Skip to main content
Adrià García
Menu

Evaluation tool

Prepare for a MarTech platform demo

Turn sales claims into questions that can be answered with architecture, evidence, and specific ownership.

Choose the claims that appear in the proposal or demo. Review the reasoning, add the relevant ones, and generate a briefing to copy or print.

Your selection remains in the browser. No answers are sent and the briefing is not stored.

Data and identity

Unified customer view

The claim

One profile brings together all customer data and makes it available to every channel.

Questions that force specificity

  1. How does the platform relate records that do not share a stable identifier, and which part remains unresolved?
  2. Which rule decides what wins when two systems disagree on an attribute, and who can change that rule?
  3. How is an incorrect merge reversed, and what happens to audiences or activations that already used that profile?
Show the reasoning

The decision underneath

A unified view depends on shared identifiers, precedence rules, and processes for correcting bad merges. A profile screen does not prove that those decisions work with real data.

Evidence I would request

A test using your own sample with conflicts, incomplete identities, and one merge that must be reversed.

Risk signal

The answer is limited to a clean profile screen and does not explain precedence, traceability, or identity separation.

Data and identity

End-to-end real time

The claim

Events update the profile and change the customer experience in real time.

Questions that force specificity

  1. Which latency do you measure separately across event, profile, audience, decision, and channel delivery?
  2. Which sources and destinations still use batch processing although the product is presented as real time?
  3. How do those latencies change under peak volume, and what happens when one stage stops responding?
Show the reasoning

The decision underneath

Ingestion, profile updates, audience evaluation, decisioning, and delivery are separate stages. A platform may be fast in one while retaining batches or queues in another.

Evidence I would request

A trace with timestamps for every stage, executed with a use case and volume comparable to the client.

Risk signal

The demonstration shows a fast-refreshing dashboard but does not measure when the channel actually changes.

Decisioning and channels

Intelligent decisioning

The claim

Artificial intelligence automatically selects the most appropriate action for each customer.

Questions that force specificity

  1. Who defines the objective the system optimizes, and how are incompatible objectives across teams resolved?
  2. Which rules constrain the model, and which fallback applies when there is not enough customer information?
  3. Can the team reconstruct why one action won over the others for a specific customer and moment?
Show the reasoning

The decision underneath

A model may rank candidates, but someone still defines objectives, eligibility, limits, priorities, and fallbacks when signals are missing. That ownership determines actual behavior.

Evidence I would request

A scenario with several competing actions, visible constraints, and a complete explanation of the final decision.

Risk signal

The feature generates content or recommendations but does not arbitrate actions or preserve a decision record.

Decisioning and channels

Cross-channel coordination

The claim

Email, push, web, app, and messaging operate as one coordinated experience.

Questions that force specificity

  1. Can a conversion in one channel stop or change a communication already scheduled in another, and with which latency?
  2. Are contact pressure and priority computed globally or configured separately in every channel?
  3. Where can the full sequence of decisions and contacts for one person be inspected without reconstructing it manually?
Show the reasoning

The decision underneath

Sending through several channels does not imply shared state. Coordination requires a response, conversion, or accumulated pressure to change pending decisions in other channels.

Evidence I would request

A demonstration where a real action in one channel visibly changes the behavior of another.

Risk signal

All channels start from the same trigger but then execute independent flows without shared state.

Integration and architecture

Native integrations

The claim

The platform connects natively with the tools the organization already uses.

Questions that force specificity

  1. Who maintains this connector, which version does it support, and how is a breaking change communicated?
  2. Which objects, operations, and directions does it actually cover, and which require additional development?
  3. How does it detect, retry, and reconcile lost or rejected events without duplicating data?
Show the reasoning

The decision underneath

The word native may describe anything from a maintained and observable connector to a generic template. Object coverage, direction, frequency, and error recovery are what matter.

Evidence I would request

Scope documentation, version history, and a failure-and-recovery test with the system that must be connected.

Risk signal

The connector only exposes a generic call, has no clear owner, or provides no operational traceability.

Integration and architecture

Composable architecture and exit

The claim

Every capability can be replaced without vendor dependency or a complex migration.

Questions that force specificity

  1. Which modules can be replaced independently, and which capabilities stop working when they are removed?
  2. In which format are profiles, audiences, rules, journeys, and traces exported when the contract ends?
  3. Which proprietary components or professional services remain mandatory after one part is replaced?
Show the reasoning

The decision underneath

Technical modularity does not guarantee portability of data, configuration, or operations. The useful test is what can be removed and which work remains when a component changes.

Evidence I would request

A dependency diagram and exit procedure covering data, configuration, integrations, and operations.

Risk signal

The suite exposes APIs, but no core capability can be replaced or exported in a usable form.

Integration and architecture

Activation without copying data

The claim

The platform activates directly from the warehouse and does not create another copy of the profile.

Questions that force specificity

  1. Which data remains outside the warehouse, for how long, and for which specific function?
  2. Which queries does the platform run, and how do they affect latency, concurrency, and cost at the expected volume?
  3. Which activations continue to work if warehouse access is removed or the warehouse becomes unavailable?
Show the reasoning

The decision underneath

Querying at source, synchronizing an audience, and keeping a temporary cache are different patterns. Each changes freshness, cost, availability, and governance scope.

Evidence I would request

A complete data flow with storage, caches, refresh frequency, and queries explicitly identified.

Risk signal

The supposed direct activation actually depends on a persistent copy refreshed on a schedule.

Consent and governance

Consent enforced across destinations

The claim

A consent change is automatically enforced across every connected channel and platform.

  1. Which purposes and channels are represented separately, and where is each signal translated for destinations?
  2. How long does withdrawal take to stop new collection, audiences, exports, and sends already prepared?
  3. What happens to data shared before the change, and which actions can propagate to third parties?
Show the reasoning

The decision underneath

The CMP may record the choice, but every tag, pipeline, CDP, and destination must receive and interpret it. Actual coverage depends on propagation, mappings, and work already queued.

Evidence I would request

A test that changes consent and follows the signal into several destinations, including queues and existing activations.

Risk signal

Only banner or email blocking is demonstrated, without testing pipelines, audiences, and external exports.

Operations and autonomy

No-code business autonomy

The claim

Marketing teams can build and publish use cases without depending on technology teams.

Questions that force specificity

  1. Which non-trivial use case does the business operate without technical support, and which knowledge is required to maintain it?
  2. How are changes reviewed, tested, versioned, and promoted before affecting a production audience or journey?
  3. Who diagnoses an empty segment, missing event, or conflicting rule, and with which tools?
Show the reasoning

The decision underneath

Not writing code does not remove the need to understand schemas, events, consent, testing, and promotion across environments. Autonomy depends on the operating model, not only the interface.

Evidence I would request

A complete path from change to production with permissions, validation, rollback, and responsibilities made visible.

Risk signal

The interface is visual, but any meaningful change requires consulting and only support can resolve failures.

Operations and autonomy

Observability and recovery

The claim

The platform shows what happened to each customer and supports recovery from any operational failure.

Questions that force specificity

  1. Can the data, rule, audience, and journey behind one action for one person be reconstructed?
  2. How long are those traces retained, and do they remain interpretable after schemas or journeys change?
  3. Which failures support retry or replay, how are duplicate effects prevented, and who performs recovery?
Show the reasoning

The decision underneath

Aggregate dashboards support measurement but do not explain an individual decision or replay a failed stage. Operations require traces, retention, alerts, and recovery procedures.

Evidence I would request

A live investigation of a prepared incident, from alert through cause and controlled recovery.

Risk signal

Only aggregate metrics are available, and explaining an individual case requires opening a vendor support request.

Evaluation tool

Your evaluation briefing

Include only the claims relevant to the meeting. Questions are grouped so they can be used during the demo.

No claims have been added yet. Select at least one to prepare the briefing.

Need to compare the answers?

I can turn requirements, vendor answers, and operating constraints into a comparison and a reasoned recommendation.

View the platform selection service → Email me →