Search for “see who someone recently followed on Instagram” and you will find two kinds of results: tools that quietly work, and promises that cannot possibly be true. Telling them apart is easy once you understand the engineering, because the entire category rests on one simple mechanism — and everything that mechanism cannot do is, by definition, a lie when advertised.
The gap in the platform
Instagram exposes a public account’s follower and following lists to any viewer, but as an unordered set: the app sorts the lists by its own relevance ranking, not by recency, and attaches no timestamps. So the data needed to answer “what changed?” is technically public, while the ordering that would make it useful is withheld. That gap is the product category.
The mechanism: collect, version, diff
Activity trackers close the gap with a two-part pipeline.
Part one is scheduled collection. The tool retrieves a public profile’s visible data — follower list, following list, counts, posts — the same class of technique described in any overview of web scraping: automated retrieval of information a human could view in a browser. The engineering work here is mostly about doing it politely and reliably at scale: respecting rate limits, handling the platform’s frequent markup changes, and retrying gracefully when requests fail.
Part two is versioned comparison. Each collection run is stored as a dated snapshot, and consecutive snapshots are diffed — the classic change detection problem, applied to set membership. An account present in today’s following list but absent from last week’s is a new follow; the reverse is an unfollow. Stack enough dated diffs and you have reconstructed the chronology Instagram never provided: not because the tool saw the follow happen, but because it can prove the follow occurred between two known dates.
This is why trackers such as IGDetective present changes as a dated feed rather than a raw list — the feed is the diff history. It is also why the first scan of a new account shows nothing “recent”: with a single snapshot there is nothing to compare yet. A tracker that claims to show you what an account followed before monitoring began is describing data it cannot have.
Reading the architecture tells you the limits
Once you see the pipeline, the honest boundaries of the category become obvious, and they are worth spelling out because marketing in this space routinely ignores them.
Private accounts are out of reach. The collection step can only retrieve what a logged-out viewer could see. A private account’s lists are not public data; no snapshot pipeline can capture them. Any service claiming otherwise is either fraudulent or doing something no user should be part of.
Direct messages are out of reach for the same reason, categorically. There is no public surface from which to collect them.
Granularity is bounded by scan frequency. A tool that scans daily can date a change to a day, not a minute. Vendors advertising real-time alerts are describing short scan intervals, which is legitimate — but precision beyond the scan interval is invented.
And the user’s own credentials should never enter the picture. A properly built public-data tool has no use for your Instagram password — IGDetective, for instance, takes a public username as input and nothing else. A “tracker” that wants you to log in with your own account credentials is asking for more access than the job requires, which is the single clearest red flag in the category.
Why the honest version is still valuable
Stripped of impossible promises, what remains is genuinely useful: a dated, provable record of how a public account’s connections and content change over time. For analysts, that turns anecdote into evidence — bought-looking follower spikes become visible as vertical jumps between snapshots; a quiet association becomes a dated fact rather than a hazy memory. The value does not come from seeing anything secret. It comes from remembering the public record better than the platform chooses to, which is an old and durable species of software value.
The next time a tool in this space makes you an offer, run it against the pipeline: could scheduled collection of public data plus diffing produce this feature? If yes, it is plausibly real. If no, you have your answer.
Sources:
- Wikipedia — Web scraping: https://en.wikipedia.org/wiki/Web_scraping
- Wikipedia — Change detection: https://en.wikipedia.org/wiki/Change_detection
