The Consent Ledger: Velti's Second Act Runs Through the Telecom Customer

Summary

• The company examined here is the current GreekVELTI SINGLE MEMBER S.A. Software Products and Services, publicly represented in English as Velti SA Software Products and Services. It must not be treated as interchangeable with the former Velti plc, incorporated in Jersey and listed on Nasdaq, nor attributed the scale, finances or ownership history of that group.
• Velti's current offering combines an omnichannel loyalty and engagement platform, Sympan, with Daedalus, a predictive customer value management engine. The visible workflow goes from subscriber data ingestion to profiling, offer selection, campaign delivery, gamification, rewards, measurement and managed operational support.
• Velti's business value cannot be judged solely on announced campaign volume or conversion rates. A buyer needs evidence of incremental revenue after incentives and channel costs, defensible consent and suppression controls, model monitoring, operational resilience and a viable exit path.
• The 2013 crisis is relevant because it raises the evidentiary bar for continuity and vendor assurance. It does not prove the current Greek company is in trouble and cannot substitute for current disclosure of financials, ownership, security or customer concentration.

A reload, many dependencies

Consider a prepaid subscriber adding credit to their mobile account. Seconds later, the subscriber may receive a reward, a data plan suggestion, a game invitation or an invitation to switch to another tariff. To the recipient, it is a single message. To the operator, it is the end of a chain that may include an account event, an eligibility rule, a consent record, a customer profile, an offer catalogue, a prediction, a campaign priority decision, a channel gateway, a reward obligation and a measurement framework.

This chain is Velti's true product surface. The company'sDaedalus pagestates that its software combines near-real-time contextual data from multiple sources into customer profiles, evaluates alternative journeys and recommends an offer, message and moment. Its broadercustomer retention offeringadds programme design, campaign operations, rewards, localisation and account management. Itsgamification productexposes integration via web view, REST API and SDK, with channels that may include SMS, email, app push, web push and WhatsApp.

The seemingly simple top-up message therefore embodies two different promises. The first is computational: Velti can rank actions better than a manual campaign team or a static set of rules. The second is institutional: the data was lawfully available, the recommendation was commercially sensible, the communication respected the subscriber's choices, the reward could be honoured and the outcome could be measured without misleading the operator.

The second promise is harder. A model can improve its observed response rate while eroding margin through discounts, annoying a profitable control population, exploiting a biased dataset, repeatedly targeting the easiest responders or displacing sales that would have happened anyway. A technically successful campaign can still be a governance failure. Conversely, a cautious campaign can create lasting value even if its immediate click-through or conversion rate is lower.

The investment case for Velti from a client perspective is therefore not 'AI sends better messages'. Rather, it is that a specialised vendor can execute a reliable decision and execution loop in a sensitive telecom environment. The procurement question is whether the vendor can prove that loop is lawful, incremental, controllable and reversible.

Start with the Greek company, not the former ticker

The current legal and operational identity can be delimited more precisely than the complicated brand history suggests. Velti'sprivacy policynames 'Velti SA Software Products and Services' as a company registered and operating under Greek law. It gives a registered address at 44 Kifissias Avenue in Athens and describes the company as part of a Velti group of companies. The site's contact footer uses an operating address at 62 Kifissias Avenue and Premetis in Maroussi.

The company's ownfinancial statements pageprovides the strongest link to the attributed single entity. It identifiesVELTI ΜΟΝΟΠΡΟΣΩΠΗ Α.Ε.—Velti Single Member S.A.—at the Maroussi address and gives the Greek General Commercial Registry number 003784801000. Anindividual 2020 financial statementuses the shorter nameVELTI Α.Ε., the same registry number without leading zeros, and the Greek tax number 099756001. These are documents published by the company, not a full independent register extract, but together they support treating the current operational subject as the Greek single-member software products and services company named in the attribution.

They do not disclose a full current ownership chain. 'Single member' tells the reader that the Greek company has a single shareholder; it does not identify that shareholder in the summaries cited. The reference to a group in the privacy policy also does not establish which entities are in that group, where ultimate control lies, or how post-2013 transactions affected ownership. These are unresolved questions, not gaps to fill with the former public company's history.

Current operational activity is visible beyond the company's static text. In March 2024, the Dutch payments and fraud prevention company Alphacomm published its own account of apartnership with Velti, corroborating Velti's announcement and identifying Sympan as the engagement layer. Velti has also continued to publicly associate its platform with the Dutch operator now called Odido, including ashortlist for a 2025 customer experience award. Odido independently confirms thatT-Mobile Netherlands became Odido in September 2023, which helps explain the client name change without implying that every historical T-Mobile claim remains current.

The clear boundary is therefore: the article concerns the current Greek Velti company and its current product surface. The former Velti plc is relevant only where historical documents shed light on continuity risk, contractual diligence and the burden of proof.

The 2013 break

The failure of the former listed group was not a simple rebrand. It was a break in capital structure, asset ownership and public company continuity.

Velti plc's2012 Form 20-Fdescribed a parent company incorporated in Jersey and listed Velti S.A. in Greece as a 100% subsidiary at that time. The filing reported $270.3 million in 2012 revenue for the consolidated public group, but also carried a going-concern warning, revealed a material weakness in the financial close process and described heavy dependence on receivables collection. Year-end trade receivables stood at $150.1 million and accrued contract receivables at $133.0 million before the crisis fully broke. These figures belong to the former Velti plc consolidated scope. They are not current figures for the Greek company.

By June 2013, the deterioration was stark. The former group'ssecond-quarter filingshowed cash of $19.4 million, while allowances for doubtful accounts on trade receivables had risen to $44.6 million and allowances on accrued contract receivables to $77.8 million. A contemporaryBusiness Insider reconstructionfocused on the mismatch between reported revenue, slow collections and cash. The primary filings are the strongest evidence; the article is useful as a document of how investors and vendors interpreted the warning signs at the time.

On 4 November 2013, Velti Inc., a US subsidiary, filed for Chapter 11 protection to organise a sale. A Bloomberg report carried byeKathimeriniindicated that a subsidiary of GSO Capital Partners, Blackstone's credit arm, would act as a stalking horse bidder and provide debtor-in-possession financing. It also reported that operations outside the US, including Greece, had not sought bankruptcy protection. This distinction is important: the US bankruptcy did not mean that every Velti legal entity entered insolvency proceedings.

Nor does it prove an uninterrupted ownership path to the current company. The subsequent sale covered specific mobile marketing operations and assets. In November, Velti plc filed aForm 6-K announcing voluntary delisting from Nasdaq, citing non-compliance with the minimum bid price, questions about listing eligibility after the proposed sale of the business, the Chapter 11 proceedings for US subsidiaries and the Chapter 7 filing of Mobclix. A December company announcement reported court approval of thesale of the mobile marketing business to subsidiaries of GSO.

The defensible conclusion is narrow. The Greek operation existed within the former listed group, the Greek activities were described as continuing during the US process, and a current Greek Velti entity now markets related mobile engagement capabilities. The cited public record does not reconstruct every intervening transfer of shares, intellectual property, contracts or control. Any client relying on continuity should demand the actual corporate chain, rights in the relevant software and the contracting entity's authority to license and support it.

This history also offers a business lesson. Enterprise software clients are exposed not only to product failure but also to vendor finances, receivable quality, lender control and asset sale mechanisms. A platform may remain technically available while ownership of the code, personnel, client contracts or support obligations changes. The correct response is not to assume history will repeat itself. It is to make continuity provisions testable before dependency deepens.

What Velti sells now

Velti presents itself as a mobile marketing, customer loyalty and content solutions company. Its'about' pagedescribes data science, machine learning and artificial intelligence as the core of an end-to-end platform serving mobile operators, brands and media groups. The site portfolio can be understood as four connected layers.

First, customer engagement and loyalty. Sympan is described in Velti'sloyalty presentation transcriptas an omnichannel platform capable of running loyalty, gamification and customer value management programmes in a single installation. It supports points programmes, missions, tiers, badges, campaign communication and third-party channels. This is the operational shell in which campaigns and customer journeys are configured.

Second, automated decision-making. Daedalus is the recommendation and optimisation engine. Velti claims it builds customer profiles from multiple data sources, evaluates alternative actions and continuously incorporates campaign responses. A company case study page goes further, naming supervised tree-based methods including GBM, distributed random forest, XGBoost and LightGBM, as well as stacked ensembles and natural language processing. These are Velti'stechnical claims, not independent architectural verification, but they indicate that 'AI' here refers substantively to predictive ranking, segmentation and offer optimisation rather than a general conversational model.

Third, execution and experience. Velti offers campaign delivery over messaging and digital channels, customer cockpits, reward catalogues, instant gratifications, games, quizzes, missions and promotional journeys. The gamification layer is not decorative. It changes the economic mechanism: an operator can use rewards to incentivise top-ups, app adoption, self-service, product trials or repeat engagement, while collecting response data that feeds subsequent targeting.

Fourth, managed service. Velti repeatedly states that its teams can handle account management, project management, copywriting, campaign planning, rewards, localisation, maintenance and measurement. This makes the offering partly software and partly outsourced operational capability. An operator is not just buying a model endpoint; it can delegate a significant portion of the campaign factory.

The public site claims more than 300 clients or projects in more than 70 countries and displays logos associated with T-Mobile Netherlands, Vodafone Germany and Otelo. These figures and logos appear on Velti'sclients pageand homepage, but the site does not provide a verified current client list, revenue segmentation or concentration table. They must be treated as company claims. The ongoing Odido documents and Alphacomm partnership provide useful corroboration that there is living operational activity, but they do not validate the full claimed scale.

The workflow Velti must make reliable

A customer value management platform is best evaluated as a workflow rather than a feature list. For a telecom operator, the minimal end-to-end sequence looks like this.

1. Establish authorised use.The operator must know which subscriber data can be used for which purpose and on which channel. An active service relationship does not automatically make every form of profiling or promotional communication permissible. Consent, legitimate interest analysis, contractual necessity, local e-marketing rules and customer objections can produce different eligibility states.

2. Map and ingest source data.Relevant inputs may include plan, tenure, top-up behaviour, usage, service events, campaign history, channel engagement, reward history and customer service signals. Velti claims Daedalus uses near-real-time contextual data from multiple sources. The hard work is not just moving fields. It is defining authoritative sources, timestamps, missing value behaviour, identity resolution, permitted lookback periods and what happens when systems disagree.

3. Build features and segments.Raw events become model inputs: top-up frequency, declining usage, churn likelihood, likely plan interest, response propensity. A '360-degree profile' is a marketing phrase until the buyer knows which attributes are included, how fresh they are, whether they are directly observed or inferred, and which are excluded for legal or ethical reasons.

4. Build the action set.An economically sensible next action by the model depends on receiving a valid portfolio of offers, messages, channels, rewards and timing options. Each candidate needs eligibility rules, inventory or reward availability, margin information, expiration dates, frequency caps and customer service constraints. A good prediction on a bad offer catalogue still produces a bad campaign.

5. Score and constrain.Daedalus is presented as ranking possible journeys or campaigns for each customer. In production, a score should not be the final authority. Business rules may need to suppress vulnerable groups, honour contact limits, prevent conflicting campaigns, reserve a control population, block unprofitable rewards and route unusual cases to human approval.

6. Execute via channels.Velti's public material describes SMS, email, app and web push notifications, WhatsApp, web views, APIs and SDKs. Each channel introduces dependencies: sender identity, template approval, delivery receipts, application versions, gateway throughput, quiet hours, language rendering and unsubscribe propagation. The decision engine may be available while the last-mile channel fails.

7. Keep the promise.A reward campaign creates obligations. Points must be credited correctly; vouchers must be valid; physical prizes require inventory and delivery; tariff benefits must be activated; winners need support. A reward failure turns a personalisation exercise into a trust incident.

8. Measure and feed back.Responses and non-responses flow back into the system. The operator must distinguish delivery from exposure, exposure from action, and action from incremental value. If the feedback loop records only successful contacts or does not account for competing campaigns, it can train the model on a distorted world.

This workflow explains why implementation quality dominates promotional material quality. Velti can provide software, models and managed operations, but the operator provides the legal authority, source system truth, offer economics and much of the customer relationship. The outcome belongs to the joint system.

Architecture visible through the gaps

Velti does not publish a full current reference architecture, but its public material exposes useful gaps.

The live gamification page states that integrations can range from a simple web view to REST API and SDK connections. It also lists several outbound channels. The loyalty material states that multiple programmes can run in a single installation and that third-party channels can be connected. A currentsoftware engineer job postingasks for skills in Java/J2EE, Spring, JPA or Hibernate, SOAP and REST web services, SQL or another database management system, and experience building maintainable components in a large multi-process system. It also references functional testing on a full installation that reproduces real-world usage and resolving customer software issues.

Together, these details suggest a conventional enterprise platform with substantial integration and deployment engineering around it. They support the existence of server-side transactional components, database-backed applications, service interfaces and customer-specific installations. They do not prove that every current Sympan or Daedalus component uses exactly this stack, nor do they reveal the hosting model, cloud providers, multi-tenancy design, streaming technology, feature store, model registry, message bus or observability system.

This distinction matters because the marketing phrase 'minimal integration' can mean several things. It can mean that Velti can sit alongside existing business support systems rather than replace them. It can mean that prebuilt adapters reduce the work. It can also mean that the first use case requires only a narrow batch flow while later real-time use cases require much deeper coupling. A buyer should ask for interface-by-interface estimates rather than accepting an aggregated description.

The likely operating model is a hybrid boundary. The operator retains the systems of record (billing, charging, customer management, consent, product catalogue, channels) while Velti receives selected data, makes or supports campaign decisions, and returns actions or orchestrates delivery. This is an inference from the product pages, not a disclosed universal topology. Some deployments may be more self-contained or more heavily managed.

The architecture has two control planes. The technical control plane moves data and actions. The business control plane decides objectives, offers, budgets, exclusions, rewards and escalations. Procurement often examines the first and under-specifies the second. Yet a campaign can fail because an expired offer remained selectable, a reward budget was exhausted, a customer service suppression arrived late, or two departments targeted the same subscriber. These are orchestration failures even when the model and API are sound.

The product is improvement, not outcome

Velti publishes striking performance claims. The Daedalus page states that campaigns achieve at least a 50% conversion rate improvement over campaigns sent without intelligence, use 25% fewer SMS sends and outperform campaigns deployed by business experts. A company case study for T-Mobile Netherlands reports increases of up to 441% for a particular cross-sell campaign. Velti has also continued to publicise its recognition with Odido, including aresult at the Engage Awards in 2024.

These are significant signals of product use and client participation. They are not a substitute for a reproducible lift study. The figures are published by the vendor, may refer to selected campaigns and are not publicly accompanied by full sample definitions, confidence intervals, offer economics or a complete account of control group construction. An award validates that judges found the entry compelling; it does not verify every financial assumption.

The first procurement discipline is to define the denominator. 'Conversion rate improvement' could mean a relative lift from 1% to 1.5%, an absolute increase of 50 percentage points, or something else. It might count activations, purchases, top-ups, clicks or retained subscribers. It may be measured across delivered messages, eligible customers or the entire targeted population. Each tells a different economic story.

The second is incrementality. A model can identify customers who were already likely to buy. This produces high observed conversion and little causal value. A credible test needs a contemporaneous control population, consistent eligibility, controlled offer exposure and enough time to observe cannibalisation or deferred churn. Competitors now make this an explicit selling point:Comviva's MobiLytix Marketing Studioadvertises control groups and causal measurement, whileOptimove's orchestration platformdescribes industrialised control groups and incremental growth measurement. Velti should be evaluated to the same standard regardless of method.

Third is net value. The correct equation includes incremental gross margin, not just revenue, minus reward cost, messaging costs, platform fees, managed service cost, fraud and abuse, customer service contacts, offer subsidy and any damage to other campaigns. A data plan upsell may appear successful while merely accelerating demand or moving a customer from a more profitable product.

Fourth is sustainability. A campaign that repeatedly targets the most responsive users can exhaust them. A model that learns from discount offers can teach the customer to wait for discounts. A churn model can become self-fulfilling if only predicted at-risk customers receive attractive retention treatment. The buyer needs evidence at the cohort level over time, not just campaign screenshots.

Velti's proposition becomes strongest when it can show that automation reduces unnecessary contacts while increasing net value. Its claim of fewer SMS sends moves in that direction. The business proof would be an agreed measurement protocol that withstands changes in season, product mix, customer base and campaign team.

Permission is a state machine

Velti explicitly sellspermission campaignsdesigned to obtain marketing opt-ins before sending messages. This is strategically sensible: the same company that optimises targeting also offers a mechanism to gather permission. But permission is not a single checkbox that can be detached from subsequent data use.

Under the EU General Data Protection Regulation (GDPR), individuals have the right to entity to direct marketing, including profiling related to it.Article 21 of the GDPRrequires that this right be brought explicitly to the individual's attention. TheePrivacy Directivesets additional rules for unsolicited electronic communications and generally requires prior consent for certain forms of automated electronic marketing, subject to national implementation and the existing customer exception. TheEuropean Data Protection Board's guidance on profilingtreats automated assessment of preferences and behaviour as a separate governance concern.

For an operator, the usable consent record should therefore include at least the source, wording, purpose, channel, timestamp, jurisdiction, customer identifier and proof of action. It should also record withdrawal, objection, expiry if any and the systems to which the suppression was propagated. If an app preference changes but the campaign platform copy is stale, the operator may continue to target a person who has unsubscribed.

Purpose matters as much as channel. Consenting to receive a top-up reminder does not automatically authorise every loyalty, cross-sell or third-party promotion. Permission to use service data for network operations does not automatically settle its use for marketing prediction. A 'customer profile' assembled from browsing, content engagement, usage and transactional data must be evaluated attribute by attribute and purpose by purpose.

Velti's website privacy policy contains useful commitments on security and retention, but it primarily governs website visitors, business contacts and applicants. It is not the data processing agreement for an operator deployment. The controller-processor split, hosting location, sub-processor chain, international transfers, retention, data subject rights assistance and incident timelines must be established in the client contract. Depending on the programme, roles may vary; it would be dangerous to assume from the public site that Velti is always merely a processor or that the operator can delegate liability.

Permission also has a product dimension. A gamified opt-in can improve participation, but the reward must not obscure the choice or create a misleading impression about what data will be used. The operator should test understanding, not just completion. Trust is higher when the customer understands the exchange: what will be observed, what benefit is offered, how often contact may occur and how to stop it.

The central design principle is simple: offer eligibility and communication eligibility are different states. Daedalus may predict that an offer is attractive while the permission system correctly blocks the message. A reliable platform must maintain this separation.

Self-learning needs brakes

Velti describes Daedalus as adaptive, self-training or self-learning. These terms promise improvement, but they also widen the governance surface. A model that changes with new outcomes can drift because customer behaviour changes, the offer portfolio changes, a channel becomes less reliable, or the system's own past actions reshape the data.

Customer value models are allocation systems. They decide who receives attention, discounts, rewards and opportunities. Even when a marketing recommendation does not produce the legal or similarly significant effect targeted by Article 22 of the GDPR, it can still create unfair or commercially dangerous patterns. High-value customers may receive richer benefits than low-income prepaid customers; a model may infer vulnerability from top-up behaviour; frequent responders may be over-contacted; groups with sparse data may receive inferior offers; a profitable but inappropriate action may dominate because the objective function is too narrow.

The US National Institute of Standards and Technology (NIST)Artificial Intelligence Risk Management Frameworkprovides a useful voluntary structure: govern, map, measure and manage. Itscore guidancecalls for pre-deployment and operational testing, documenting uncertainty, monitoring third-party resources and preparing override, incident, recovery and decommissioning mechanisms. These are appropriate procurement expectations, even for a traditional predictive model.

For Velti, an adequate governance package should address practical questions:

• Which models are active for which campaigns, and who approved each objective?
• What features are used, where do they come from, and what prohibited or sensitive proxies have been tested?
• How are training, validation and production populations separated?
• What baseline, calibration, accuracy, lift and fairness metrics are monitored?
• How quickly is drift detected after a tariff, brand, channel or customer-base change?
• Can business users impose frequency caps, protected group rules, margin floors and mandatory exclusions?
• Is there a human-readable reason for a recommendation, and is that reason faithful enough for operational review?
• Can a model version be rolled back without disrupting unrelated campaigns?
• Are negative outcomes, complaints and non-responses fed back, or only conversions?
• Who can stop automated selection, and how is that action recorded?

Vendor and buyer must also distinguish model error from policy error. If the model accurately predicts that an aggressive discount will convert but the discount destroys margin, the algorithm may be statistically correct and commercially wrong. If the model selects a prohibited customer because the consent flow was late, the failure is one of integration and governance, not necessarily of prediction. Incident reviews need a taxonomy that identifies which layer failed.

Human oversight must be meaningful, not ceremonial. A marketing manager who can only approve hundreds of automatically generated campaigns without sufficient context is not exercising control. Oversight requires manageable exceptions, clear constraints, sampling review and authority to modify or stop the system.

Managed service is both value and dependency

Velti's strongest differentiation may be operational rather than algorithmic. Its public material says teams can handle projects, copywriting, campaigns, rewards and localisation, from setup to maintenance. For an operator with limited data science or campaign operations capacity, this can shorten time to value. It also places vendor staff inside the rhythms of business planning.

The client nevertheless retains substantial work.

The operator owns the accuracy and authorised use of source data. It must maintain product and tariff truth, fund incentives, define acceptable margin, approve communications, coordinate channels, train customer service teams and manage regulatory liability. It must ensure that a recommendation does not conflict with credit, fraud, vulnerable customer, roaming, network or service assurance policies. If a campaign promises a benefit that the billing system fails to activate, the operator (not an abstract AI engine) faces the customer.

Implementation should start with a limited, measurable use case rather than a broad '360-degree profile' ambition. A sensible first deployment might use a small set of well-understood events, a narrow offer catalogue, explicit exclusions and a randomised control group. This allows the parties to validate identity matching, consent propagation, latency, delivery, reward execution and outcome capture before adding more data or autonomy.

A second step can broaden orchestration: competing campaigns, multiple channels, dynamic rewards and more frequent decisions. This is where organisational conflict emerges. Loyalty, acquisition, digital, prepaid, postpaid, brand and customer service teams can all claim the same customer moment. The platform needs a documented hierarchy for priority and arbitration.

A third step is ongoing operation. It requires version management, model review, campaign calendars, data quality alerts, incident response, support escalation and business performance review. Velti's job posting references agile development, testing on complete installation and resolving customer issues, suggesting that customer-specific operations remain part of the engineering model.

A useful contract would include a responsibility matrix covering every critical activity: source data, legal basis, feature approval, campaign design, model validation, channel availability, reward funding, execution, customer support, rights requests, breach notification, regulatory response, backup, recovery and exit. 'Fully managed' should never mean 'responsibility transferred'.

Business model and pricing logic

Velti does not publish a price list for Sympan, Daedalus or its managed programmes. The customer loyalty page describes tailored packages and performance-based programmes. This points to negotiated enterprise contracts rather than self-service software pricing.

The likely cost structure (an inference, not a disclosed Velti rate) has several components. There may be implementation and integration fees; recurring platform or licence fees; managed service fees for campaign and account work; usage fees related to profiles, decisions, campaigns or interactions; and transit costs for messages, rewards, content, payments or fulfilment partners. A performance component may be tied to agreed outcomes. Buyers should request the actual schedule rather than treating a single component as the total price.

Each model creates incentives. Per-message fees reward volume unless contact policy constrains it. A share of incremental revenue appears aligned but depends on a credible counterfactual and agreed treatment of cannibalisation, discounts and seasonality. Fixed managed service fees may encourage efficiency but underresource unusual campaigns. Per-profile fees can make unused data costly. The commercial design should reinforce fewer, better, provably incremental actions.

The lack of current public financial disclosure is significant for vendor assessment. Velti's website only publishes individual Greek financials up to 2020. The 2020 filing reports net turnover of €4.48 million and a loss after tax of €1.21 million, with 2019 comparatives of €3.82 million and a loss of €1.01 million. It also contains substantial balance sheet amounts that cannot be responsibly interpreted without more complete notes and later periods. These figures are stale, individual and unsuitable for estimating current group revenue, profitability, cash or creditworthiness.

They do, however, show why a buyer should ask for up-to-date information. The website claims of hundreds of clients and projects cannot be publicly reconciled with current revenue, contract value or customer concentration. A long-standing relationship with Odido can be both a powerful reference and a concentration risk; the public record does not reveal its economic weight.

Enterprise software for telecoms can have lumpy economics. Sales cycles are long, integration is expensive, and a few operator programmes can require dedicated staff. Managed service increases recurring revenue but also labour intensity. International deployments add local content, regulation, support hours, rewards and channel complexity. A buyer assessing continuity should therefore examine order book, renewal pattern, customer concentration, cash conversion, insurance, headcount and financial capacity to support a multi-year deployment.

The 2013 history makes this diligence more relevant, not determinative. The former group's crisis involved slow receivables collection, covenant pressure and a growing gap between recorded revenue and liquidity. A modern contract can reduce exposure through phased acceptance, prepayment limits, escrow where applicable, transition assistance, assignability checks and clear treatment of customer data and configurations in the event of ownership change.

Where switching costs accumulate

Velti's platform may initially sit alongside existing systems, but switching costs grow with successful use.

The first layer is integration. Source mappings, identity resolution, consent flows, product catalogues, event schemas, channel adapters and delivery acknowledgements take time to stabilise. Documentation often lags behind working configuration.

The second is campaign knowledge. Offer rules, exclusions, calendars, copy variants, reward mechanisms, localisation and escalation routines encode years of business learning. Part of this knowledge resides in the software; another part resides in vendor staff and informal operational routines.

The third is model history. Training data, feature definitions, response labels, control outcomes, calibration and drift history make the recommendation system more operator-specific. Even if raw data belongs to the client, portability of derived features, model artefacts and decision logs may be contractually or technically limited.

The fourth is customer state. Points, tiers, missions, reward eligibility, pending prizes and campaign suppression cannot simply be abandoned. A loyalty migration must preserve balances, histories and customer expectations.

The fifth is organisational dependency. If Velti manages copywriting, campaign planning, reward management and maintenance, the operator may reduce its own capabilities. Replacing the platform then also means rebuilding a team or finding another managed operator.

An exit plan should be designed before launch. It should specify export formats and frequency for customer state, consent and suppression history, campaign definitions, decision logs, outcomes, reward liabilities and configuration. It should state which customer-specific features and model artefacts can be transferred, which remain Velti intellectual property, and what documentation is needed to reproduce business rules elsewhere. It should include a data deletion certification, sub-processor removal, transition support, API version notice and a parallel operation period.

The operator should test an export during the contract, not at termination. A file that technically exists may omit identifiers, relationships or timestamps needed for migration. The buyer should also retain an independent record of offers, eligibility and treatment assignments so that historical measurements can be verified after switching vendors.

Architectural modularity can reduce lock-in. An operator could keep consent, product eligibility and channel execution in its own systems while using Velti for decision and orchestration. Alternatively, it can accept deeper integration in exchange for faster managed operation. Neither choice is inherently superior. The important point is to price the exit and know which functions can continue if the recommendation engine, campaign layer or managed team becomes unavailable.

Security, compliance and resilience

Velti announced anISO 27001 certification in July 2019with a scope covering the management, design, development, installation and support of software products, software as a service and IT projects. Its currentsecurity policystates that the company operates an information security management system aligned with ISO 27001:2013, with risk assessment, access controls, incident logging, training and management review.

The privacy policy adds company claims regarding encryption, penetration testing, service provider cyber assessments, crisis exercises, infrastructure controls and data processing agreements with sub-processors. These are relevant commitments. They still require contractual and audit evidence.

By 2026, a buyer should ask for the current certificate, certificate number, certified legal entity, locations, scope, statement of applicability, surveillance status and transition to ISO/IEC 27001:2022. DQS explains that anISO 27001 certificate is generally valid for a maximum of three years, subject to surveillance and recertification. A web page referencing a 2019 award and the 2013 edition does not by itself prove current certification status or that every deployment environment is in scope.

The public documents do not disclose a SOC 2 report, current sub-processor list, detailed hosting and residency map, public service status history, standard service level agreement, recovery time and recovery point objectives, penetration test summary or product-specific incident record. This is an evidence gap, not proof that controls or documents do not exist. Telecom procurement should obtain them under confidentiality where appropriate.

The cited public record also does not establish a current Velti product breach or major outage. It would be wrong to infer perfect reliability from the absence of a documented event. Managed enterprise platforms often disclose incidents directly to clients rather than on a public status page. The test is whether the contract requires prompt notification, root cause analysis, corrective action, regulatory support and transparent service credits.

Supply chain risk is particularly important because Velti may sit between sensitive customer data and several fulfilment or reward partners. ENISA's analysis ofsoftware supply chain attackshighlights that clients can be compromised via vendors even when their own controls are strong. A buyer should map libraries, cloud services, messaging providers, application or support systems, analytics tools, fulfilment partners and vendor privileged access.

Minimum security diligence should cover encryption and key control; tenant isolation; privileged access approval; multi-factor authentication; logging; secure development; dependency analysis; secrets management; vulnerability and patch timelines; backups; disaster recovery; data masking in non-production; deletion; employee access; sub-processors; and independent testing. Because the platform can influence customer communications at scale, integrity is as important as confidentiality.

An attacker or erroneous deployment that alters campaign rules could send unauthorised messages, issue rewards, expose segmentation or damage the operator's brand without stealing a database.

Resilience testing should include degraded modes. Can the operator stop outbound decisions while keeping customer state data intact? Can static rules take over? Are duplicate rewards prevented after replay? What happens when consent, billing or product catalogue feeds are stale? Does the platform fail safe for marketing eligibility? Recovery procedures should be exercised with the operator, not just documented by the vendor.

Competition and substitutes

Velti competes in overlapping markets rather than a single clear category.

Telecom-specific software vendors offer customer value management, real-time decisioning, loyalty and campaign orchestration. Comviva markets a modular platform with unified customer profiles, next-best-action recommendations, policy controls and causal measurement. Amdocs integratesomnichannel engagement and next-best-action capabilitiesinto a much broader suite of telecom software and services. These vendors may benefit from existing relationships with operators' billing, catalogue and customer systems.

Horizontal marketing platforms are another substitute.Optimovecombines customer data, predictive segmentation, multichannel orchestration, APIs and managed services, and publicly explains pricing factors such as customer profiles, data volume and feature scope. It is therefore a relevant comparator when a buyer wants these functions outside a telecom-specific suite.

The third substitute is an operator-built stack. An operator can keep data in its own warehouse or data lake, develop churn and propensity models in-house, use a commercial campaign manager and deliver through existing channels. This offers control and may reduce vendor dependency, but requires data engineering, model operations, experimentation, campaign expertise and 24/7 support.

The fourth is a services-led arrangement using simpler tools. Not every campaign needs adaptive machine learning. Rules, segmentation and disciplined experimentation can outperform a complex platform when data is scarce, offers are few or governance is immature.

Velti's potential differentiation is the combination of telecom experience, managed campaign operation, loyalty, gamification, permission acquisition and predictive decisioning. The long Odido/T-Mobile Netherlands association suggests the company can stay embedded in an operator programme for many years. A small specialist can also adapt faster than a large suite vendor and provide more hands-on support.

Its potential disadvantages are the other side of this model: dependence on expert staff, less public financial and architectural disclosure, a seemingly smaller research and development base than large suite vendors, and the risk that client-specific work becomes hard to standardise. The public marketing does not establish how quickly the platform can onboard a new operator, how many releases it supports, or how support coverage varies by region.

A fair selection should therefore use a paid bake-off on the buyer's data. Vendors should receive the same eligible population, same offer set, same constraints and same historical window. They should be scored on incremental value, reduction of unnecessary contacts, explainability, campaign build time, data quality handling, integration effort, operational control and total cost—not just on a curated demonstration.

Twelve procurement tests

1. Contractual identity.Obtain a current extract from the Greek register, ultimate beneficial ownership information, group structure, signing authority, intellectual property and the exact entity responsible for service, data processing and support. Confirm what happens if the contract or software is assigned.

2. Client references.Request at least three current references for comparable scale and use case, including a recently completed implementation and a client that has renewed. Interview references on data work, launch delays, support, measurement disputes, outages and exit rights—not just campaign results.

3. Data and consent mapping.Trace every input field from source to feature to decision. Test withdrawal and objection propagation across all channels. Require a clear controller-processor matrix, processing purposes, residency, retention, sub-processors and deletion evidence.

4. Offline model review.Before production, reproduce training and validation results, inspect for leakage, assess stability by cohort and verify that excluded or sensitive attributes and proxies are treated as agreed. Compare Daedalus against a simple rule baseline and the operator's existing model.

5. Live causal trial.Run randomised control groups with pre-agreed metrics. Measure incremental gross margin after incentives, channel and service costs. Preserve treatment assignment and analyse complaints, unsubscribes and long-term behaviour in addition to conversion.

6. Guardrails and override.Demonstrate frequency caps, suppression, margin floors, offer eligibility, conflict resolution, protected group constraints, human approval, kill switch and version rollback. Intentionally introduce stale and conflicting data to observe failure behaviour.

7. Integration proof.Build the actual interfaces, not simulated endpoints. Measure latency, throughput, retries, duplicate handling, changes and observability. Identify which parts are web view, API, SDK, file transfer or manual operation.

8. Operational readiness.Define release windows, campaign approval, localisation, reward inventory, customer service scripts and escalation. Perform a full dress rehearsal from source event to honoured reward and reconciled financial outcome.

9. Security assurance.Verify current ISO certification and scope, request independent test evidence, review secure development and vulnerability processes, map privileged access and sub-processors, and test incident notification. Require fix timelines by severity.

10. Resilience.Agree availability metrics for each critical component rather than an aggregated SLA. Test backup restoration, regional or site failure, channel failure, stale consent flows, duplicate events and recovery without duplicate communications or rewards.

11. Commercial transparency.Model total cost over implementation, recurring platform, profiles or decisions, managed service, messages, rewards, partner fees, change requests and exit. If pricing is performance-based, define the counterfactual and audit rights.

12. Exit rehearsal.Export a representative tenant or programme during the contract. Rebuild campaign rules and customer state in a neutral environment, verify deletion and confirm transition staff. Make ongoing access to logs and measurement history part of termination assistance.

These tests are demanding because Velti's software is not a peripheral dashboard. In full deployment, it can become integral to how an operator allocates offers, contacts customers and interprets business performance.

What the evidence proves—and what it does not

The current evidence proves several useful things. A current operational Greek Velti company publicly identifies itself with a legal name close to the attributed entity and publishes the Greek single-member company form and registry number. It maintains a current product site, recruits software engineers, markets Sympan and Daedalus, and has recent external corroboration via Alphacomm. The T-Mobile Netherlands relationship has visible continuity into the Odido era, at least at the level of company and award documents.

The evidence also supports a functional product model. Velti describes customer data ingestion, profiles, predictive ranking, campaign automation, APIs, SDKs, multiple channels, loyalty mechanisms, rewards and managed services. Its technical recruitment material is consistent with an enterprise application and integration business.

The evidence does not independently verify the claimed client count, geographic scale, universal campaign improvement or current financial strength. It does not disclose current ownership, customer concentration, order book, pricing, cloud and multi-tenancy architecture, service levels, current certificate validity, model governance documentation or a comprehensive incident history. The public case studies are selective and largely vendor-hosted.

The former Velti plc records prove a severe corporate and financial break in 2013. They do not prove that the current Greek company inherited every asset, liability, client or governance problem. Nor do they justify attributing the former public group's revenue to the current entity. Historical continuity is a diligence question, not a shortcut to current assessment.

This evidence boundary should inform the tone. Velti is neither an unverified shell hiding behind a former name nor the uninterrupted continuation of a once-significant listed adtech group. It is a current Greek enterprise software and managed services operator whose most credible public case is a deeply embedded telecom customer engagement programme. The unanswered questions concern scale, governance and resilience, not the existence of a product surface.

Watchpoints

The first watchpoint is disclosure. Current audited or filed financial statements, an ownership chart and a clearer group structure would materially improve vendor continuity assessment. As would an updated public security page showing the current ISO edition, certificate scope and validity.

The second is independent performance evidence. A case study that publishes population definitions, control group design, incremental gross margin, confidence intervals, contact reduction and long-term churn would be more valuable than another headline conversion percentage. Continuing named references after renewals would help distinguish lasting value from a historically successful campaign.

The third is product governance. Buyers should watch for published information on model inventories, drift monitoring, explainability, human approval, feature restrictions and incident handling. As the company uses stronger 'AI' language, evidence of controlled automation should grow with it.

The fourth is the Odido relationship. Continued operation after the client's name change is a positive continuity signal. Any expansion, renewal, replacement or public service issue would be informative as the relationship appears central to Velti's visible evidence.

The fifth is platform modernisation. Public evidence of deployment options, API lifecycle, observability, data residency, cloud portability and independent resilience testing would clarify whether Sympan and Daedalus can compete with newer orchestration platforms while preserving the specialist managed-service advantage.

The sixth is commercial concentration. More named clients and recent deployments would reduce dependence on a small number of flagship references. Conversely, a shrinking set of public clients or a persistent absence of current filings would increase continuity questions.

Velti's second act will not be won by invoking its longevity or rhetorically distancing from 2013. It will be won one governed decision at a time. If the company can show that a subscriber was eligible, authorised, treated fairly, contacted through a resilient system and moved in a way that created genuine incremental value, then the platform earns trust. If a link is missing, 'next best action' becomes just the next unproven claim.