diff --git a/CHANGELOG.md b/CHANGELOG.md
index 923de14..e207d47 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -82,6 +82,11 @@ The format is based on [Keep a Changelog].
   `selectrum-secondary-highlight` are now a part of
   `selectrum-prescient.el` rather than Selectrum proper ([#94])
 
+* The user option `prescient-sort-full-matches-first` was added. If
+  non-nil, candidates that are fully matched are sorted before
+  partially matched candidates, though all candidates still follow the
+  order of recency, frequency, and length. See [#95].
+
 [#66]: https://github.com/raxod502/prescient.el/pull/66
 [#67]: https://github.com/raxod502/prescient.el/pull/67
 [#70]: https://github.com/raxod502/prescient.el/pull/70
@@ -89,6 +94,7 @@ The format is based on [Keep a Changelog].
 [#76]: https://github.com/raxod502/prescient.el/pull/76
 [#77]: https://github.com/raxod502/prescient.el/pull/77
 [#94]: https://github.com/raxod502/prescient.el/pull/94
+[#95]: https://github.com/raxod502/prescient.el/pull/95
 
 ## 5.0 (release 2020-07-16)
 ### Breaking changes
diff --git a/README.md b/README.md
index 03c70d0..8790e48 100644
--- a/README.md
+++ b/README.md
@@ -97,6 +97,10 @@ different one by customizing `prescient-filter-method`.
   to define your own custom filter methods, and use them by adding the
   appropriate symbol to `prescient-filter-method`.
 
+* `prescient-sort-full-matches-first`: Whether `prescient.el` sorts
+  candidates that are fully matched before candidates that are
+  partially matched.
+
 ### Company-specific
 
 * `company-prescient-sort-length-enable`: By default, the standard
diff --git a/prescient.el b/prescient.el
index ef8c7af..8d79c02 100644
--- a/prescient.el
+++ b/prescient.el
@@ -184,6 +184,15 @@ after the cache data is updated by `prescient-remember' when
 `prescient-persist-mode' is activated."
   :type 'boolean)
 
+(defcustom prescient-sort-full-matches-first nil
+  "Whether to sort fully matched candidates before others.
+
+Prescient can sort by recency, frequency, and candidate length.
+With this option, fully matched candidates will be sorted before
+partially matched candidates, but candidates in each group will
+still be sorted like normal."
+  :type 'boolean)
+
 ;;;; Caches
 
 (defvar prescient--history (make-hash-table :test 'equal)
@@ -525,18 +534,27 @@ must match each subquery, either using substring or initialism
 matching. Discard any that do not, and return the resulting list.
 Do not modify CANDIDATES; always make a new copy of the list."
   (let ((regexps (prescient-filter-regexps query))
-        (results nil))
+        (results nil)
+        (prioritized-results nil))
     (save-match-data
       ;; Use named block in case somebody loads `cl' accidentally
-      ;; which causes `dolist' to turn into `cl-dolist' which creates
-      ;; a nil block implicitly.
+      ;; which causes `dolist' to turn into `cl-dolist' which
+      ;; creates a nil block implicitly.
       (dolist (candidate candidates)
         (cl-block done
-          (dolist (regexp regexps)
-            (unless (string-match regexp candidate)
-              (cl-return-from done)))
-          (push candidate results)))
-      (nreverse results))))
+          (let ((fully-matched nil))
+            (dolist (regexp regexps)
+              (unless (string-match regexp candidate)
+                (cl-return-from done))
+              (when (and
+                     prescient-sort-full-matches-first
+                     (equal (length candidate)
+                            (length (match-string 0 candidate))))
+                (setq fully-matched t)))
+            (if fully-matched
+                (push candidate prioritized-results)
+              (push candidate results)))))
+      (nconc (nreverse prioritized-results) (nreverse results)))))
 
 (defmacro prescient--sort-compare ()
   "Hack used to cause the byte-compiler to produce faster code.