We prove ℓ p Z d bounds for p ∈ (1, ∞), of r-variations r ∈ (2, ∞), for discrete averaging operators and truncated singular integrals of Radon type. We shall present a new powerful method which allows us to deal with these operators in a unified way and obtain the range of parameters of p and r which coincide with the ranges of their continuous counterparts.where B t = {x ∈ Z k : |x| ≤ t} and t > 0. We will be also interested in discrete truncated singular integrals.Assume that K ∈ C 1 R k \ {0} is a Calderón-Zygmund kernel satisfying the differential inequality |y| k |K(y)| + |y| k+1 |∇K(y)| ≤ 1 for all y ∈ R k with |y| ≥ 1. We also impose the following cancellation condition B λ 2 \B λ 1 K(y) dy = 0 (1.2) for every 0 < λ 1 ≤ λ 2 where B λ is the Euclidean ball in R k centered at the origin with radius λ > 0. Define, for a finitely supported function f : Z d0 → C, the truncated singular Radon transformsThe basic aim of this paper is to strengthen the ℓ p Z d0 boundedness, p ∈ (1, ∞), of maximal functions corresponding to operators (1.1) and (1.3), which have been recently proven in [15], and provide sharp r-variational bounds in the full range of exponents.Recall that for any r ∈ [1, ∞) the r-variational seminorm V r of a sequence a n (x) : n ∈ N of complexvalued functions is defined byThe main results of this article are the following theorems.Theorem A. For every p ∈ (1, ∞) and r ∈ (2, ∞) there is C p,r > 0 such that for all f ∈ ℓ p Z d0Moreover, the constant C p,r ≤ C p r r−2 for some C p > 0 which is independent of the coefficients of the polynomial mapping P.[1], he was able to circumvent this issue for the operators A P N , in the one dimensional case k = d 0 = 1, by controlling their oscillation seminorm. Given a lacunary sequence (n j : j ∈ N), the oscillation seminorm for a sequence a n : n ∈ N of complex numbers is defined by O J a n : n ∈ N = J j=1 sup nj 0 and c < 1/2 such that for all J ∈ NVariational estimates have been the subject of many papers, see [9,10,11,16,22] and the references therein. Our motivation to study r-variational seminorms is threefold. Firstly, for any sequence of functions a n (x) : n ∈ N , if for some 1 ≤ r < ∞ V r a n (x) : n ∈ N < ∞