Great situational awareness on your part — identifying the exact problem (losing the thread when skimming, and running out of time when reading fully) is actually the harder half of fixing a DI timing issue. Here's what worked for me and what I've seen work for others with a similar profile.
Your core challenge isn't skimming vs. reading — it's that you haven't yet built a structured reading protocol for each DI question type. The Data Insights section has four distinct types (Data Sufficiency, Graphs & Tables, Multi-Source Reasoning, Two-Part Analysis), and each one demands a different pacing approach.
For your 15-day sprint, here's what I'd focus on:
For Graphs & Tables specifically — stop reading the data before you read the question. Read the question stem first, identify what you're looking for (a comparison? a maximum? a calculation?), and only then pull the relevant number from the chart. This alone cuts 30–45 seconds per question for most people.
For Multi-Source Reasoning — skim the tab titles and the first sentence of each source tab, then read only the tab that's relevant to each question. Do not try to read all three tabs front to back. This is the biggest time sink in DI.
For Two-Part Analysis — don't fill in column 1 then come back for column 2. Work the logic once and lock in both answers simultaneously. Going back wastes 60+ seconds per question.
On the "running out of time" front: with DI78 (83rd percentile), you're not making careless errors — you're getting harder questions and spending too long on them. The fix is strict time boxing: give yourself 2:30 per question maximum, and when you hit that threshold, make your best guess and move. Leaving 3-4 questions blank is a much bigger penalty than a wrong guess.
Given your Q83 and V85, your DI is genuinely the bottleneck to 700+. A DI score of 82–85 would likely push you to the 675–700 range. Totally achievable in 15 days with this kind of targeted work.
You've got this.
— Kavya | GMAT Focus 725 (99th percentile)