Practical tips for presenting data-led research at conferences

The archived IASSIST 2017 programme captures a valuable conference truth: researchers often arrive with sophisticated evidence and leave the audience trying to reconstruct the point. Held at the University of Kansas in Lawrence from 23 to 26 May, the event brought big data, deep learning, digital agriculture and global food security into a shared programme. Its plenary sessions, schedules and presentation listings illustrate the breadth of work that can sit behind a single research question. The task is to make that work legible without flattening its complexity.

Useful data communication acts as a bridge between analysis and application. A strong presentation enables attendees to identify the main finding, understand the evidence, recognise the uncertainty and explain why the result matters. It does not require displaying every model output or reproducing the full methodology. Instead, it requires a deliberate sequence that respects both scientific rigour and the audience’s time.

Build the argument before choosing the charts

Begin with the decision your research should inform. A useful starting sentence has four parts: the problem, the evidence, the finding and its significance. For example, a digital agriculture study might connect unreliable irrigation forecasts with satellite-derived soil-moisture measures, then show how the combined data could support earlier irrigation decisions. This gives each visual a job and prevents a sequence of disconnected results from taking over the talk.

The theme of IASSIST 2017, “Data in the Middle: The Common Language of Research,” captures the need for that common language. Findings from different disciplines rarely become useful simply because they share a dataset. Presenters must translate statistical patterns, model behaviour and domain assumptions into a coherent analytical narrative. The audience should be able to follow the reasoning even when it does not share the presenter’s technical vocabulary.

An Australian research setting adds useful dimensions to this argument. A project relevant to dairy farms near Toowoomba may have different adoption constraints from one supporting horticulture in Atherton, while a broad policy audience in Canberra may focus on national implications rather than regional operations. Naming those contexts helps the audience see the relevance of the evidence. It also encourages discussion about infrastructure, skills, costs and implementation rather than treating a statistically significant result as a finished solution.

Make every visual understandable at a glance

Every visual should answer a question that has already been raised in the presentation. Headings such as “Results” or “Model comparison” provide little guidance; a descriptive heading such as “Irrigation forecasts reduced unnecessary watering by 18 per cent” tells the audience what to look for. Use position, colour and annotation to establish a clear visual hierarchy. Keep decorative elements, dense legends and automatic gridlines out of the main analytical path.

Quantitative uncertainty belongs beside the finding it qualifies. Include sample sizes, confidence intervals, missing-data rates or model-performance ranges when they affect interpretation. Showing an error band beside a forecast or a range around an estimated effect makes the strength of the evidence easier to judge. It also reduces the temptation to treat every measured difference as equally reliable. Colour should support this information rather than carry it alone, particularly in rooms with poor projection or for attendees with different colour-vision needs.

Complex analyses usually need translation into a small number of purposeful visuals. A big-data pipeline might be represented as a flow showing where errors, delays or bias enter. A deep-learning model might be explained through representative examples, failure cases and changes in performance rather than a wall of architecture diagrams. For digital agriculture, a regional map can show where a model performs well, while field-level examples reveal what those patterns mean. Captions should make charts understandable when projected at a distance, and essential context should remain visible without relying on tiny footnotes.

Show methods, uncertainty and research integrity

Methods should appear in the amount needed to support confidence. Explain the sample, relevant variables, comparison group, baseline and validation approach. If a causal claim is being made, identify the design that supports it and acknowledge important assumptions. Technical precision remains valuable, yet a lengthy methodological explanation can bury the result unless each detail helps the audience assess the evidence.

Data provenance and reproducibility deserve explicit attention. Record where each dataset came from, which transformations were applied, how missing observations were handled and which software versions produced the result. Code, protocols or synthetic examples may be shared when access to sensitive data is restricted. This practice allows other researchers to inspect the workflow and helps participants distinguish a reproducible finding from a result that cannot be checked. A concise supplement can carry details that would consume too much time during the main presentation.

Research integrity also involves describing whose data are represented and whose decisions may be affected. This matters in global food security work, where an apparently precise indicator can conceal regional differences, data access gaps or conflicting priorities. Australian collaborators may include universities, government agencies, technology suppliers, producer groups and regional communities, but no single partner speaks for the entire system. Present claims at the confidence the evidence supports and distinguish observed associations from tested mechanisms or projected outcomes. Candid limitations can strengthen a presentation by showing mature judgment rather than weakness.

Deliver for the room, not the script

Many Australian professional conversations value directness, friendliness and practicality. Speakers often earn engagement by explaining the research in plain language, acknowledging evidence that does not fit neatly and responding to a concern with a concrete example. That style need not become informal to the point of losing scientific precision. A concise delivery, clear definitions and a willingness to say “the evidence does not settle that point” can make technical material more accessible.

Timing is part of the message. Rehearse aloud, mark the point where the central finding appears and leave enough time to explain it before reaching the final slide. A useful rule is to make the main result understandable within the first few minutes, while reserving specialist detail for the questioning period. During questions, answer the concern directly, refer to the relevant evidence and state when further analysis is needed. Repeating the central finding after a difficult question can also redirect attention toward the study’s actual contribution.

Conference operations influence how well an idea lands. The archived IASSIST site includes programme information, plenary details, check-in guidance, transportation and local recommendations because practical access shapes participation. Australian presenters face comparable issues with venue signage, travel across time zones, room changes and hybrid participation. Arriving early, checking the display connection and carrying both presentation formats and an accessible PDF can prevent avoidable disruption. The talk begins before the first slide appears.

Match the message to the conference format

A plenary requires a shared narrative, while a parallel session permits greater methodological depth. Posters support repeated conversations, and workshops depend on participation and application. The research may remain the same, yet its evidence must be reorganised around the purpose of the session. A single dense deck placed into several formats usually serves none of them well. Instead, design the central argument first and adapt its supporting material to the expected level of audience expertise and time.

Australian translation pathways also shape the presentation. Cooperative Research Centres, competitive grants, government programmes and industry partnerships often require researchers to show feasibility, commercial or social value, and a credible route from pilot to adoption. This can sharpen the final portion of a talk. Rather than ending with “more research is needed,” identify the next test, implementation constraint or decision that the current evidence can support.

Format	Primary aim	Effective evidence mix	Delivery rhythm	Common risk
Plenary	Establish relevance for a mixed audience	One central finding, one explanatory visual and a concise limitation	Broad narrative with a memorable pause	Abstract conclusions without enough context
Parallel talk	Explain a defined study or method	Two or three purposeful visuals with selective technical detail	Technical account followed by implications	Presenting every result in sequence
Poster	Generate discussion around a specific problem	Legible workflow, headline result and representative evidence	Short walk-around repeated for several groups	Text too small for independent reading
Industry workshop	Test application and next steps	Workflow evidence, constraints and implementation measures	Short briefing, structured discussion and action sequence	Turning preliminary results into promises

The strongest presentation keeps the same standard across every format: a clear claim, inspectable evidence, honest uncertainty and a conclusion proportionate to the research. The presenter’s role is not merely to display data, but to help others interpret what that data can responsibly support.