than the removal of a data cap? Would customers accept a higher price if 5G access were included? How does the framing of the price change affect perception: is "unlimited data for $10 more" received differently from "premium tier with unlimited data at $65"?
These questions can be posed to synthetic panels segmented by age, income, current plan type, or any other relevant variable. The speed of iteration means that product teams can test dozens of pricing configurations in the time it would previously have taken to test one.
3. New Plan and Bundle Testing
Bundling, the packaging of mobile, broadband, television, and increasingly smart home services into combined offerings, is a strategic priority for most large operators. The logic is sound: bundled customers churn at significantly lower rates than single-service subscribers. But designing bundles that customers actually want, rather than bundles that look tidy on a product roadmap, requires understanding which service combinations create genuine perceived value.
Synthetic research enables rapid concept testing of bundle configurations. Which combination of services would make you consider switching provider? What would you be willing to pay for a bundle that included mobile, fibre broadband, and streaming? Is a smart home add-on appealing or does it feel like unnecessary complexity?
The ability to test these propositions across different customer segments, without the lead time and cost of traditional concept testing, has proven particularly valuable for operators entering adjacent markets such as home security or content streaming.
4. Customer Satisfaction Deep Dives
NPS surveys tell you the score. They do not tell you the story. When an operator's NPS drops three points in a quarter, the executive team wants to know what happened, not merely that something did.
Synthetic panels, filtered to match the demographic profile of detractors, can explore the experiential drivers of dissatisfaction with a depth that standardised surveys cannot achieve. Was it a billing dispute? A network outage? A frustrating chatbot interaction? The perception that the operator's advertising promises what its network cannot deliver?
One advantage of synthetic research in this context is the absence of survey fatigue. Real customers, already irritated with their provider, are unlikely to spend 20 minutes explaining their frustrations in a satisfaction survey. Synthetic personas, constructed to reflect the attitudes and circumstances of those customers, will.
5. 5G Adoption and Migration Research
The 5G transition presents operators with a peculiar marketing challenge. The technology is genuinely transformative for certain use cases, but the average consumer, who uses their phone primarily for social media, messaging, and streaming, may struggle to perceive the benefit. "Faster downloads" is a difficult proposition to sell when current 4G speeds are already adequate for most daily tasks.
Understanding what 5G messaging resonates with different customer segments is critical to driving adoption and justifying the enormous capital expenditure involved. Synthetic research can test messaging angles rapidly: does "future-proof your connectivity" work better than "no more buffering at concerts"? Does the business customer respond to latency improvements while the consumer cares about coverage?
This is not merely a marketing exercise. The speed of 5G adoption directly affects network planning, spectrum utilisation, and the business case for infrastructure investment. Getting the messaging right has capital allocation implications.
6. Bundling Optimisation and Cross-Sell Strategy
Distinct from the initial design of bundles, ongoing optimisation of cross-sell and upsell strategies requires continuous insight into customer preferences. Which existing single-service customers are most receptive to bundling? What is the right moment in the customer lifecycle to make the offer? How should the value proposition differ for a customer who has been with the operator for six months versus six years?
Synthetic research panels segmented by tenure, current service portfolio, and demographic profile can inform these strategies with a granularity that broad-brush market research cannot. The iterative nature of the tool means that as the competitive landscape shifts, so can the research, without commissioning a new study each time.
The Speed Advantage, Quantified
It is worth pausing on the practical implications of speed, because this is where synthetic research most dramatically departs from the status quo.
A traditional qualitative research project in telecommunications follows a well-worn path: scope definition (one to two weeks), vendor selection and briefing (one to two weeks), fieldwork and recruitment (two to four weeks), analysis and reporting (two to three weeks). End to end, a competent agency will deliver findings in six to twelve weeks. A rushed project might manage four.
Synthetic research, using a platform such as FishDog, compresses this to hours. A study can be designed, fielded, and analysed in a single working day. This is not merely a convenience. It changes the category of decisions that research can inform.
At traditional research timelines, only strategic decisions justify the investment: annual pricing reviews, major product launches, brand repositioning exercises. Tactical decisions, the ones that arise weekly and demand answers in days, proceed on intuition and internal opinion.
Synthetic research makes tactical research economically and temporally viable. Should we match the competitor's new plan? What retention offer will resonate with this segment? How should we message the network maintenance scheduled for next month? These questions can now be answered with data rather than instinct.
The effect is cumulative. An operator that makes 50 small decisions per quarter informed by research rather than guesswork will, over time, outperform one that reserves research for quarterly strategy reviews. Compounding marginal gains, as the cycling world discovered some years ago, can be transformative.
What Telstra's Adoption Signals to the Market
Telstra's use of Simile, FishDog's enterprise offering, is worth examining not because it is the only example but because of what it signals about market maturity.
Telstra is not a startup experimenting with novel tools. It is a $23 billion revenue operator serving over 22 million customers across mobile, fixed-line, broadband, and enterprise services. Its adoption of synthetic research represents a considered judgement by a procurement and insights function that has no shortage of traditional research vendors competing for its budget.
What drives a company of this scale to adopt synthetic research? Three factors, based on publicly available information and broader industry conversations.
First, speed to insight. Telstra operates in a market where Optus and TPG Telecom compete aggressively on price and bundling. The ability to understand customer reactions to competitive moves in hours rather than weeks has direct commercial value.
Second, cost efficiency at scale. When you need to understand customer sentiment across dozens of segments, regions, and product lines, traditional research costs scale linearly. Synthetic research costs do not. Running 20 studies costs modestly more than running one.
Third, integration with existing workflows. Synthetic research platforms increasingly offer API access that allows insights to feed directly into product management, pricing, and marketing systems. This is not a separate "research project" that lives in a PowerPoint deck. It is an input to operational decision-making.
The broader signal is that synthetic research is transitioning from "interesting innovation" to "standard tool in the insights stack." When tier-one operators adopt, mid-market operators take note. The diffusion pattern is familiar from every enterprise technology cycle.
For those interested in how independent reviewers have assessed the platform, the Simile reviews provide additional perspective.
Limitations and Honest Caveats
No technology solves every problem, and intellectual honesty requires acknowledging where synthetic research falls short in the telecommunications context.
Network experience is physical. Synthetic personas can articulate attitudes toward network quality, but they cannot replicate the visceral frustration of a dropped call during a job interview or a video stream that buffers during a football match. Research into network quality perception benefits from being paired with actual network performance data. Synthetic research provides the "what do customers feel" layer; it does not replace the "what is the network actually doing" layer.
Extreme niche segments. While synthetic panels handle demographic breadth well, very narrow segments, such as enterprise customers with specific vertical industry requirements or users of highly specialised IoT applications, may push the boundaries of what synthetic personas can credibly represent. The more unusual the customer profile, the more caution is warranted in interpreting results.
Regulatory and contractual nuance. Telecommunications is a heavily regulated industry, and customer attitudes are shaped by regulatory frameworks that vary by jurisdiction. A synthetic panel filtered to Australian consumers will reflect Australian attitudes, but the specifics of ACMA regulations, local competition dynamics, and market-specific promotional norms may be imperfectly captured.
Behavioural prediction versus attitudinal insight. Synthetic research excels at understanding how people think and feel about telecoms products and services. It is less reliable as a predictor of specific behaviours, particularly switching behaviour, which is often determined by inertia, contractual lock-in, and the hassle factor rather than stated preferences. Operators should treat synthetic insights as one input alongside behavioural analytics, not as a replacement for them.
These limitations are real but they are also, for the most part, the same limitations that apply to traditional qualitative research. The question is not whether synthetic research is perfect. It is whether it is better than the alternative, which in many telecom contexts is no qualitative research at all.
The Pricing Research Connection
For operators specifically interested in how synthetic research applies to the pricing dimension of their business, a more detailed treatment is available in our companion article on AI-powered pricing research. Pricing is arguably the highest-stakes application of synthetic research in telecommunications, given the direct revenue implications of plan design and the competitive sensitivity of pricing moves. The techniques described there, including iterative price sensitivity testing and value-based segmentation, apply directly to the telecom context discussed here.
Where This Goes Next
The trajectory of synthetic research adoption in telecommunications is likely to follow a pattern familiar from other enterprise technology categories. Early adopters, the Telstras and their peers, are using it primarily for discrete research projects: a pricing study here, a churn analysis there. The next phase involves integration into continuous insight programmes, where synthetic research runs alongside quantitative analytics as a persistent layer of customer understanding.
Beyond that, the more speculative but plausible future involves synthetic research informing automated decision systems. An operator's pricing engine, for instance, might automatically test proposed adjustments against a synthetic panel before implementing them, creating a feedback loop that operates at machine speed rather than human speed.
Whether that future arrives in two years or ten is uncertain. What is clear is that the gap between how quickly telecom markets move and how slowly traditional research delivers has become untenable. Operators that close this gap, by whatever means, will make better decisions. Those that do not will continue to discover, six weeks after the fact, what their customers were trying to tell them.
Phillip Gales is a co-founder of FishDog. He has a financial interest in the platform discussed in this article and has attempted to present the analysis fairly despite this conflict. Readers should consider this when evaluating the claims made herein.